“All of a sudden you can program the world” — it’s the continuation of the software eating the world thesis we put out over five years ago, and of the trajectory of past and current technology shifts. So what are those shifts? What tech trends and platforms do we find most interesting on the heels of raising our fifth fund? Are we just building on and extending existing platforms though, or will there be new platforms; and if so, what will they be? Well, distributed systems for one…

This episode of the a16z Podcast covers all things distributed systems — encompassing cloud and SaaS; A.I., machine learning, deep learning; and quantum computing — to the role of hardware; future interfaces; and data, big and small. Podcast guests Marc Andreessen and Ben Horowitz (in conversation with Scott Kupor and Sonal Chokshi) also share the one piece of advice from a management and go-to-market perspective that all founders should know. And finally, why simulations matter… and what do we make of our current reality if we are all really living in a simulation as Elon Musk believes?

Show Notes

• How advances in hardware and reduced prices are pushing A.I. and other technological advancements [0:27]
• The current state of A.I. and where it’s headed [8:39]
• Real-world applications for technology (life sciences, SaaS, and company creation) [20:21]
• The firm’s philosophy around team-building [32:59] and advice for founders [37:50]

Transcript

Sonal: Hi, everyone. Welcome to the “a16z Podcast.” I am Sonal. And I’m here today with a special podcast we have on the heels of announcing our fifth fund for Andreessen Horowitz. And we thought we’d talk more broadly about what’s changed between the first fund and now and, more importantly, some of the technology trends and trends we’re seeing with founders. And to have that conversation with us, we have our co-founders, Marc Andreessen and Ben Horowitz, and our managing partner, Scott Kupor. Welcome, guys.

Scott: Hey.

Advances in hardware

Sonal: Okay. So let’s just kick things off. One of the things that I want to understand is that it’s been — since fund one, which is, what, six, seven years ago?

Marc: Seven.

Ben: Seven years.

Sonal: Yeah, seven years ago. A lot’s changed in seven years, and I’ve actually heard you argue, Marc, that things have accelerated in that time period, more so than previous decades before. So what do you guys think are the biggest shifts now that are important to us in this newest fund, and what changed in that period, like, the biggest things?

Ben: So, in fund one, when we started, we thought that our timing was really good, despite the fact that I think the world thought our timing was really bad in starting a new venture capital fund. And the reason why we thought that was that there were three gigantic new platforms hitting all at the same time, which was kind of unprecedented in the history of technology. One was mobile, the second was social, and the third was cloud. And that really proved out, through the course of the early history, that the applications on top of those — particularly mobile and cloud — were just spectacular. And I think we’re coming a little bit to the end of the first phase of, you know, some of the obvious applications that could be built on those things, and we’re moving into some new areas.

Marc: Yeah. So, let me go kinda to the foundations. So, there’s different ways of looking at it. The foundational levels — one is Moore’s law has really flipped, and this actually has happened. I think this actually has happened over the last seven or eight years, actually, almost exactly over the life of the fund. Which is, you know, for many, many years, Moore’s law was a process of the chip industry, bringing out a new chip every year and a half, that was twice as fast as the last one at the same price. And that continued for 40, 50 years, and that’s, by the way, what resulted in everything from mainframes, mini-computers, PCs, and then smartphones. About, you know, 7, 8, 9, 10 years ago, that process actually started to come to an end the way that it had worked up until then. So, chips have kind of topped out at a speed of about three gigahertz, and a lot of people have said, therefore, like, progress in the tech industry is gonna stall out, because the chips aren’t getting faster. I think what’s actually happened is, Moore’s law has now flipped.

The dynamic now, instead of increased performance, is reduced cost. You now have this dynamic where, every year, a year and a half, chip companies come out with a chip that’s just as fast but half the price. And so, this is the, sort of, just this massive deflationary force, I think, in the technology world, and I actually also suspect in the economy more broadly, where, basically, computing is just becoming free. Basically, what we do in this business is we just kind of chart out the graphs and then just kind of assume, at some point, you’re gonna get to the end state, and the end state is gonna be the chips are gonna be free. Which means chips will be embedded in everything. You’ll be able to use chips for, literally, everything. And we’ve never lived in a world before where you can do that. So, that’s the first one.

Second one is just the obvious implication from that, which is, all those chips will be on the network, right? So, all those chips will be connected to the internet. They’ll all be on Wi-Fi, or mobile carrier networks, or wired networks, or whatever, but they’ll all fundamentally be on the internet, you know. That’s something that’s now happening at a very rapid pace. And then the third is the continuation of the piece that I wrote, actually, five years ago, which was called “Software Eats the World,” which basically just [says], if you’re gonna live in a world in which there’s gonna be a chip in every physical object, and if you live in a world in which every physical object, therefore, is going to be networked — it’s gonna be smart because it has a chip, and it’s gonna be connected to the network — then basically, you can then program the world. You can basically write software that applies to the entire world. So, you can write software that, all of a sudden, applies to all cars, or you can write software that applies to all, you know, everything flying in the sky, or you can write software that applies to all buildings, or you can write software that applies to, you know, all homes, or all businesses, or whatever, all factories.

And so, all of a sudden, you can program the world. That’s really just starting, and I think a lot of the — there’s a number of things that make the entrepreneurs we’re seeing these days, in many ways, more interesting and more aggressive than entrepreneurs we’ve seen in the past. And part of it is, they just assume — if there’s something to be done in the world, there must be a way to write software to be able to do it. That’s at a new level of power, sophistication. It’s a new scope of what the tech industry can do. The consequence of that for us, as a fund, is that we find ourselves evaluating business plans and funding companies that are in markets where, I think, seven or eight years ago, we would have never anticipated operating.

Scott: So, Marc, does that mean that there’s no new innovation in platforms themselves, and everything — all the innovation will be applications that ride on that existing infrastructure, or do you think there’s also the opportunity to build a new platform, even given some of those trends?

Marc: I think there are new platforms, and I think there will be new platforms. I just think they’ll be different kinds of platforms than we’ve had in the past. The idea of a platform in the tech industry, as you know, up until, you know, 5 or 10 years ago, was there is a new chip that has new capabilities, is faster, and then, therefore, you build a new operating system for it. And that might be Windows, or it might be, you know, might be iOS, or whatever it is. The platforms that we’re seeing getting built these days are distributed systems. So, scale-out systems, sort of, being built on a chip necessarily with new unique capabilities. They are platforms that are getting built across lots of chips. And so, in computer science terms, they’re distributed systems. Cloud is one of the first examples, right? So anybody who uses AWS can now go on and can program an application on AWS that will run across 20,000 computers. And they can run it for an hour, and it’ll cost, you know, 50 bucks. And that’s a kind of platform that did not exist before.

And, by the way, there are many specific elements to that. So, for example, we’ve seen the rise of, in that category, we’ve seen the rise of Hadoop, and now the rise of Spark for distributed data processing. We’ve seen — in financial technology, we’ve seen the rise of Bitcoin and cryptocurrency, which is, literally, a distributed platform, you know, for currency and for exchanging value. And now, we’re seeing the emergence of a major new platform, which is A.I. — machine learning and deep learning, which is inherently — the great thing about machine learning and deep learning is they’re inherently parallelizable. They can run across many chips, and they get very powerful as you do that. And you can do things in A.I. today as a consequence of being able to run across many chips that you just couldn’t even envision doing 5 or 10 years ago.

Sonal: So, let’s talk about the rise of the GPU as part of this next platform chip. I mean, I think the biggest surprise people have had is that this is the graphical processor unit, which is something that was developed in the gaming industry for really high-resolution graphics processing, and is now finding, I guess, unexpected — is it a surprise to us that it’s finding uses in these new platforms, like VR, AR, deep learning?

Marc: It’s actually, interestingly — it’s a new application of an old idea. Back when I was getting started 30 years ago, working in physics labs, if you wanted to run just a normal program, you just buy a normal computer and run the program. But if you wanted to run a program — many physics simulations had this property where you would want to run a very large number of calculations in parallel, right. So, you could basically divide up a problem, simulating anything from a black hole or to different kinds of biological simulations. You could basically write these algorithms in a way that you could basically parcel the problem into many different pieces and then run them all in parallel. There was actually, in the old days, there was actually a whole industry of what were called vector processors, which were, literally, these kind of sidecar computers that you would buy and you would hook up to your main computer, and they would let you run these parallel problems much faster.

And so, literally, 30 years later, the GPU is — it’s basically a vector processor. It’s basically a sidecar processor that sits along a CPU and runs these parallel problems much faster. And graphics are a natural application of that, but as it turns out, graphics aren’t the only application.

Ben: Yeah. Actually, interestingly, and I was at a company making one of these called Silicon Graphics — and the applications then were, as Marc was saying, a lot of physics applications, computational fluid dynamics, and simulating, you know, flight simulation, and all these kinds of things that are hard physics to calculate. When you go into the virtual world, and you’re simulating the physics of the real world, guess what? You need the exact same processor to do it. So, it’s a super logical conclusion to what’s been going on, but I think we’re also in the world of big data, seeing kind of more reasons to do just lots of math in parallel. And so, it’s an exciting application.

Marc: Yeah. You talk about platforms — one of the really interesting hardware platforms that’s emerging right now is Nvidia, which is a very well established public chip company, but very successful, to your point, doing graphics chips for a very long time — has become seemingly overnight — it’s really, of course, the result of years of work — but seemingly overnight has become the market leader in both not just GPUs but also in chips being used for A.I.. And it’s basically extensions of the GPU technology. And we see this overriding theme, which is kind of an amazing thing, which is, basically, every sharp A.I. software entrepreneur that comes in here is now building on top of Nvidia’s chips. Which is, of course, a very different outcome than entrepreneurs of previous years, who would have built other kinds of programs primarily on top of Intel chips.

Advances in A.I.

Scott: We’ve mentioned A.I. and machine learning a couple of times here. And one of the interesting things, at least, that I think we see in the industry is, at the same time we’ve got startups doing it, we also see some of the very largest established players investing significantly in A.I. and machine learning. So, certainly Facebook, and Google, Apple, and others are obviously building big operations. How do you think about the universe from an investment perspective? What are the kinds of things that actually lend themselves well to startup opportunities in the A.I. space, versus things that actually might make sense kind of living inside of one of the larger companies, like a Facebook or a Google?

Ben: Yeah. So, you know, A.I. is extremely broad, and I think one of the challenges that people have with it is they try to paint it as a narrower thing than it is, but one can think of it as an entirely new way to write a computer program. And so, then, it’s applicable to, you know, the universe of problems. So, there are things that advantage a big company. You know, if you’re building A.I. to analyze consumer internet data, like, that’s hard to take Google on at that. They do have an awful lot of data. And you know, Facebook, you know, with A.I., computing power matters and the dataset matters. Having said that, there are a lot of areas where nobody has any data yet, in the areas of healthcare and the areas of autonomy. So, you know, there’s lots and lots of opportunities, and you know, there’s also interesting ideas about, “Well, is there a better user interface than the smartphone using A.I. techniques? And then, what is the form of that?”

Sonal: What do you mean by that, when you say there’s a better user interface?

Ben: Well, yeah, if you think about a smartphone, it was kind of an advance over what we used to call the WIMP interface. Windows, icons, what was it?

Marc: Menus.

Ben: Menus.

Sonal: Oh. What was the P?

Marc: Pointer.

Ben: Pointer, yeah.

Sonal: Oh, pointer, right.

Ben: Which, you know, was, like, a big advance over the text-based interface of DOS. And then, you know, the smartphone with the touch interface, it was more of a direct manipulation — was an advance over that. And so, you go, “Okay, well, but that’s not actually what people do in life,” right? It’s, anthropologically — it’s a backward step, in terms of the natural interface that we’ve become accustomed to, like, for example, natural language. With A.I., you get into a world where things like natural language, and natural gestures, and so forth, become much more plausible. So, there’s, you know, potentially an opportunity to build interfaces for things that you couldn’t before. I mean, I think there’s one, like, really interesting thing, which I’m sure — and I know that Google, and Apple, and all the giant companies are very focused on — which is, how do you replace the current set of user interfaces with it? But there’s another dimension, which is, what are all the applications that you just couldn’t have before, because you couldn’t build a workable user interface for it. And A.I. seems very promising in those areas.

Sonal: You didn’t mention Amazon, which is sort of the stealth player here, with Echo and Alexa. I mean, really, Trojan Horse of the home.

Ben: Well, you know, in a way, they’ve got an interesting advantage in that they’re not tied to the last generation of user interfaces, so that they don’t have to pay the strategy tax for shoehorning in their A.I. into, say, the iPhone, and that’s something.

Marc: Yeah, that’s worth pointing out. There’s sort of two, kind of, classic rules of thumb in this industry. One is for major new advances, especially in things like interfaces, if you don’t own a platform, you can’t do them. And so, the assumption, I think, had been up, until recently, you know, that it would have to be Google or Apple that does these kinds of natural language or interface advances, because they own iOS and Android. The other rule, of course, is the exact opposite rule, which is the one that Ben mentioned, which is the problem that big established companies get into — is what he referred to as the strategy tax, which is, basically, big companies with existing agendas have to, sort of, fit their next thing into their existing agenda, and they often compromise it in the process.

And so, it’s sort of this ironic twist of fate that Amazon has, all of a sudden, taken the lead from Google and Apple, even though Amazon, you know, famously flopped with their phone, right, which is sort of the obvious place where you have a voice interface. It didn’t matter because they came out with this new product, which was, basically, the speaker, the smart speaker called Echo, and the fact that, all of a sudden, Amazon didn’t have a phone, all of a sudden, became an advantage because they could just do the clean actual breakthrough product without worrying about tying it into the existing strategy.

Sonal: Right. And those are all still big companies, though. I’m not really hearing where startups can really play in this space, especially when you are describing this huge data network effect that all these big companies have.

Marc: A year ago, we would have probably been sitting here and say that A.I. was going to be likely would be a domain of big companies, because of this sort of thing of, like, “Okay, only big companies can afford the very large number of engineers that are required to do A.I., only big companies can afford the amount of hardware required to do A.I., and then only big companies can get the giant datasets required to do A.I..” In the last 12 months, what we’ve seen, basically, is all three of those changing very fast, and to the advantage of startups. We’ve seen a lot of A.I. technologies, actually — now, interestingly standardizing — so going to open source. And then the next step is going to be, they’re gonna go to cloud, and that we’re right — because we think we’re right on the verge of that. We think all the major cloud providers are going to be providing A.I. as a service, and they’re gonna really radically reduce the amount of technical knowledge you need to apply A.I.. And so that plays very well to the startups.

Sonal: So, there will be, like, an AWS for A.I..

Marc: Yeah, exactly. And that may be literally AWS, or it may be Google, or Microsoft, or all three of them, and you know, in some combination. Or, it may be other, you know, other companies yet to emerge.

Sonal: An example of the open source, like TensorFlow, Google releasing TensorFlow.

Marc: Yeah. And this is a big deal, of course. Yeah, that’s right. So, Google open-sourced a pretty significant part of how they do deep learning, and that, actually, now, is something other companies can pick up and use directly. And we see, actually, not only a lot of companies but, like, a lot of university — a lot of student projects now just kind of pick that up and run with it. So, this technology is kind of trickling down very fast.

Sonal: Just this past weekend, we had a Hackathon. And I think most of the teams had some machine learning, A.I. component into their hacks. And these are college kids.

Marc: Yeah, yeah. You know, if you’re a 21-year-old junior in college and you’re doing some project, just, kind of — it’s rapidly becoming very obvious that you would have A.I. be part of it, which was very much not the case even 12 months ago. And that’s a direct, to your point, that’s a direct consequence of the open sourcing and kind of this knowledge spreading out. The second thing was the hardware cost, and there, again, the cloud, A.I. in the cloud — just the existence of the cloud is bringing down hardware costs across the board, but A.I. in the cloud is gonna bring that down even further. And by the way, these trends all slam together. So, you get what I think, in a year, is gonna be very common to these sort of A.I. supercomputing chips, with A.I. algorithms in the cloud available to anybody for a dollar, right? And so, there’s gonna be this massive deflation of hardware cost on that side. These big datasets are interesting.

Ben made the case that the startups can assemble big datasets, and I think that there are, certainly, examples of that. We also see another thing happening, which is the newest generation of experts in deep learning, or many of them are specializing in the idea of deep learning applied against small datasets. If you talk to those folks, what they’ll tell you is — [what] they’ll basically say is — primitive and crude deep learning require big datasets, but the really good stuff doesn’t. Small datasets are fine. And so, that’s still very early, but it’s extremely enticing. It’s an extremely enticing idea, because it really brings a lot of these problems, to your point, further into being tractable for small companies.

But actually, one of the things you can do with these — especially with these GPUs, is you can literally use the same tools that are used to make video games, and you can create simulated versions of the real world, and then you can actually let the A.I. train inside the simulation. And so, if you’re building a new self-driving car, or a drone, or something like that, you can actually create simulated worlds in which there are everything from earthquakes, to floods, to, you know, thunderstorms, hailstorms. You can create birds, swarms of birds. You can literally simulate the real-world environment, and then you can let the A.I. actually train inside that world. And actually, it’s funny. The A.I. actually has no idea it’s training in the virtual world. It’s learning just the same as if it were learning in the physical world. And so, again, for startups with access to cloud-based A.I., you could potentially run, basically, millions of hours of simulated training at very low costs, and all of a sudden catch up to big companies.

Ben: Interestingly, you know, the very famous A.I. project that Google did with DeepMind, that whole dataset came from the game playing itself. So, you know, it wasn’t some dataset that Google had collected over 20 years. It was the game playing itself.

Sonal: So, you guys have both mentioned simulations a few times. Why are they so important? Because I feel like there was this period, like, you know, maybe even a decade ago, where simulations were almost frowned upon as this promised thing that didn’t really actually deliver in what you needed to be able to navigate complex environments in real life.

Ben: Yeah. Well, it’s interesting, so was A.I. — was frowned upon 10 years ago, saying it was all — it didn’t work. I mean, particularly, neural nets and deep learning were the most frowned upon area. And there’s been similar, kind of, breakthroughs for simulation, first of all. So, if you think about the field of data science and what you do with data, you have a giant set of data, which is always historical in nature, and you can analyze that. And maybe it’s predictive of the future but oftentimes, it’s not. We see this, in particular, in things like really dynamic things, where the past affects the future, like, say, stock picking or the weather, or other kinds of things where data analysis doesn’t get you an accurate answer. Simulation is the flip side of that, where you can say, “Okay, here are all the entities in the world, and let’s generate their behavior over time,” and then their actual behavior feeds back into the simulation, which is critical — you know, a critical component.

Historically, that’s been difficult at scale, but there have been some really important breakthroughs lately, particularly from a company that we’re invested in called Improbable, which is able to do very large scale scale-out simulation, you know, using cloud computing techniques and some very important new technology that they’ve developed. And so, you can get a really complete picture of the world. And as Marc was saying, you can actually generate your own dataset, rather than collecting it for certain kinds of situations.

Marc: Yeah. Let me add one thing to that. So, one way to think about it is it’s expensive to make things happen in the real world. Like, it’s expensive to change things in the real world, because the real world is physical, and causing physical changes to happen — I mean, everything from building roads to flying planes, all these things are very expensive. And then things in the real world — changes have serious consequences, right? And so, you know, depending on where you put the dam, or where you put the airport, or what your evacuation plan you have for the city if something bad happens — like, you know, these decisions have huge consequences.

Ben: Which banks you bail out.

Marc: Which banks you bail out, which banks you don’t bail out. And so, you always have these consequences, and people who have to make these decisions are often flying blind, because they don’t have any real sense of what’s gonna happen as a consequence of their decisions. In contrast, if you can simulate a world, and if you can run an experiment — if you can simulate the real world or some portion of it, like the highway system, or the banking system, or whatever, and then you can basically introduce change into that simulation, and you can see what the consequences are — it’s very cheap to do that because Moore’s law, the collapse of chips, and the rise of cloud computing, all these other things we’ve been talking about, all of a sudden, make it very cheap to run these simulations. It’s much cheaper to do it in a simulated world, and then there are no consequences. You run a simulation and everything goes, you know, wrong, and everybody dies, or the entire financial system collapses, or whatever. It doesn’t matter. You just erase it and you run it again.

Sonal: Yeah. You have infinite testability.

Marc: Great. Yeah.

Ben: I wanna challenge that. There is Elon Musk’s simulation, in which case, the consequences are quite dire.

Marc: There is a scenario that we’re all living in a simulation…

Ben: Right, we’re living in one.

Marc: …in which case, I would argue it’s gone badly awry, as evidenced by the current political situation.

Ben: There’s no do-over button in this simulation.

Marc: Yes. And then you, basically, again, you look at the progress of Moore’s Law and the rise of these new technologies, and you say, “Okay, how about instead of running one simulation, let’s run a million simulations, or let’s run a billion simulations? And let’s try every conceivable thing we can possibly think of, and let’s imagine — let’s literally model all potential future states of the world, and then let’s decide which one of those — which path is the one that leads to the best consequences.” And so we can then make these very big real-world decisions with a lot more foreknowledge of what will unfold afterwards.

Real-world applications for technology

Scott: Maybe just to get concrete on some opportunities, what are the other areas in — maybe it’s life sciences, or what are some of the other kind of more tangible areas that you think near-term, as you think about kind of deploying this fund or beyond over the next, you know, 5 or 10 years that might be interesting for, you know, people to think about in the context of real-world applications of this technology?

Ben: Yeah. So, as Marc was saying, we’re coming into this era of new platforms, and with the intersection of health and computer science, what we’re seeing is really exciting new platforms around data and around, basically, you being able to get much more information about someone’s health from a variety of techniques that had been developed, you know, based on the, kind of, historic breakthroughs and sequencing the genome. But beyond that as well, where we can get really, really powerful data about people and understand them better. And once you have that data about people, wherein you can be predictive of diseases that they might get or things that are wrong, and you aggregate that into a platform, then you can actually make new scientific discovery off it as well. So, that’s one interesting area.

If you think about the A.I. platform itself, one of the things about it is the hardware that’s been built for it, or that’s been built historically, is for a completely different kind of computer programming. And we’ve seen Google already announce a chip to power their deep learning cloud. And you know, similarly, there’s new breakthroughs in quantum computing, which, at least on the surface, look like they may be very promising for much more powerful deep learning systems, and so forth. So, there’s a lot of things that are coming out of these platforms. And then, you know, as we get to chip and everything, the platforms to run and manage and understand those chips are equally as exciting.

Sonal: So, you know, one of the themes that’s come up through here is that tech is reaching into places it never did before. I mean, every company is becoming a tech company, or they have tech inside. Or, as Benedict likes to say, “Tech’s outgrowing the tech industry.” The reality is it’s permeating everywhere. And the question I have for us is that we are founded on this thesis that software is eating the world, that’s our premise. And yet we seem to have been making a lot of hard investments, you know, if you count things like Soylent, Oculus, Nutribox. So, are we changing our thesis about hardware as a result of this software eating in the world?

Ben: No, I don’t think so. I mean, I think that what we see with the companies that you’ve named are interesting. So, Oculus, I think we would all agree that the software component of Oculus is both more complex, has many more people working on it, and is kind of the core of the investment. Sometimes, if you have a breakthrough technology, then you require new hardware to actually support it. And that’s the case there. And I think that Soylent and Nutribox, both of them apply computer science techniques and information technology to get people to optimal health, and that’s what we’re doing there. So, I think we’re big, big believers that, you know, in the last 100 years, the great breakthroughs in knowledge have been the breakthroughs of people like Alan Turing and Claude Shannon, who gave us a new model of the world and how to understand it. And companies that build on that fundamental knowledge breakthrough are what we’re about, and we’ll continue to be about that.

Marc: Even if some of them may ship their products in a box.

Ben: Yes, a package is not a technology.

Scott: Let’s talk a little bit about SaaS. As you’ve probably seen, there’s been actually a bunch of acquisitions in the space recently, but what’s left to do there? So, is the new platform the salesforce.coms and others of the world, or are there actually both, kind of, vertical applications and/or are there other platforms that actually might exist over time in that market?

Ben: So, there’s SaaS as the metaphorical in-the-cloud version of all the stuff that we had built over the previous, you know, 30, 40 years. So, that’s, like, Workday, Salesforce, SuccessFactors, you know, the kind of big categories. The thing that we believe that’s changed as you go from on-premise to the cloud is, the technology is so much easier to adopt that we’re now seeing software applications for things that you just would never do as a software application, because the cost of — as we used to say in the old days, screwing it in, and paying the army of eccentric consultants to get it going — just wasn’t worth it for, say, expense reporting, which, you know, Concur, of course, built a really powerful product in that.

But, like, there was no packaged software for expense reporting in the same way that there is now. And I think there’s a gigantic number of categories in everything that you do in business that can be automated in that way. In addition to that, you can scale down to very, very small companies. Companies below thousands of employees never bought Oracle Financials. It would have been insane to do so. But they’re absolutely buying, you know, NetSuite and things like that. And then beyond that, now it becomes economical and very interesting to build vertical applications for industries. So, to build an application that revolutionizes, say, the real estate industry, or something like that, or the construction industry, is becoming extremely viable. And not just as a niche business, but as a real venture capital-based kind of activity.

Marc: One of the consequences that will be interesting to watch play out is that, historically, enterprise software has been described as represented by companies like Oracle, SAP, IBM. Like, that stuff was really only accessible to the largest companies, the top 500, 1,000 companies in a country. And then, in particular, only in a handful of countries. Those businesses, their revenue and their customer base have always been dominated by, you know, 2,000 or 3,000 companies globally that are these, you know, these giant multinational companies that we’ve all heard of. So, big companies had this sort of inherent advantage versus a lot of midsize and small companies, and then companies in the U.S. and Western Europe had this big advantage versus companies in other parts of the world, where the large companies and the large companies in the U.S. and Western Europe could just afford to make technology investments that small and midsize companies all over the world couldn’t make.

The sort of changes in SaaS that Ben described, they lead to an interesting conclusion, which is it may actually be interesting for a smaller company, or a company not in the U.S. or Western Europe, to be able to adopt the next generation of SaaS and cloud technology. It’s almost like, the folks who’ve been able to skip landline telephones and just go straight to mobile phones. You can just leapfrog the old stuff because you never had it, and you can just start using the new stuff out of the box. And then the big established companies might have a harder time adapting, because they’ve made these giant investments in the old systems, and it’s hard to just jump to the new thing. And so, there may be a power shift happening from, on the one hand, large companies to small and medium companies that can now more aggressively adopt technology faster — and then from companies in the U.S. and Western Europe to companies all over the world that can also do the exact same thing. And so, at the very least, a leveling of the playing field and possibly even a national shift in balance for small and midsize companies all over the world may all of a sudden get a lot more competitive.

Scott: So you’ve got, kind of, democratization, on one point. And then, to your point, there’s one version of internationalization, which is adoption across international communities. So, how do you think about, then, the other aspect of internationalization, which is company formation? Should we, then, expect to see more new company formation outside the U.S., partly as a result of some of these trends? And why won’t we see or will we see 50 Silicon Valleys, you know, over the next, you know, 20, 30, 40 years? And how do you all think about what the strategy should be vis-à-vis those opportunities?

Ben: That would be probably the most amazing thing for the world that could happen in the realm of business and economics. So, we’re hoping for it, and certainly, building — kind of, help trying to build technologies that would facilitate it. And I think the world has never been kind of more ripe for that kind of thing. Having said that, look, there are real network effects, geographical network effects, and Silicon Valley, obviously, has the biggest one in technology. And you always have to keep in mind, and this is something that gets lost, is — there are no local technology companies, right? There’s nobody who sells, you know, internet search to Wyoming. That’s not, like, a viable thing. So, when you’re competing globally, it does matter, you know, “Do you have the best people? Do you have the best executives? Do you have the best engineers? Do you have access to money?” Like, all these things become real competitive things. So, we still are believers in Silicon Valley, and we’re very hopeful that the rest of the world grows and that we can, you know, participate in that as well, but that’s TBD.

Marc: There’s an interesting macro kind of thing that’s happening. You know, one of the really, kind of, negative stories is that there’s, basically, the world is starved for innovation and growth. One of the data points you point to on that is, there’s now $10 trillion of money being held in government bonds, governments all over the world, trading at what’s called negative yield. This is literally, like, the equivalent of a savings account where, instead of a bank paying you interest, you have to pay the bank interest to hold your money. And so, there’s literally $10 trillion of capital parked around the world that is actually losing money as it sits there, which means people cannot find enough productive places to deploy capital.

The conventional view, if you just pick up the newspaper and read the economics section, how horrible this is and how it means the world is just starved for growth — the optimistic side of it is there’s $10 trillion of money sitting on the sidelines waiting for something productive to be done with it. What could be productively done with it, right? New kinds of health care, new kinds of education, right, new kinds of consumer products, new kinds of media, new kinds of art, new kinds of science, you know, new kinds of, you know, self-driving cars, new kinds of housing, all these things that need to be done all over the world. And so, the world has never been more ripe for a, you know, very large wave of innovation that would actually be quite easy to finance.

A lot of the time, you just can’t get things done because you don’t have enough money, right? That’s just kind of the constant state of the world for a very long time. And now, ironically, we live in a world where the opposite is true. There’s actually “too much money.”

Ben: Yeah, more money than ideas…

Marc: More money than ideas.

Ben: …which really can’t be true.

Marc: It can’t be true, right.

Ben: You have to unlock the ideas.

Marc: Human creativity is boundless. And so, if you can get more smart people around the world educated, and with the skills required to do these things, and if you can get them in environments, either create new environments to do that or figuring out how to get more of the people from other places in environments where they can do new things, we could do all kinds of new things, globally. And that’s something that we hope to contribute to, but I think is a very big opportunity for the world.

Scott: And so, do you think we’re getting to the point where it’s kind of geopolitical risk and rule of law issues that limit adoption or deployment of some of these new technologies in other countries outside the U.S.? It sounds like it’s less so technological advancement.

Marc: Well, I would say there’s bad news and good news. So, the bad news is, we frequently have delegations of folks coming into the valley from all over the U.S. and all over the world. And they basically come in, and its economic delegations, of different kinds of politicians, or whatever. And they come in, and they’re like, “Okay, what can we do to have our own Silicon Valley?” And then you kind of sit down with them, and you kind of go through, you know, ABCDEF, all these things. “Well, you want rule of law, you want ease of migration, you want ease of trade, you want deep investments in scientific research, you want no non-competes, you want fluid labor laws to let companies very easily both hire and fire, you want the ability for entrepreneurs to be able to start companies very quickly, you want bankruptcy laws that make it very easy to move on and start another company.” And at some point, the visitors give this stricken look on their face, and they’re like, “Whoa.” At the end of it, they’re like, “Okay, but, like, what if we want Silicon Valley but we can’t do any of those things?” And so, that’s the bad news.

Ben: And they can hire Donald Trump to run their country.

Marc: It’s ironic that we have this guy running for president who would seriously move us backwards on a number of those topics. So, even we struggle with these things, right? Like, I would argue, the formula is fairly well known. It’s just, people do not want to apply it for reasons that have a lot to do with politics and have a lot to do, you know, with other issues. The good news is it can be done, and then the other good news is it is happening, and there are very, very, very exciting things happening throughout much of the world. There are, you know, very active now startup scenes all through, you know, South America, Brazil, Argentina, Buenos Aires. Amazing things are happening in India. There’s all kinds of startup activity throughout the Middle East. There’s startup activity now throughout Africa. There’s, you know, obviously, China’s been a gigantic success story. Korea has all kinds of interesting things happening. So there are lots and lots of extremely positive early indications of what’s possible in many places all over the world. That said, there are very big political questions about whether or not those founders are gonna be able to operate in an environment that’s willing to let them succeed to the level that they should be capable of doing.

Ben: A big reason that we raised the fund and are excited about the fund is, it is a backing of our core belief system here, which is, we believe in the creativity, and ingeniousness, and intelligence of human beings and the entrepreneurs that we see and come to Silicon Valley and around the world. And we believe that these people absolutely have the ability to change things, and are changing things. And there’s plenty of room to improve the world, and there’s plenty of ideas to do so. And that’s really what we’re about with Fund V.

Team-building philosophy

Scott: So, let’s talk a little bit about, kind of, company-building and founders, in particular. So, you know, undoubtedly, you had a very distinct view of what types of founders you wanted to back when you started the firm, now, seven years ago. How has that evolved, if it all, over time? You know, what has changed either in terms of the types of founders you see, or the types of qualities you see that actually make founders successful, that’s caused you to either augment or rethink some of the initial, you know, foundations for the firm?

Ben: You know, I think a lot of the things — we had this great advantage when we started the firm that, you know, we, ourselves, were founders. I think that we’ve probably gotten, I would say, more risk-tolerant in our view of founders over time, even though sometimes…

Sonal: Wait, what do you mean by that? What do you mean by getting more risk-tolerant?

Ben: Well, we have this thing we say at the firm, which is we’re much more interested in the magnitude of the strength than the number of the weaknesses. We always believe that intellectually. I think that some of the number of weaknesses were fairly terrifying early on, just because, you know, you do have a lot of founders with a very small amount of experience these days, which is also, you know, part of their strength, in that it’s hard to rewrite the world if you’re too steeped in the world.

And so, I think, over time, we’ve kind of doubled down on that. And really, the founders who have figured out something really important, or who are true geniuses, or have will to power that we can’t even contain in the room — when they bring those things to the table, whatever is wrong with them, we tend to overlook and work with them on that. And if they’re strong enough in those areas, you know, the really interesting thing for us has been those weaknesses do go away pretty quickly. And that’s probably the biggest learning, is I’d say, we went in thinking that, but we’ve gotten even more extreme in our commitment to that kind of philosophy.

Scott: So almost in financial terms, you’re buying volatility to a certain extent.

Ben: Well, I think buying volatility, in the sense that we’re buying people who have world-class strengths where we care about them, and regardless of whatever else. There is volatility in that, but you can have a different kind of volatility. You know, you can have people who have gigantic weaknesses that are spectacular without having the strengths. And we’re not trying to buy that kind of volatility.

Sonal: How do you know, though, that they’re going to be the ones to actually build the companies that scale? Because there seems to be this inflection point, where the very thing that makes you a founder that’s gonna punch through this tough industry, is also the thing that’s pretty much gonna hold you back from really building your company in a really meaningful way if you think you can do everything, you know, your way. And there seems to be an inherent contradiction in that.

Ben: I think that that would be right if founders did not evolve. So, I think what…

Marc: And some don’t.

Ben: And some don’t. And some don’t. Like, some don’t and may get stuck, and they can’t get past that point. But you know, it’s a real common characteristic in great founders that they want to know absolutely everything about the company and how it works, and, you know, every knob and every button. And they really would, like, have a strong desire to actually be able to do every job in that company themselves, if it came down to it. But those kinds of founders also have great ambition, and it’s very logical and easy to understand that there’s never actually been a gigantic long, you know — a really important long-lasting company that had, like, five employees. Those just don’t exist.

And so, if you’re gonna have to have a bigger company than that, you have to think about the company not only, you know, from the scale perspective, but from the perspective of the people working there. And how are you gonna get great people to work with you if you’re literally making every decision in the company? And I think that not every founder can let go of that, and sometimes it’s a psychological flaw rather than a desire for greatness. And if it’s a psychological flaw that they can’t overcome, then, you know, it’s just like any flaw that any of us have — you know, where we can’t stop eating ice cream or whatever. And you know, there’s nothing we can do at that point. Like, we can give them the logical explanation, but they’ve got to fix themselves.

Scott: One of the things that we’ve seen even in the short time that the firm has been in business is companies staying private longer, or taking a longer time to IPO. What are some of the implications of that on the company building process? How do you, kind of, balance that new reality, if it is a new reality around how companies stay private, with how you think about building management teams and other issues around the company?

Ben: I think this gets back to probably one of the more neglected parts of company building, which is, like, “What is the company culture? What does it believe? What’s our way of doing things, you know, when we come to work every day? What does quality mean? How do we prosecute an opportunity, and the kind of philosophy, onboarding, training into that culture, and so forth?” And so you kind of have to develop a philosophy. Like, what kind of employees do you want? How do you want them to behave when they get there? How do people contribute?

Scott: As we’re getting close to wrapping up here, what would be one piece of advice that you might give either from a management perspective, from a go-to-market perspective? What would be a takeaway for people listening to this podcast?

Ben: From a management perspective, I think the most common mistake that founders make is, they make decisions based on — management decisions and organizational design decisions — based on very kind of proximate perspective. So, what’s my perspective, what’s the person I’m talking to’s perspective, what’s my HR person’s perspective, without, like, taking the time to go, “Okay, like, how does everybody in the entire company see this decision, and how will they see it once it’s made? Is it motivating people in the way that I think it will? And let’s look past the person I’m talking to feeling good about what I’m saying, and really make this for the long-term health of the organization.”

Marc: Yep. The single biggest strategic piece of advice we just see across all of our companies is, literally, people just need to raise prices. People need to charge more for their products and services. The good news is you have all these new founders with many different backgrounds who have come in, many of them have never run companies before, run salesforces before. And so they have these extremely sophisticated views on things like products and design and engineering, and then I think, in some cases, relatively naive views on how to actually prosecute a campaign to be able to get the world to use your product. And so, the temptation we see from many founders is to have a one-dimensional view — what I call a one-dimensional view of the relationship between price and volume. Which is, if I price my product cheap, then I sell more of it, because the assumption is just that people just make purchase decisions based on cost. And so, you drive down prices, you drive up volume. And by the way, a lot of the history of the tech industry, like the chip industry, is “drive down prices, drive up volume.”

But a lot of startups really suffer from having that view. Instead, we encourage companies to adopt what I call, kind of, the two-dimensional view, which is the advantage of raising prices. Actually, there’s a couple of advantages. So, one big advantage — if you raise prices, you can afford a bigger sales and marketing effort. A lot of companies have prices that are actually too low to be able to mount the kind of sales and marketing campaign required to get people to ever actually buy the product. And I call this the “too hungry to eat problem,” right? I’m not selling enough, but I’m not selling enough because I don’t have the sales and marketing coverage required to actually get the product out there, and I don’t have that because I’m charging too little. As a consequence, I’m not selling any despite my low prices.

The other really interesting thing is that, for a very large number of products, it turns out, if you charge higher prices, the customers take the product more seriously. They impute more value into it when they’re making their purchase decision. And then once they’ve purchased, they’ve made a bigger commitment to it. And in particular, anybody selling anything to businesses, businesses will take something that they had to pay a lot of money for a lot more seriously than something that they didn’t have to pay very much money for. So, you can get a much higher level of engagement and stickiness, and actually use of your product, if you charge more. Going through this, this definitely has felt like swimming upstream for the last several years. We see some glimmers that more folks are starting to figure this out.

Sonal: Okay. Well, that’s all we have time for. I think this is the first time I’ve actually had all you guys together on the podcast since we did our fifth anniversary podcast a couple of years ago. Kind of amazing how much has changed even in that short amount of time. So, thank you. Thanks, everyone.

Marc: Thank you, Sonal.

• 

  Marc Andreessen is a cofounder and general partner at a16z. Marc co-created the highly influential Mosaic internet browser and cofounded Netscape.

  Follow Twitter
• 

  Ben Horowitz is a cofounder and general partner at the venture capital firm Andreessen Horowitz. He is the author of The Hard Thing About Hard Things and What You Do Is Who You Are.

  Follow Twitter
• 

  Scott Kupor is an Investing Partner at Andreessen Horowitz where he is also responsible for all operational aspects of running the firm.

  Follow Twitter
• 

  Sonal Chokshi is the editor in chief as well as podcast network showrunner. Prior to joining a16z 2014 to build the editorial operation, Sonal was a senior editor at WIRED, and before that in content at Xerox PARC.

  Follow Twitter

Who has the advantage in artificial intelligence — big companies, startups, or academia? Perhaps all three, especially as they work together when it comes to fields like this. One thing is clear though: AI and deep learning is where it’s at. And that’s why this year’s newly anointed Andreessen Horowitz Distinguished Visiting Professor of Computer Science is Fei-Fei Li [who publishes under Li Fei-Fei], associate professor at Stanford University. Bridging entrepreneurs across academia and industry, we began the a16z Professor-in-Residence program just a couple years ago (most recently with Dan Boneh and beginning with Vijay Pande).

Li is the Director of the Stanford Vision Lab, which focuses on connecting computer vision and human vision; is the Director of the Stanford Artificial Intelligence Lab (SAIL), which was founded in the early 1960s; and directs the new SAIL-Toyota Center for AI Research, which brings together researchers in visual computing, machine learning, robotics, human-computer interactions, intelligent systems, decision making, natural language processing, dynamic modeling, and design to develop “human-centered artificial intelligence” for intelligent vehicles. Li also co-created ImageNet, which forms the basis of the Large Scale Visual Recognition Challenge (ILSVRC) that continually demonstrates drastic advances in machine vision accuracy.

So why now for AI? Is deep learning “it”… or what comes next? And what happens as AI moves from what Li calls its “in vitro phase” to its “in vivo phase”? Beyond ethical considerations — or celebrating only “geekiness” and “nerdiness” — Li argues we need to inject a stronger humanistic thinking element to design and develop algorithms and AI that can co-habitate with people and in social (including crowded) spaces. All this and more on this episode of the a16z Podcast.

Show Notes

• Where AI research is today in terms of hardware and algorithms [0:51]
• Discussion of creativity and artistic, “generative” intelligence [11:14]
• Where startups have opportunity in AI [15:51] and a discussion of self-driving cars, including the ethical issues [19:00]
• How AI needs to learn to interact with humans in a socially-acceptable way [24:24]
• Adding a humanistic element to AI research to attract more diverse young people [29:11]

Transcript

Sonal: Hi, everyone. Welcome to the “a16z Podcast.” I’m Sonal, and I’m here today with a16z partner, Frank Chen. And we’re interviewing our newest professor in residence. This is actually the third year — the first year we had ViJay Pande, who’s now the general partner on our Bio Fund, and then we have Dan Boneh. And now, we are so pleased to welcome Dr. Fei-Fei Li, who is the director of the Stanford AI Lab, the Stanford Toyota AI Center, and the Stanford Computer Vision Lab, which is pretty much the most important work happening, at least we think.

Current state of AI

Frank: AI is the white hot center of both a lot of startup activity, as well as academic research. And, Fei-Fei, why in the world has it gotten so hot again?

Dr. Li: From my perspective, AI has always been hot. AI is a discipline about 60 years old. In the past 60 years, I call that the “in vitro AI time,” where AI was developed in the laboratories and mostly in research centers. We were laying down the mathematical foundations of AI, we were formulating the questions of AI, and we were testing out the prototypes of AI algorithms. But now, going forward, we’re entering what I call the “AI in vivo time,” [in] which AI is entering real life. So why now? What’s triggering the switch between in vitro to in vivo? I think several things are happening. 

First is that AI’s techniques have come of age, but what’s driving that? There are two more very important factors. One is the big data contribution to AI. It’s, you know, the information age. The internet age has brought us big data, and now even boosted by just trillions of sensors everywhere. And the third factor that’s contributing [to] this is the hardware — the computing hardware, the advance of the CPUs, of the GPUs, and the computing clusters. So the convergence of, I’d say, mathematical foundations and statistical machine learning tools, the big data, and the hardware has created this historical moment of AI.

Frank: Why don’t we unpack those in turn? Because I think each one of them in themselves are interesting trends. So why don’t we talk about hardware? We have CPUs, we have GPUs. So it turns out deep learning is great to do on GPUs, because it’s linear algebra and parallelizable. Are we going to see deep learning chips?

Dr. Li: I think so, and I hope so.

Sonal: What would deep learning chips look like? Just obviously, much more — the ability to do much more parallelization, but what does that actually look like? Is it like what’s happening with Nvidia’s chips right now, or something different?

Dr. Li: Nvidia is definitely one of the pioneers in deep learning chips, in the sense [that] their GPUs are highly parallelizable — can handle highly parallelizable operations. And as it turned out, much of the internal operations of a deep learning algorithm, which technically we call neural networks or convolutional neural networks, involves a lot of repeated computation that can be done concurrently. So the GPUs have really contributed a lot in speeding up the contributions, because this can be done in parallel. GPUs are wonderful for training the deep learning algorithms. But I think there is still a lot of space in rapid testing or inference time chips, where it can be used in recognition, you know, in embedded devices. So, I see there is a trend coming up in deep learning chips.

Sonal: So more specialized hardware dedicated for that.

Dr. Li: Yeah.

Frank: Yeah. And we’ve already seen the startups do it like Nervana. Obviously, Google announced the TensorFlow processing unit, right? So they’ve got dedicated silicon as well. So…

Sonal: So GPU to TPU, basically?

Frank: Yeah, exactly. Once you know that you’re going to do something over and over again, then you want it in silicon — both for the performance, and then very importantly on the embedded side, for power consumption. Which is — you want your iPhone eventually to be able to do this.

Dr. Li: But I still think, Frank, that this is a little bit — I wouldn’t say it’s too early, but I think we’re still in the exploratory stage, because the algorithms are not matured enough yet. There’s still a lot of exploration about what to do, in the best way. So, you know, like — this year ICLR, one of the top deep learning conferences — one of its best papers is on [the] particular work coming out of Stanford. Not my lab actually, somebody else’s lab — Professor Bill Dally’s lab — where they’re exploring a sparse algorithm that can enable a specific design of a chip. So, this conjunction of improving algorithms in order to also design the innovative chip is still happening right now.

Sonal: Is that a new thing?

Dr. Li: You mean like algorithms driving the design of the chips…

Sonal: Right, versus the other way around, sort of the chicken egg thing and what comes first.

Frank: Chip design is already so complicated that you have to do it with algorithms. Humans can’t actually lay out chips.

Sonal: Oh, I don’t mean algorithm, like a design. I thought what you were saying was, designing the chip for a particular type of almost universal algorithm, which is how I heard that thing.

Dr. Li: It is designing the chip for a type of algorithm, but it’s a family of algorithms.

Frank: Your argument is that because we’re not sure what the winning algorithms are going to be, we’re still in this very productive period where we’re trying lots and lots of algorithms. It might be too early to design chips, because to put something in hardware, it’s obviously incredibly expensive to get to an ASIC, right? It’s $50 million to tape out. And so unless you’re sure you know what algorithms are gonna run, you can’t optimize the chips for it. Is that…

Dr. Li: Oh, actually, I think it’s really important [that] this thing is happening right now. This R&D has to happen concurrently. It’s just, like Sonal said, there’s a chicken and egg dynamic here, that algorithms affect the way chips are designed, but the constraints of the chips could in turn affect the algorithm. I think this is [the] time to explore this. This is the time to devote resources. Of course, in terms of business model, one has to be careful.

Frank: So the second thing — or another of the three things that you mentioned was that we’ve laid the mathematical foundations for artificial intelligence. And I want to come back to this idea of, look — the hottest thing right now is deep neural networks. But over the 60 years of AI research, we’ve actually used many, many different techniques, right — logical programming. We’ve used planning algorithms, we’ve tried to implement planning algorithms as search algorithms. And so, is deep learning it? Is this what the community has been waiting for or is this just, “Okay, it’s hot now but there’s going to be something else later, too?”

Dr. Li: I get this question a lot — is deep learning the answer to it all? So, first of all, I’m very happy you actually brought up other algorithms and tools. So, if you look at AI’s development, in the very early Minsky, MacArthur days, they used a lot of, you know, first-order logic and expert systems. And those are very much driven by cognitive designs of rules. But what really, I think, was the first AI spring face is the blossoming of machine learning, statistical machine learning algorithms. We’re looking at, you know, boosting algorithms, Bayesian nets, graphical models, support vector machines, regression algorithms, as well as neural networks. So, that whole period — there is about 20, 30 years of blossoming of machine learning algorithms [that] laid the statistical machine learning foundation to today’s AI. And we shouldn’t overlook that.

In fact, many, many industry applications today still use some of the most powerful machine learning algorithms that are not <inaudible>. Deep learning is not the newest. It’s actually developed in the ’60s, ’70s by people like Kunihiko Fukushima, then carried out by Geoff Hinton, and Yann LeCun and their colleagues. I think there [are] some really powerful ingredients of the neural network architecture. It is a very high-capacity model that can take almost any function, and they can do end-to-end training that takes data and all the way to the task objective and optimize on that. But is deep learning it? I think there’s quite a few questions [that] remain that would challenge today’s deep learning architecture and hopefully challenge the entire thinking of AI going forward. One of the more obvious one everybody talks about is supervised versus unsupervised training.

Sonal: And this is I think so important, because a drawback of the current narrative is that it focuses so much on the supervised cases — that we don’t have computers that learn the way children learn.

Dr. Li: Exactly. First of all, we don’t even know much [about] how children learn. There’s a vast body of education, developmental psychology literature, and that’s not getting into computer science yet. You know, supervised learning is powerful when data can be annotated, but it gets very, very hairy when we want to apply a more realistic training scenario. For example, if one day a company builds a little robot that sends to your home, and you want the robot to adapt to tasks that your family wants to do. The best way of training is probably not to open the head of the robot and put in all the annotated data. You want to just, you know, like show and talk about what tasks there is, and have the robot observe and learn. That kind of training scenario, we cannot do in deep learning yet.

But there’s more than just supervised training versus unsupervised training. There is also this whole definition of what is being intelligent, right? Task-driven intelligence is really important, especially for industry. You know, tagging pictures, avoiding pedestrians, speech recognition, transcribing speech, carrying goods. Specific task-driven applications are part of AI and [are] important, but there is also the AGI, artificial general intelligence, of reasoning, abstraction, communication, emotional interaction, understanding of intention and purpose, formulation of knowledge, understanding of context — all this is still largely unknown in terms of how we can get it done.

Creativity and generative intelligence

Sonal: Where would you put creative AI on that list from — okay, there’s the problems that are yet to be solved — unsupervised, supervised, generalized intelligence, and now also creative intelligence?

Dr. Li: Actually, you know, here’s one question we should ask ourselves. What is creativity? If you look at the four, five matches of AlphaGo, there were multiple moments when AlphaGo made a movement — Master Lee Sedol was really surprised. And if you look at the Go community, people were just amazed by the kind of creativity AlphaGo has, in terms of making the moves that most people cannot think of. From that point of view, I think we’re already seeing creativity. Part of creativity is just making right decisions in a somewhat unexpected way. That’s already happening.

Sonal: I’m meant, actually — more interested in the type of creativity where it defies logic, because that’s an example of logical creativity. I’m thinking of something like Jackson Pollock. There is no way a computer is going to waste paint and splatter it, because it’s the most inefficient, irrational thing to possibly do. That’s the kind of creativity I want to know about. I mean, I’m seeing examples of, like, AI-written short films, AI poetry — your own lab, there are people who are writing captions for images. That’s, like, maybe still mechanistic, and Kevin Kelly would even argue that creativity in itself is largely mechanistic, and it’s not as human as we think, anthropomorphic as we think it is — but I really mean like, artistic creativity.

Dr. Li: Yeah, that’s a great question. So interestingly, you already see some of the deep learning work of transferring artistic style. You can put in a Van Gogh painting and turn a photo into that, but I agree that’s very mimicking.

Sonal: Mechanistic.

Dr. Li: Mechanistic. The kind of creativity we’re talking about — blending our logical thinking, emotional thinking, and just, you know, intuitive thinking — and I haven’t seen today’s — any work that builds on the kind of mathematical formulation that would enable that.

Frank: Yeah, it comes back to one of the three things that you use to set up, “Why is AI winning now?” And that is about data, which is if you’re just going to feed the system a bunch of data and then have the neural net train itself, can that ever lead to something that’s truly creative, which isn’t in the data itself?

Dr. Li: Right. Exactly. So, this is…

Sonal: Exactly. Or, maybe it could, by the way, because maybe it can follow the same type of logical arc of history, where you go through a classic phase, a traditionalist phase, an impressionist phase, a post-impressionist phase, an abstract phase. And then you actually go through Jackson Pollock — kind of, Modern Art phase. Like, I almost wonder if you could technically train on that type of history of art and see what happens. I know that’s crazy, and this is completely abstract. And it’s not in any way tied to the actual computer science, but just theoretically.

Frank: We already have systems that can paint in all of those styles, because there was enough in the data so that it could form a classifier that said, “Here’s the style of Van Gogh,” or, “Here’s the style of an impressionist,” and then we can mimic those styles. So, the question is, down that road, using deep learning, can you ever get to breakthrough new things?

Sonal: Right. Generative intelligence. Not general, but generative.

Dr. Li: Generative. So, there’s a lot of thinking on that. We’re pretty far from going from Impressionism to Cubism and all this. But coming back to a more mundane class of work, for example, we are doing computer vision. And some of our work recently is to write a brief captioning or a few caption sentences about images. And then the next thing we did is to start doing Q&A of a picture. And at this point, we start to think, “Can we actually develop algorithms that’s not just learning the training data but learning to learn?”

Sonal: Exactly.

Dr. Li: Learning to ask the right question. For example, we just submitted a paper that is — if we show the computer a picture and ask a question about, “What is the woman doing?” —  instead of directly having the computer learn to answer, the computer needs to actually ask a series of questions in order to answer this. So the algorithm needs to — not learning to answer the question directly, but learning to explore the potential space to ask the right question to arrive at the final answer. So the ability of learning to learn is what we want children to have. And this is what we’re exploring in our algorithms.

Sonal: Okay, so then let’s go back for a moment to something you said earlier, Fei-Fei. You know, I really like how you describe that these phases — the, sort of — the in vitro, like, the laboratory phase, and then the in vivo, like, the in-real-life phase. It’s a wonderful way of clumping the work and the moment we’re at, but there’s always been industry and lab and company, you know, collaboration since the beginning of computing. So, what is different now that startups can play in this space, in vivo?

Dr. Li: I think several factors. One is that the algorithms are maturing to the point that industry and startups can use it. You know, 20 years ago, it’s only a few top places in the world, top labs in the world, that hold some algorithms that can do some AI tasks. It’s not percolated to the rest of the industry or rest of the world. So, for any startup, or even company, for that matter, to get their hands-on those algorithms is difficult. But there are also other reasons. Because of the blossoming of [the] internet, because of the blossoming of sensing, we now have more use cases. In order to harness data, we need to manage and understand this information. This created a huge need for intelligent algorithms to do that. So, that’s a use case. Because of sensing, we start to get into scenarios like self-driving, and, like, cars. And now suddenly, we need to create intelligent algorithms to have the cars drive. So, that’s what’s creating this, in my opinion, blossoming.

Frank: The fun thing to watch unfold will be startups versus big established labs and companies. And on the one hand, we’ve got George at comma.ai who built a self-driving car by himself, like, one person. And then on the other side, you’re involved with the SAIL-Toyota Center for AI Research, which is sort of the big industrial approach to this. So, what do you think the relative contributions will be between startups and big organizations?

Dr. Li: In terms of self-driving cars, who is gonna win the self-driving car competition, right? I think the advantages of the big companies are some of the following. A company like Toyota, as soon as they are committed to this, I hope that they put cameras in their cars. They can already get data very quickly, whereas a startup, this is a lot more difficult.

Sonal: The data, again, is the big differential.

Dr. Li: Companies like Google, even though they didn’t have cars at the beginning, they had algorithms. They started this early. So, they now have both data and algorithms.

Sonal: They were a software company first, as opposed to a car company trying to become a software company.

Dr. Li: Exactly. The software is such an important part. They actually have an edge there. What about startups? Do they still have an edge? I think there is a lot of business scenarios that might be not so critical on the path for these big companies. But the startup can come in through a more niche area, or more vertical space, and build up their data and algorithm that way. Or, the startup company can do what Mobileye does. Instead of building the entire system — [the] entire car — they build one critical component that’s better than anybody else. And that’s another angle they can come in. 

Self-driving cars and ethical issues

Frank: Your colleague, Andrew Ng, who used to be at Stanford and now runs the AI lab at Baidu, has called Tesla’s autopilot system irresponsible, because it got into a crash. Because there are well known scenarios, basically, where the system wouldn’t perform safely. And so, Andrew said, “Look, it’s premature.” So, I wanted to get your thoughts on this, especially since you’re involved with the Toyota program.

Dr. Li: So, when Tesla’s autopilot came out, I watched some of the YouTube videos. As a mom, I would never want to put my kids or myself into those cars. So, from that point of view, I did, kind of, react — you know, squeamishly on that. But what I’m hoping here, is a really clear communication strategy between the business and the consumers. I don’t have a Tesla, so I don’t know what Tesla told the users. But if the communication is extremely clear about when you should trust the system, and when you should use it, when you shouldn’t — then we get into a situation, you know, when customers are not doing the right thing — who is to blame? And we’re getting more and more into that in AI and ethics — is that, who is to blame? Because every single machine, if used in a wrong way, would have its very scary consequences. I think that’s a societal conversation we need to be having.

Sonal: Yet another example of how technologists and technology needs marketing. I mean, we tell our company CEOs all the time about the importance of these functions. It just continually reinforces that.

Frank: Yeah, marketing and training and the right user experience.

Sonal: Right design.

Frank: So, this is going to be one of the hardest areas to design for, which is, if we’re on this continuum somewhere between intelligence augmentation and full autonomy, how do you design a system so that the driver knows, “Oh, it’s time for you to pay attention to again, because I don’t know what to do.” Does the steering wheel vibrate? Is there an auditory cue? Like, these are gonna be tricky systems to design.

Sonal: I agree. And I think this is actually where there is a really important conversation to be had. Nissan has an anthropologist on staff, Dr. Melissa Cefkin. I forgot how to pronounce her last name, but she’s an anthropologist whose full-time job is to study these issues in order to build it into the actual design. And it’s not just, like, software engineers who are designing this. It’s a conversation to be had.

Dr. Li: In our Stanford-Toyota Center, this center has a group of professors working on different projects. And there is one big project that is led by [the] Human-Computing Interaction Group.

Sonal: It’s HCI, right?

Dr. Li: Yeah, it’s HCI because of this.

Frank: Yeah, it’s great to see, sort of, anthropologists, maybe philosophers come back into the mix, because these complex systems — you’d really want the full 360 degree view of design. It’s not just what technology enables, but what are human expectations around it.

Dr. Li: And one thing to really keep in mind. Compared to computers, humans are extremely slow computing machines. The information transfer in our brain is very slow compared to transistors and, add on top of that, our motor system — you know, from our brain to our muscles — is even slower. So, when we are talking about human machine interaction and split second decision making, we should really factor in that.

Frank: Yeah, it sort of brings to mind the famous trolley problem. You knew I was gonna here, Sonal, right? Because I can’t help bringing this up.

Sonal: And I edited Patrick Lin, who is, like, a long time thinker in this space. And he…

Frank: Yeah. And the YouTube video that Patrick created is great. So, if you want to sort of see the full exposition, go see his YouTube video. But in summary, the challenge is this — humans are slow. And so, if you get into an accident because your response time was too slow, you’re definitely not liable, right? Like, you just couldn’t control the car breaking in front of you. An autonomous car can actually make a decision. So, imagine that you’re an autonomous car, and then your algorithm needs to decide, “All right, the truck in front of me suddenly braked. I could plow myself into the back of the truck and injure my passengers, or I could swerve to the right and maybe take out the motorcyclist, or I could soar to the left and hit a minivan.” The computer will need to make an explicit decision. And it has the reaction time to actually make an explicit decision. And so, if that decision is explicit, can it be held liable? Can the designer of that algorithm be held liable, because it made an explicit decision rather than having a split second response.

Sonal: When people bring up the trolley example, it gets really frustrated, because it’s so abstract. But I actually think that the act of going through this thought process is exactly what gets you to answering these questions that you’re asking about the liability — who’s accountable, the emotional tradeoffs that we make, and how to understand even our own limitations, as you point out, Fei-Fei,

Dr. Li: This actually brings up the topic that in the past few years, I’ve been really advocating in the education and research of AI. We need to inject a strong humanistic thinking element into this, because our technology is more and more in vivo. It’s touching people as real lives. And how do we think and develop and design algorithms that can, you know — hopefully better humans’ lives, but really have to cohabitate with humans. We need that kind of humanistic thinking.

AI and social interactions

Sonal: I actually want to ask about a paper that you guys recently just put out. I actually included it in our last newsletter. It was about autonomous cars navigating social spaces. So interesting, because this is lab research in the wild. This is no longer — you know, we can have these algorithms work perfectly fine. But to have them navigate — I’m thinking of streets like in India where, you know, there will be a cow and like 10 buffaloes behind you in the middle of all this, and I don’t know any computer that’s accounting for that. So, I’d love to hear how you guys came to that paper and some of the thinking.

Dr. Li: This is a project [where] the main PI is Silvio Savrese. It’s the social robot they created called Jackrabbot, so to honor California’s jackrabbit. And the purpose of Jackrabbot is an autonomous driving robot or vehicle that’s taking care of what we call the last miles of driving, where it tends to be in much more social spaces rather than highways. You know, sidewalks, busy cities, campuses, airports, and all this. So, when we look at the problem of last miles of driving, or just the social space, we quickly realize the problem is — you know, not only you have to do everything that a highway driving car needs to be doing to understand the layout of the seeing the pedestrians, the lanes, and all this — you also have to navigate in a way that is courteous and acceptable to people.

So, one naive solution, people say, “Well, you know, just make a really low speed and stop whenever there’s people.” We tested that. If we do that, the robot will never go anywhere. Because in a very crowded space, there’s always people. If the robot just follows the most naive rule of, “I’ll yield to people all the time,” the robot would just be sitting there from the starting point and not getting anywhere.

Sonal: Frankly, if that robot was used in San Francisco, it would be kicked, too, probably a couple of times. Maybe people will be really irritated about it — or New York, they’d be irritated in Time Square about it moving so slowly.

Dr. Li: Yeah, right. So, we thought about that, and we haven’t thought about, you know, what to do yet. We think with — the robot has to have an SOS kind of call. So, what we want to do is to create a robot that understands human social dynamics. So, it can carry it’s on task, for example, going from A to B to deliver something on campus, but do it in a courteous way. So, we started to first record human behavior by data on campus and look at how people gathered together when they talk in small groups, or how they walk — especially, you know, 9:00 [on the] Stanford campus, there’s so many students going into so many classes. But they’re not going in a completely random way. They tend to form interesting patterns, depending on the direction they’re going.

So, we gather all this data, we feed it into the algorithm. Have the algorithm learn about this — especially from injecting some social rules, such as, people tend to follow others in going the same direction. You do not break two people or several people when they’re talking. So, we injected all these and learned the right way of doing it. And then we put it into the algorithm. And then the algorithm started to learn by itself how to navigate.

Sonal: Just to probe on that — how to navigate, not how to learn those social cues itself.

Dr. Li: Right. How to navigate. We give them some social cues, but we only give them high level cues, The detail, for example — the algorithm still has to learn, “When I avoid two people talking, how far do I avoid? Do I avoid them by 10 feet or two feet?” These are the things that are learned just by observing.

Sonal: Have there been any new surprises yet for you guys, out of this?

Dr. Li: No. Sorry.

Frank: When I read the paper, the question that immediately came to mind for me — which is that social norms vary from place to place.

Sonal: That’s what I was thinking too, the cross cultural aspect, especially.

Frank: And so when we ship these robots that observe social norms, is this going to be the new localization? In other words, here’s the self-navigating robot, Mumbai edition. Here’s a self-navigating robot, Boston edition.

Dr. Li: Excellent question. So, my answer to that is, as of now, we have to train them location by location. We have to gather data. But, as I was saying earlier, you know, the next dream I would have is to teach robots how to learn — learning to learn — rather than just to mimic training data. At that point, it should be online learning. It should be incremental learning so that the robot can adapt to different…

Frank: Right. So you wouldn’t have to train it on a particular city’s actual traffic patterns. You just drop it in there and the robot will figure it out.

Dr. Li: Exactly.

Humanistic thinking in AI research

Sonal: Like the way humans do when you travel like to be — when in Rome, do as the Romans do, so to speak. I mean, I come from the world of developmental psychology, and the development of moral and social mores requires not just a regular cognition, but a metacognition and an awareness of your own thinking — that is a whole new layer that it just complicates things. So it’s super fascinating. Okay, so I want to go back, then, to something you said, Fei-Fei, about this humanistic side of things. Tell us more about what you’re thinking when you say that. Like, do you mean that we should be injecting humanities into computer science, or art — like, you know, I’ve heard of this move from STEM to STEAM. Like, what are you actually talking about when you say that?

Dr. Li: So, here’s where it all came from. About three years ago, I was thinking — I was observing that in my professional life, there are two crises people tend to talk about, and they seem to be completely disconnected, these two crises. The first crisis is that terminators are coming next door, and AI’s are turning evil, and all this. We’re summoning evil, and AI is gonna just one day rule us all. That’s one crisis. Another crisis we hear here also is about the lack of diversity in STEM, and computing. And from where I stand, the total lack of diversity in AI. And it dawned on me that these two crises are actually connected by a very important hypothesis, which is the lack of humanistic thinking and humanistic mission statement in the education and development of our technology.

So, let’s look at the first one. Why do we ever think technology might turn evil? Well, technologies are always in the hands of people. Technologies themselves are neutral. You know, be it nuclear weapons or nuclear physics, or just a knife, you know, that can cut [an] apple — you know, in the hands of people, technology can have consequences. So, in order to have responsible and benevolent technology, what we really want is to have a society, have a group of technologists, who have the humanistic awareness and thinking — so that we can use technology responsibly. So, that’s related to the first thing. The second thing is, why are we not — millions and millions and millions of dollars are put into attracting diversity into computing and STEM. And where I stand, I find it very hard to convince women and underrepresented minorities to work in AI.

Sonal: This is, by the way — despite being at Stanford, which has, what, 50/50 parity in the computer science program with women and men?

Dr. Li: Oh, no, it’s not 50/50. It’s about 25% to 30%, in undergraduate that we have women. And then this thing just goes down as you…

Sonal: Oh, goes down as you go higher. Okay.

Dr. Li: Oh, yeah. The attrition at every stage is grim. And so, looking at Stanford students, they’re extremely talented. Almost any student coming to Stanford, whether it’s an undergrad or a Ph.D, they’re talented enough to be analytical, but also have, you know, great writing skills, care about the world. I suddenly realized here, in our field, as well as Silicon Valley, we’re not sending the right messages to attract people of all walks of life.

Sonal: What do you mean by that?

Dr. Li: We tend to just celebrate geekiness, nerdiness. But when you have an ambitious young woman coming into our department, or into the AI lab, she might be thinking about the aging society. She might be thinking about curing cancer. She might be thinking about a lot of socially important topics. If we present ourselves just as geeks loving to do geeky things, we’re missing a huge demography who actually want to turn technology into [a] humanistic mission. So then, suddenly, I realized, we’re missing [a] huge opportunity attracting diversity, because we’re not talking enough or thinking enough of [the] humanistic mission in AI. And that united my two themes I’ve been thinking about.

Sonal: Just to put a sharp and a point on this. I don’t want to be cliché about “only women and underrepresented minorities would take on ‘the soft problems,’” because there are also other people who might want to take on those challenges of aging, and some of the other interesting shifts that are happening. But to your point, we’re not necessarily inclusive enough — we’re not thinking about this enough, period, regardless of background — to be able to really welcome that type of thinking.

Dr. Li: I think it’s all walks of life. They come with their experiences and value systems.

Sonal: That’s fair.

Dr. Li: The one thing I start to notice. I have a lot of friends who are extremely successful Silicon Valley entrepreneurs and technologists. And, given my own age, all of my friends — many of them are entering the age that they have aging parents.

Sonal: Yes, this is so top of mind.

Dr. Li: Suddenly they’re talking about health care.

Sonal: Which they never did before.

Dr. Li: When they were [in their] 20s, they’re thinking about beers. You know, they’re not talking about health care. Yeah.

Sonal: Your point is that having that access to that experience is really important to that perspective.

Dr. Li: Right. So all walks of life add to our collective thinking and creativity…

Sonal: Right. It’s a great point.

Dr. Li: …in our technology.

Frank: I know one of the things that your lab does is an outreach to high school girls who come to campus for two weeks.

Dr. Li: This is the brainchild of me and my former student, Dr. Olga Russakovsky. Our hypothesis is, let’s catch girls at the age that they’re starting to think of who they are and what they want to do. And we find the age group of high school freshmen to sophomore thinking about what they want to focus on. So, we created this AI. camp that specifically — we aim for two things. One is, we want to be very technical because we want to get — inspire the future leaders of AI, and talented math and computing students. But we want to attract these students who otherwise might not think of AI, because they didn’t know such a strong humanistic mission is in AI. We actually [ran] a very rigorous hypothesis testing over the summer and wrote a technical paper about this.

Sonal: I like this approach, by the way, because I get really tired of hearing all the different “camp for this, camp for that, program for this, program for that,” and I feel like, “Come on, guys, are we really solving the problem?” It’s kind of refreshing to hear that you’re taking a much more rigorous approach to it.

Dr. Li: Right. So our campus designed — in the morning, the students go through rigorous lectures and work with the TA’s and Ph.D students and postdocs on the technical problems of AI. In the afternoon, the girls were divided into four research groups. And each of the research projects is a technical AI project — for example, computer vision or NLP or computational biology. But we put a very strong humanistic statement into each of the projects. For example, last year, we had four projects. The computer vision project uses depth sensors to look at hospital environments and help doctors and nurses to monitor hand hygiene scenarios. The NLP natural language project uses Twitter data during natural disasters, for example, earthquakes to — the girls’ aim is to do the right data mining to find messages that help to do disaster relief. And the self-driving car project, we designed an aging problem of a senior that needs to retrieve drops…

Sonal: That’s amazing.

Dr. Li: …and go there and come back. So, everything is very technical, but suddenly they learn that they connect these technologies to humanistic purposes. We have a team of three researchers. Two undergrad, one Ph.D student, and myself — we conducted a rigorous evaluation project on this hypothesis, can humanism increase the interest in AI? And we found a statistically significant difference from the beginning to — before and after for these girls’ thinking. And that particular paper is published in the computer science education conference to show this makes a difference.

Sonal: That’s great. It’ll be interesting to see what happens when you expand that to other groups.

Dr. Li: Yeah, we’re running it again this year. And we really hope that this can become a continuous program.

Sonal: Okay. Well, Fei-Fei, I’m excited to have you join us and bring all these perspectives to our own firm and the entrepreneurs we work with. And we’re so excited. Thank you for joining.

Dr. Li: Thank you.

• 

  Sonal Chokshi is the editor in chief as well as podcast network showrunner. Prior to joining a16z 2014 to build the editorial operation, Sonal was a senior editor at WIRED, and before that in content at Xerox PARC.

  Follow Twitter
• 

  Fei-Fei Li
• 

  Frank Chen is an operating partner at a16z where he oversees the Talent x Opportunity Initiative. Prior to TxO, Frank ran the deal and research team at the firm.

  Follow Twitter

Everything old is new again when it comes to startup ideas and how technology innovation happens. But practically, how does that apply to starting and/or working at startups — especially since the default state of every company is “dying in obscurity”?

In this episode of the a16z Podcast, Marc Andreessen and 21 co-founder Balaji Srinivasan cover everything from deciding what ideas to work on and the optimal type of startups to work at, to the funding environment and pendulum swings of deciding when to IPO. They also discuss the VC “formula” of weighting product vs. market vs. team; the full-stack approach to cracking industries that tech could never enter before; and recent tech trends and news including The DAO, AI, VR/AR and the “Instagrammification of everything”, more.

And where does Andreessen stand on the “moral dilemma” of whether entrepreneurs should drop out of college or not? Would Srinivasan still do a PhD today? People’s early career goals should be about maximizing learning skills and minimizing “personal burn”, they argue. But no matter what, Andreessen believes, smart people — from all industries, not just tech — should build things. It’s also easier to get through startup hard times when there’s an ideological mission motivating you, observes Srinivasan.

This episode is based on a May 2016 conversation that was recorded as part of the Annual Distinguished Speaker Series with Thought Leaders in Technology, hosted by engineering honor society Tau Beta Pi at Stanford University.

photo credit: Ryan Jae/ The Stanford Daily

Show Notes

• How Andreessen Horowitz chooses investments, and balancing team vs. market [0:18]
• Current trends in the startup space, delaying going public [7:40] and regulatory concerns [12:38]
• Discussion of new technologies, including Bitcoin [13:35], AI, and VR [20:37]
• Advice for current college students interested in forming companies [24:57] and an overview of how venture capital evolved [29:12]
• Questions from the audience [38:51]

Transcript

Sonal: Hi, everyone. Welcome to the “a16z Podcast.” I’m Sonal. And today’s episode features Marc Andreessen and a16z board partner and co-founder of 21 Balaji Srinivasan. This conversation took place in front of a group of Stanford engineering students as part of the engineering honor society Tau Beta Pi’s Distinguished Annual Speaker Series with Thought Leaders in Technology.

Balaji: Let’s talk about one of the things I’m sure every student here wants to — or not everyone, but a lot of them, you know, think about startups, think about technology, as an entrepreneur, as a founder, as a potential employee. How should students today, you know, graduating from Stanford think about startups as in both the founder context and an employee context?

Choosing companies for VC investment

Marc: Yeah. So the traditional — venture capitalists all have, like, a secret sauce kind of formula of how they think about what they want to fund. And then, it turns out, I think, the formula is all reduced to the same handful of factors, with the exception maybe of Peter Thiel, who has, like, six other factors in his head that he hasn’t told anybody about. But for everybody else, basically, it always reduces down to some combination of market, product, and team. If you talk to people who have been in venture for a long time, what they’ll tell you is, basically, the difference between venture firms, you know, in a lot of ways, is based on how do they rank the importance of market, product, and team.

You know, as an example, Sequoia was legendary in prioritizing market over team, right? So Don Valentine has — if you go online and Google Don Valentine’s talks, he talks a lot about how the key to success of a startup is to land yourself in a giant market. Like, land yourself in a market that’s about to become explosively large. And, basically, once the startup is in a position where it is the leading company in an explosively large new market, the people become somewhat fungible, like, you can swap the people out. And he would cite Cisco as one of the great case studies of that, which, you know, was actually a Stanford spin-off. A husband-and-wife team, very sharp founders, but they got booted very quickly, actually, by Don Valentine, who brought in professional CEO John Morgridge, who was phenomenal, and actually, you know, built Cisco the company. And so that’s one model.

The diametrically opposed model is prioritizing team over market. Basically, saying that, you know, the right market or whatever — can you really even know what the good markets are gonna be? Like, how well can you predict? Really what you’re doing is you’re going into business with people, either going to business with really good people or not. If you’re going into business with really good people, one of the things that should make really good people really good is they should be able to find themselves, you know, a good opportunity, right. A lot of startups end up, you know, they succeed based on something different than what they started doing. And so, if you get in business with the right people, they’ll be able to sniff out the opportunity.

Peter, I don’t want to put words in his mouth, but I think he’d probably prioritize product over market and team, which is — you have to be doing — you have to be making a fundamental advance in technology. He can tolerate a lot of flaws in the people, and he can tolerate a lot of uncertainty around the market. If the product breakthrough is big enough, he’ll make those other bets.

It’s kind of an angels-dancing-on-the-head-of-a-pin thing. You, kind of — as a VC, you sit around and talk about this a lot. And then if you want an investment company, you kind of figure out some rationalization, I guess, your formula to do it. So, I don’t want to overstate it, but I wanted to go through it, because I do think that is the framework, you know, as you think about startups as either a startup that you might start, or as you think about a startup you might go to. I think that’s a pretty good framework.

In terms of where — if you’re here as a student, if you’re gonna be graduating — my personal recommendation would be to focus much more on team. And the reason is just because I think we struggle from a distance to evaluate market, and we also actually start to evaluate product. But if you can get yourself in business with really good people, I think, number one — like, if it works, it’s great because those are really good people to be a business with, and they, with you, can build something great. But even if it doesn’t work, even if it’s the wrong market or the wrong product, you’ll still learn so much working with the right people, and you’ll build such a valuable network for whatever you do next.

It would also apply if you start a company. Like, who do you start the company with? You may end up in a situation where it’s like, do you start the company with a super genius who’s cantankerous and hard to get along with, or do you start the company with the person who’s, like, maybe not quite as incandescently bright but maybe is much more collaborative? And by the way, I don’t know that there’s a right answer. I do know it helps a lot early in your career to be working with really good people, because it really gives you a sense of what good really means and gives you the ability to learn.

Balaji: Yeah. I would say, one thing that we’ve talked about is that it should be exceptional, at least one dimension.

Marc: Yes.

Balaji: It can’t be, like, just pretty good and all these different things. At least one dimension needs to be, like, truly 10x and, you know, amazing to make the bet.

Marc: That’s exactly right. We talk a lot in our firm about — we have this concept — we say, “We invest in strength, not in lack of weakness.” And again, it’s one of these things that sounds obvious, but it’s proved to us to be a pretty big deal. So, there’s a lot of startups you’ll run into, or you probably have friends who are at these companies or know people at them, and it’s like, team’s good, product’s good, market seems good, they’re making some progress, they’ve got some customers, the customers are pretty happy. Okay. Where is that really gonna go, and where is it really gonna go? Because what’s spectacular about it, right? What’s the thing that’s gonna cause it to jump out from the other hundred, or other thousand companies where you can say the exact same thing?

So, then you say, “Okay, great. Now, I want to invest in strength. Okay, that’s easy.” The problem with investing in strength, or the problem with running a company, is that the strongest startups — at the point of contact, what you discover is the strongest startups aren’t strong at everything. They’re strong at something, and then they often have — the term we internally use, ironically, is they have hair on them. Which people are always kind of surprised when I start to use that metaphor, but they often have serious team issues.

Many successful startups have a founder divorce at some point. Like, literally, the founders go to war. And you would think that would be a very bad indicator, and actually, sometimes it’s a really good indicator, because it means that things are really starting to work, and like, it’s time to get serious. And one founder wants to get serious, another one doesn’t, or you’ll have these — some of our best companies are, like, stellar at product and engineering and cannot go get a deal with a customer to save their life, and like, labor for years under the illusion that the way the world works is that, you know, if you have the mousetrap, everybody beats a path to your door, and then three years later, they’re like, “Oh, we have to get salespeople to go sell things.” 

And so there’s these things, and they’ll just drive you nuts. But if the strength is strong enough, they can really punch through. And so much about this — another thing maybe worth saying is, the default state of every company is just dying in obscurity. And so, so much of that is, how do you punch through? How do you punch through in the minds of the people you’re gonna have to recruit? How do you punch through in the minds of the investors? How do you punch through in the minds of the customers? How do you punch through to the press? Like, how do you actually get yourself visible, such that you can start to attract the kinds of, you know, business, and momentum, and talent, and money that you need to be successful? And so that sort of model of strength versus a lack of weakness I think is pretty important.

Balaji: Every startup and every project starts as a hallucination, right? Like, it’s a word on a napkin. It literally doesn’t mean anything, and you have to believe it can become much bigger than it is. And always, at every stage, it has to become — you have to believe it’s bigger than it is.

Marc: Yeah.

Balaji: Okay, so…

Marc: That’s right. By the way, it means, in our business, if we’re doing something right, there’s something, basically, horribly wrong with every company we fund. One of the reasons, like, investment banks or the hedge funds don’t just come in and do venture capital is because they’re just horrified at every single investment we do. The one saving grace that we have with that model is, we have a portfolio. So, we get to make, you know, basically 30 grossly irresponsible bets, right, in our portfolio. And then, basically, the math is if we’re doing our job right, 15 work and 15 don’t. And in almost any other area of investing, or any other area of business, if you have that kind of failure rate, right, with that kind of risk level per decision, you would just throw up and go home. If there’s one edge that we have, it’s the ability to kind of indulge in these situations where the strength is crazy, but the weaknesses are also frankly crazy.

Balaji: Yeah. I mean, like, the thing is, if it gets de-risked all the way, then it’s just a safe investment and there’s very little upside. But I think it also holds for technology, in the sense that, if you read about something in the Wall Street Journal or the New York Times, and technology is on everybody’s lips, it’s probably — not always, but it’s probably started to, you know, have some of the value taken out of it, in the sense that there’s a lot of companies that already built in the space, it’s very competitive, and the technology to look for are often the ones that haven’t got a lot of press yet, you know, that are near inception that are in the labs of places like Stanford.

Marc: If it’s a buzzword, if it’s something that’s on people’s lips, if there’s magazine articles about it and newspaper articles about it, or, God help us, if it’s on TV, like, the time has passed. Like, we better look for something new.

Current trends in startups

Balaji: So related to the subject we just talked about, how people should think about pursuing startups, what does it mean for — so, folks who are, you know, employees, what does it mean when companies stay private longer? And what do you think of the root cause of this relatively new phenomenon, really the last 10, 15 years or so?

Marc: So, the model for Valley startups, right, used to be very straightforward, which is you’d raise an A round, and then you’d raise a B round to kind of build out your sales force once the product started working. You raise a C round to maybe expand in a couple of other countries, maybe do a little acquisition or something. And then, within, you know, four, five, six years, get to about, you know, 30, 40, 50 million in revenue, and you go public. It was sort of, you know, that was sort of the rite of passage. And then a bunch of things became possible once you were public that weren’t possible before. So, one was liquidity, which is — early investors and employees could start to sell stock. But there are other very important ones. One was, it was viewed as a legitimizing event, especially for companies that sell products to other companies. It was viewed as an event that basically was, you know — a lot of big customers of technology would much prefer to buy technology from public companies, because they feel like they can understand the vendor they’re buying from, whereas these private companies, they don’t know if they’re still gonna be a business or not.

And then, also, M&A, mergers and acquisitions, you know, it was considered a great virtue of being public — is to have an acquisition currency, right, to be able to issue stocks, and a lot of the great tech acquisitions over the years were done with stock because, you know, you get <inaudible> and go public, and you can use that value to buy things, even if you don’t have the cash. The stereotype is that everybody wants to go to work for a startup in the Valley. I think the reality is, a very large number of people actually don’t want the true early-stage risk. They want to go to a company that’s doing interesting things, but they don’t want to have to, like, go look for another job in six months if something goes wrong, because they’ve got, like, a family. They’ve got, like, a spouse, and they’ve got a mortgage, and they’ve got kids, and they’ve got bills they have to pay. And so there’s actually a lot of talent that got unlocked, once you became public, that you could actually recruit. And so, those were the old days.

Interestingly, in the U.S., the number of public listed companies in the U.S. peaked in 1997, weirdly enough. And you might think it peaked in, like, 2000 or 2002, or something, but it actually peaked in ’97. And basically, the number of public companies in the U.S. has now dropped by two-thirds since 1997, and that has coincided with a bunch of other things. I mean, one was, you know, we had the stock market crash, and then we had the credit crisis — but it’s also coincided with some other changes. One of the big changes, for example — a lot of tech IPOs actually were individual investors, right. A lot of historical investors and small tech companies were individuals who would read about these things and get excited and invest. If you just look at the statistics on this, the percentage of ownership of tech stocks by individuals has dropped like a rock since 2000. It’s basically now all funds, right, and funds are inherently more conservative than individuals, because funds have, you know, they feel like they have a responsibility to be sober, and so they’re not that excited about the next hot IPO.

And so, the public market, like, just a lot of the enthusiasm has been drained out of it. The market has changed dramatically. And so, it’s sort of, you know, to Balaji’s question, it’s kind of become in vogue, or in style, to either not go public or at least not go public as fast as before. The good news about staying private longer is that there is something about going public that puts you on a treadmill with quarterly results. They’re like, “Well, you know, I’m not gonna get on this treadmill with quarterly results where I have to hit all these quarterly earnings targets. I’m still gonna be able to do long-term things.” So, the good news about staying private is that you can do these big ambitious projects over long periods of time. And you know, you either get them right or you don’t, but you’re not under any specific quarterly pressure to deliver any particular set of financial results.

My view is that the pendulum has actually swung too far now in the direction of not going public. Like, too many companies are now staying private too long. It used to be that it was a contrarian view that you should stay private. It’s now become a contrarian view that you should go public. And my argument of why more companies should go public is, at some point, it’s good to not just have all of your results be in the future, but to actually have to deliver in the present. And at some point, it’s good to have an organization that actually, like, knows how to work properly, and knows how to sell things to people, and knows how to, like, have financial plans and hits, and knows how to make money. And it’s all hypothetical until you have to prove it, and I think a lot of companies that are staying private for too long risk getting sloppy and undisciplined. And in the beginning, that’s fine, but at some point, you have to get serious. And if you can go for 10 years without getting serious, I think there’s a real risk that you never get serious. So that’s one.

And then number two, you know, it’s become massively differentiating to go public, because you get these big advantages. You still can then tap the public markets for more money. People talk about Elon Musk, and you know, SpaceX is still private, but Tesla is a public company. So, Elon Musk puts out this thing, the Tesla Model 3 — the pre-orders, and it gets half a million pre-orders, all of a sudden. Everybody hated Tesla before, because nobody wanted to buy the car. Now, all the investors hate Tesla because, now, there’s too much demand for the car, right, which is apparently equally bad. And so, he just now said he’s gonna do a $2-billion secondary offering, right, in the stock market, and like, even in modern, like, venture capital, it’s hard to raise $2 billion at a shot. Not very many people can do it. And so, he can actually, like, raise that amount of money publicly. He can access debt. And then, you know, you go back to the acquisition currency. Like, we’ve probably been in a slow period for M&A for a while, but there is no question. There’s gonna be a lot of M&A in the years ahead, and the companies that have public currencies, they’re gonna be able to be the acquirers and able to get big and become much more important. So, I think the pendulum is gonna swing back in the other direction. There’s a crop of companies, good companies definitely gonna go public.

Balaji: I think another part is also Sarb-Ox, and all the rules, and then Dodd-Frank, and so on, has made it quite difficult to be a public company from a compliance perspective, and the fixed cost associated with that.

Marc: Yeah. So, there’s this thing, Sarbanes-Oxley, which I see somebody in the audience yawning, and this topic is gonna make everybody yawn, and so I’m not gonna go into detail. You can Google it if you really want to learn about it. But it’s the regulatory, kind of, threshold that public companies need to hit on how they deal with risk and do reporting, and all this stuff. And the knock on Sarbanes-Oxley has been exactly what Balaji said, which is it’s basically a burden that falls disproportionately on small companies, because big companies have huge staffs of lawyers and finance experts, and so forth, who can do all this stuff, but small companies, the burden falls directly on the management team.

Our partner, Ben Horowitz, now argues the opposite side of this, having seen a lot of companies — which he argues, if you’re good enough as an operating team to actually comply with Sarb-Ox, then you’re good enough, basically, to do anything. Like, basically, not everything in it makes sense, but it sets a bar for what it means to be an operating business that’s operating in a responsible way. So, I think he’s actually flipped a little bit on that, and I think he would argue it’s actually part of being a responsible company at some point.

Future of Bitcoin, AI, and VR

Balaji: Interesting. It actually kind of gets into our next question. We’re gonna talk about a few important technologies. One thing that I’ve thought a lot about is that the ultimate, kind of, solution to this is gonna be something related to the Bitcoin/Ethereum crowdfunds that are happening now on the internet, where the regulatory stuff has to be worked out about that. But you do have a very large potential pool of capital that people can use for this kind of thing, and that might be, you know — it’s is an essay that Naval and I wrote a couple of years ago about, like, an app coin. So, you’d actually start a company and actually issue a coin that could be used to redeem for calls of that SaaS service. So, that’s one model that might help.

Marc: You might just mention — this is a whole new model for how to think about, sort of, crowdfunding taken to another level. You might just mention the DAO and what that is.

Balaji: Yeah. So, this is a pretty interesting concept on where — so Ethereum, it’s something that was based on Bitcoin, initially, and is sort of like a more programmable version of Bitcoin in some ways. There is a thing called the DAO, which raised almost $130 million online in a purely distributed way, just with digital currency, without any stock market or what have you. There’s all kinds of regulatory hair on this animal, and people can pull their money out of it. So, it’s sort of like a VC fund, where the LPs don’t actually commit until they see the first investment. So, I think there’s gonna be all kinds of stuff that happens with it. Nevertheless, I think it’s a very interesting experiment, and something which will probably be relevant for you guys, not this year, not next year, but in maybe 5 to 10 years, in terms of potentially an alternate way to get financing for something. So, actually, that leads us into important technologies, right? So, let’s get a quick riff on them one by one. So, starting with maybe, you know, talk about Bitcoin and blockchain, then FinTech more broadly.

Marc: Yeah. So I’m gonna turn the first one around. So, Balaji is the founder of one of our two big Bitcoin investments, so.

Balaji: Sure.

Marc: Balaji, how’s Bitcoin doing?

Balaji: How’s Bitcoin doing? Yeah. So, you know, like the Gartner Hype Cycle, right, something we think about a lot. We think of it as this fundamental thing in technology that is — you’ve got this trigger, and then people get really amped about a technology, and everyone’s doing it, oh, you know, bots are at that stage right now. And then you try to actually do it, and you find it’s actually hard, and everyone gets demoralized, and they quit. And you’ve got the trough, and then it’s those guys who stick it out in the trough and pull up over here that, you know, things actually happen. So, that happened with, like, the dot-com bubble. Everyone was hyped about it in 2000, it crashed. And then, actually, you built all these massive businesses. And it happens on, like, larger and larger cycles as well. Carlota Perez — she’s got this whole theory about why that happens. And it, kind of, happens at different scales. And we, sort of, think that’s happening for Bitcoin in the sense of, you know, there’s a huge amount of excitement like 2013, 2014, you know, “Oh, my god, new paradigm.” Then, you know, like, “Oh, the price crashes.” And now it’s coming back up with a lot of, like, micropayment stuff, interesting things happening this year.

I think the blockchain stuff is actually right at the top of the Gartner Hype Cycle, and I think it’s gonna crash down, like, towards the second half, you know, of this year when people actually try to implement it. That’s where I kind of think Bitcoin and blockchain is, and I would say that, you know, in addition to our kind of point earlier about, like, you know, getting technologies that nobody knows about at all, that are in the lab right now. I think other kinds of technologies to really look at are those that people have written off, right, like, you know, VR after Second Life. And so, that’s the kind of thing to look for — the stuff that people think of as, you know, dead or didn’t work, or what have you, and find out why.

Marc: It’s actually very funny. You don’t remember the first time VR got written off.

Balaji: Oh, no, that’s true.

Marc: You only remember the second time it got written off.

Balaji: I remember the second time it got — yes, that’s right.

Marc: No, actually, you remember the third time it got written off. I remember the previous two. It got written off after VPL. It got written off after the VR — there was a whole VR wave in the late ’80s — one of the great all-time hacker movies, “One More Man.” It was kind of a peak of that cycle. And then we bought a VR company, Netscape, in ’95 to do VR/ML, which is VR on the browser. You may note that that didn’t work. And then, right, there was Second Life, which was, like, the third cycle.

Balaji: Right.

Marc: One of the things we talk a lot about is, say, two operating principles in how we think about technology. One of the things I’ve come to believe — there are almost no actual new ideas, right. Basically, everything that is gonna be a big deal in the next 30 years is in a lab somewhere, probably here in a lab at Stanford. And so, the eureka moment is, like, an almost non-existent thing. Maybe every once in a while, but there’s almost always a 20- or 30-year backstory of research that often, by the way, turns out to be 50, 60, 80 years backstory of research before something pops. And then the second thing is just, yeah, things take time. There’s this concept called the AI winter, and literally, there have been surges of enthusiasm and crashes in AI. And I think we’ve counted there were, like, 5 AI winters between 1950 and basically 10 years ago.

Balaji: Even the term AI has only come back recently after neural networks themselves came back, because everyone was like, “Oh, AI is all rule-based, and ML is the new thing.” And [we’re] having another mini-cycle within that where, like, Chris Dixon and I joke that so many AI companies are just a collection of if-else statements. And you know, it’s like, “Okay.”

Marc: Which are very compelling on first demo.

Balaji: Very first, yeah, but it’s always on rails, right? And then when you try to get it a little bit off, then it’s like, “Cannot compute. Great.”

Marc: Yeah. And so I think, Balaji, that’s a very important kind of fundamental point, which is it’s not — I mean, what’s new is important, but it’s often what’s new where there is a track record of intellectual depth that’s gone into it over a long enough period of time that people really have thought hard about it. And it turns out, that track record is almost always multiple decades. And then, whatever happens to be hot or not in any particular moment, is really not predictive of what’s actually going to happen.

Balaji: Exactly. I think, you know, in particular, there’s two things, if you ask me, you know, what, like, to look at for startup ideas, and so on. So first, I’d say, don’t do a startup unless you’re ideologically driven to make it succeed beyond the economic motivation, because it’s actually very hard. But if you do want to just find startup ideas, there’s this book, “The Sovereign Individual.” It came out in the late ’90s. It’s the most prescient thing in the world. Most bestsellers, you can take the 300 pages and compact them into, like, a one-page summary, and there’s actually websites that do that, right? Whereas, this book is the opposite. You can take, like, a page and turn it into a Ph.D thesis. And what’s awesome about it is, you know, we kind of think Satoshi read through “The Sovereign Individual” and actually made Bitcoin, in part, on that basis, because the description of it is so lucid. But what’s interesting is, there’s other pages of it which haven’t yet been implemented. So it’s like, the “Book of Prophecies,” and you just flip through it, “Oh, let me do that line,” right? So…

Marc: So then the kicker of, you know, that book ripped off another book, an older book.

Balaji: What’s that?

Marc: It’s an older book called “The Twilight of Sovereignty.”

Balaji: Interesting.

Marc: Which was written by a guy named Walter Wriston, who was the founder of Citibank, who spent 40 years in banking, 40 years in, like, big New York institutional banking, and his conclusion at the end of it was, it was all bullshit. And he basically wrote a book predicting, basically, the rise of networks and distributed finance, distributed money. This is like 30 years ago.

Balaji: Yeah. So, I mean, what’s interesting is, a lot of those guys got the general direction right, and then there was some aspect that actually turned out to be much more difficult than they thought. For example, like, autonomous robotics. Well, actually, that’s really hard because of the number of degrees of freedom and the probabilities, but it’s doable with enough training data. I think the other thing that, you know — I think of it like a “Back to the Future” thing that’s very important — is this thing called Tiebout sorting. So, like, a while back, we found this guy who’d done it in 1956, and he had a bunch of assumptions for this model of how people could sort into, like, basically, many governments around the world, and he assumed like, “Okay, you have search. You have perfect information. You have perfect mobility of this, you have that.” And he basically, like, assumed the smartphone. They wouldn’t have put it that way at that time, but 1956, he assumed the smartphone is like, “Oh, wow, you can solve all these problems with governance and so on.” So, like, literally 60 years later, you can go back, you know, dust off this “Raiders of the Lost Ark” stuff and just, you know, go with it, right? And you’ll sound really smart because you can just, like, read off the “Book of Prophecies.”

Okay. So, other important technologies, all right. So AI, right? We just kind of talked about this a little bit. So, autonomous cars, drones, ML, and software, what is your take on this?

Marc: Yeah. So, magic is happening, and I think everybody here probably knows this by now, but something has changed. And actually, what that something is, is a matter of some debate, and it’s probably multiple somethings. But an entire battery of techniques that people have known about for a long time, plus some new techniques in machine learning and deep learning have really started to work. 2012 was kind of the tipping point for that. And now it’s really building steam. And then it also feels like something changed —.part of the passage of time in our industry is just Moore’s law, allowing processors to kind of catch up with our ideas, and the rise of this new generation of GPUs that are able to run neural networks and deep learning algorithms is a really big deal. And then, you know, we now have existence proofs of, you know, fully running autonomous cars using deep learning. We’ve got autonomous drones with deep learning. We’ve got, you know, AlphaGo, the great accomplishment that Google recently had, that DeepMind had. Like, significant breakthroughs are happening. I would say something both very dramatic happening, but also something very real happening.

Balaji: Yep. I would add to that. Actually, just data. Like, because, you know, like, many of these algorithms you just put 10x of data at them and they work, and 1/10 of them don’t. And so, like, just the ease of collecting massive amounts, right?

Marc: Yeah.

Balaji: So, VR and AR. So you know, Oculus and Magic Leap, and stuff like that, what are your thoughts on that area?

Marc: Yeah. So, very exciting. So VR, right, is the idea of the headset that you, basically, are in a completely computer-generated world. I’d like to say, the world’s now divided into two groups of people. People who haven’t tried the shipping consumer version of Oculus, who think VR is stupid, and then people who have tried it, who think it’s the future of everything. And so, if you haven’t tried it, find somebody — they just started shipping. Find somebody who has one and try it. It’s a really profound thing.

The other idea people are playing with is augmented reality, or AR, which is the idea of — you still see the real world, but you have computer-generated imagery kind of populating it. And there’s a company called Magic Leap in Florida that’s doing this, and Microsoft has a thing. We actually argue there’s two kinds of AR. There’s the kind that people are talking about, because they find VR too scary — and that’s why all the news articles on VR are all, like, very emotionally loaded, because it’s invariably a picture of somebody with this thing strapped to their face, right? You don’t actually get to see what’s inside the VR. You just get to see the idiot sitting there in the chair with, you know, the alien Facehugger, like this, and then everybody thinks it’s funny. To a lot of people who find VR too weird, AR feels like it must be more normal, because I still get to see everybody — and I think it’s actually a little bit of an intellectual crutch for people who just can’t quite come to grips with VR.

That said, there’s the other form of AR, which is, like — if we can get AR to really work, right, and if we can get to the vision that I think everybody in the industry has, which is — get a pair of, you know, very light eyeglasses or, even better, contact lenses that overlay computer imagery on the real world. Like, that is a big deal. There are teams — there are a handful of companies now that have teams that are super focused on this.

Balaji: Two thoughts, one on AR and one on VR. One thing that I think about AR is if that kind of thing can work, I think you can have what we think of as, like, the “Instagramification” of many more things, in a sense of, what is Instagram? So, yeah, it’s a photo app, but then it also is something that takes somebody who has no skill in photography and gets them to, like, an eight, because you got a programmer on your shoulder, and you know, he’s like, “Oh, put the f-stop there and whatnot, and don’t generate, and so on.”

Marc: There’s always at least one filter that makes any photo look good.

Balaji: Exactly, that’s right. No, I actually think, like, the next version of Instagram will make people prettier, right? Like, I call it Tinder for Instagram. So…

Marc: Just keep swiping until you get attractive enough.

Balaji: Well, yeah, exactly. You just got a filter that just morphs it just a little bit, right?

Marc: It’ll come in handy.

Balaji: Exactly. The thinking is, though, that Instagramification — you could apply to many other areas with AR, right? Like, so the classic examples are you’re a mechanic, and you put on the glasses, and now, you know, every part lights up, and you see the 3D schematics, and you tap here to order the replacement from Honda, and so on. Or you’re a surgeon and you can actually see the person’s x-ray superimposed on them. And so, it’s like you’ve got a superpower, right, in that sense. Which actually, you know, <inaudible> a while back. 

And then, on the VR end of things, you know, one thing when people, you know, kind of dismiss VR, I always ask them, “Okay, how much time do you spend looking at a screen? How much time do you spend looking at, like, a laptop or a phone?” And they’ll say, you know, “Okay, maybe, you know, six hours a day.” And so, I’ll say, “Okay, well, that’s like 50% of your waking hours.” And we’re probably gonna replace a significant percentage of monitors with VR, with something to the 2D world, right, and there’s gonna be a new Windows that’s based on the 3D universe, which has totally different GUI metaphors. So, that’s an interesting kind of company to build that doesn’t exist yet. But that company — okay, so when you’re wearing this VR thing to do work, not just to play video games, well, actually, most of your life is in the matrix. So, that’s gonna be kind of interesting in, like, 5 or 10 years. Everyone’s wearing these kind of things.

Marc: It’s coming.

Advice for college students

Balaji: Great. Okay. What should Stanford students be thinking about doing after graduation or, dare I say, instead of graduation? That’s question number one. And then related, what advice would you give if you’re at Stanford right now? And what should a student walking down this hall do right now?

Marc: Yeah. So, I used to — people used to ask, you know — so, obviously you’ve got — the example is of Mark Zuckerberg, and all these founders who dropped out, and so, therefore, you know, everybody should drop out and start a company. And people used to ask, you know, “Should I stay? Should I drop out? What should I do?” And it used to be a very — I used to feel, like, a real moral challenge answering that question, because I felt like, if somebody really should drop out and start a company, and I tell them not to, I’d be committing a moral crime. But most people probably should stay in school and actually get degrees, and it feels immoral to suggest otherwise. So, I felt trapped. I thought about it. And the absolute straight advice — 100% of the time, you should stay in school, finish your degree, not drop out. And I’ve concluded that because the people who are gonna drop out and start a company are gonna do it regardless of what I say, or what anybody else says. And so, by definition, it’s good advice. I can’t possibly steer anybody wrong.

In general, actually, not only is it a good idea to get the degree. The thing that it’s the most underrated right now. I think the archetype/myth of the 22-year-old founder — it’s been blown completely out of proportion. The thing that is underestimated now in the Valley — and, frankly, Stanford is the ground zero of this — I think skill acquisition — literally, the acquisition of skills on how to do things — is just, like, dramatically underrated. People are overvaluing the value of just jumping in the deep end of the pool, because, like, the reality is, most people who jump at the deep end of the pool drown. Like, there’s a reason why there are so many stories about Mark Zuckerberg. It’s because there aren’t that many Mark Zuckerbergs. Like, most of them are still floating face down in the pool. And so, for most of us, it’s a good idea to get skills, you know, your degree or whatever, but then there is a lot to learn.

If you want to, like, ultimately start a company, or go to a startup, there’s a lot to learn about how companies operate, right? There’s a lot to learn about how to deal with people. There’s a lot about how to manage. There’s a lot about, you know, leadership. There’s a lot about, by the way, finance. There’s a lot about legal. There’s a lot about marketing. There’s a lot about sales, HR. Like, there’s a whole skillset. Like, if you meet, you know, the really great CEOs, if you spend time with them — and you would find this to be true of Mark today, or of any of the great CEOs today or the past — like, they really are encyclopedic in their knowledge of how to run a company, and it’s just very hard to just, kind of, intuit all that in your early 20s. And so, I think the path that makes much more sense for most people is to spend 5 or 10 years getting skills. So, the problem with <inaudible>, it sounds great but, like, most startups are, like, really screwed up. Like I said, most of them just die in obscurity. And I don’t know exactly what you learn from dying in obscurity, but it’s not very much. A lot of people are at startups that don’t work well. They actually don’t carry away a lot of useful skills.

Conversely, you know, you leave school, you go to a big company. A lot of what you learn in a big company is how to function at a big company, right? But the problem with people who have been at a big company too long is, in the cold light of day, when they go off to do their own thing, they literally don’t know how to function without all the infrastructure and support of a big company. And so, I think there’s a sweet spot, like a new high-growth company or the company that’s scaling. That’s probably the best place to go. And of course, you’re at Stanford, you have a huge advantage of being in the environment. You already know who those companies are, and, you know, you have a pretty good chance of getting jobs there. So, I think that’s generally really good advice.

The other thing that I would say is, I have a favorite book I’ve never read, and actually, I’m worried about reading it because I think it can only disappoint me at this point, because I like the title so much. And the title of the book is “Smart People Should Make Things.” And like, as far as I’m concerned, like, that’s the entire value of the book. Like, I don’t even care what else he says. Like, just for engineers, it’s very obvious. Like, engineers should build things, should build products. And that could be open source, it could be, you know, working with a company or with a friend on something, but, like, going to a company that’s building something. But I think the same thing is true of everybody else, right, and people build all kinds of things. And by the way, the things that people build might be art, right. The things that people built might be, you know, businesses. The thing that people built might be an organization inside a company, or it might be a great explanation of something, but tangible output. I just always kind of really encourage people, like — when in doubt, fall back on building something tangible.

Balaji: Yeah. And, like, we’ve got that thing at Andreessen Horowitz, right, like, works in practice, not in theory. So much stuff that I saw, you know, as a scientist, a Ph.D at Stanford, worked in theory but just not in practice, and there’s lots of stuff that’s just the converse, and only if you actually build it can you see it. Why did you and Ben, then, decide to start a VC fund rather than doing another startup?

Evolution of venture capital

Marc: Yeah. So, we were customers of venture capital, or at least I’ve thought about it that way. They thought they were giving us the money. I thought we were the customer. We had maybe occasional disagreements about that. And so we were customers of venture capital. I first raised venture capital in 1995, with my partner Jim Clark, from John Doerr — who was, actually, you know, an excellent VC for us at Netscape, and then we raised money from Benchmark in ’99 for Loudcloud, and that went really well. And then, between Ben and I, we also helped probably 100 friends of ours over the course of, sort of, a 15-year period, raise venture capital. You know, we were angel investors. We would help our friends go through it. And so you kind of view it, like, as almost going to the same department store every day for 15 years or something. After a while, you’re like, “You know, I think maybe I could do this, and I think maybe I have a few ideas from being on that side of the table.”

So, we started really thinking about entering the business, and then we thought really hard about, you know — the traditional way to enter venture capital is to join an existing firm, because the history of venture capital is that the successful firms have all been around for 30 or 40 years, and we considered that. And then we basically got bit by the startup bug — me for the four-and-a-halfth time — and we decided that it was actually a good idea for a startup. We spent about a year and a half actually thinking about Andreessen Horowitz as a startup, and we spent a lot of time studying the models and talking to people who had been in the industry for a long time. And we ultimately resolved on what we thought could be two big differences. One was actually a little bit of a “Back to the Future” thing, which is — we decided that the general partners at Andreessen Horowitz would all be people who have been founders, or CEOs, or both, of tech startups. And, that kind of sounds like it might be obvious. Like, if you’re gonna have somebody on your board, and they’re gonna give you advice on what to do in your company, that maybe it would be helpful if they had actually done it before.

It actually turns out, first of all — it had been a good idea in the ’60s and ’70s. The top VCs in the ’60s and ’70s, when venture capital was created had, for the most part, all been operators, and they had been legendary characters. Gene Kleiner had been famously one of the Fairchild, one of the original Fairchild people, one of the famous “traitorous eight,” who left Shockley to start Fairchild, left Fairchild to start Intel. Tom Perkins had actually been a general manager at Hewlett-Packard, which was actually, at the time, a source of a lot of the CEOs of the new companies in the Valley and actually, himself, had been a founder. He started a laser company, which was the kind of thing people did in the 1960s, and he actually raised venture capital himself and was a founder. Don Valentine. You guys had, I think, Mike Morris here last year. The founders of Sequoia Capital, Don Valentine and Pierre Lamond, both of whom are famous chip executives and entrepreneurs. And so, it actually was how venture capital got formed. Our analysis was basically, over the course of time, venture capital — a lot of the traditional venture capital firms had evolved where the successors to the founders were, in many cases, very successful investors, but were people who had not started and built companies themselves. And so, we kind of decided to bring that idea back.

The other big idea that we had, that we’ve really pushed hard, is the idea of giving founders, and especially founders who have not been CEO before — we would use the term, sort of, give the founders superpowers — in the form of, basically, the world’s best network. And this is an observation that, you know, we’ve seen over the years. We’ve seen founders start companies, and then, at some point, the founder gets fired, and you bring in a professional CEO. One of the questions we always had is, what’s the catalyzing thing that causes the founder to get fired? And then what is a professional CEO? And professional CEOs, it’s always a type, right? It’s always like, you know, square shoulders, blue suit, six-foot-two, gray hair, fantastic teeth. Like, it’s a type. And what do these professional CEOs have that the founders didn’t have? And actually, some of it is, they have experience running a company, and we think we can help with that. But the other part is, they have these networks. They have been in the industry for 20 years, longer. They’ve got 20 years’ worth of, basically, network built up, right, and so they know customers, and they know other investors, and they know all the big tech companies. And if the company is to get sold, they know all the buyers, and they know all the reporters who cover the space, and they know all — if it’s a regulated business, they know all the government regulators. And so, they have these giant networks that they built.

So, what we decided to do in our firm is, basically, essentially, pre-build the best possible network that any startup could have, and then basically let our founders plug into it, and basically get the superpower of having a giant network. The way that we did that is we actually have — we have a very kind of nontraditional structure. We have full-time professionals in our firm who are not general partners or investing partners, who are operating partners in six teams that build and run networks across categories, customers, investors, acquirers, executive talent, engineering talent, PR, and now, policy and regulatory affairs. So, we’ve got 85 people in the office every single day, and what they’re doing is they’re basically building and grooming a network on behalf of the firm, which then works on behalf of all the portfolio companies.

Balaji: Andreessen-Horowitz is actually a network as a service.

Marc: Yeah.

Balaji: So then, one interesting point is — a16z was actually started in, you know, ’08, ’09, and it’s been, like, 7 years now, right? And the industry has changed, you know, the firm has changed, VC, more broadly, has changed. What are your thoughts on, kind of, that evolution?

Marc: I would say there’s been more change — there’s no more change in venture capital in the last 7 years than probably in the preceding 20. And I’d also argue there’s probably been more change in the tech industry in the last 7 years than probably the preceding at least 15 or 20. There’s a bunch of new firms now that people are starting that are exciting. Another thing is seed investors — angel investors have always been important. Like, a big part of the history of the Valley is the willingness of people who have made, you know, some amount of money to write a check, and sort of fund the next idea. And you know, a lot of the original companies in the Valley, there was angel money involved. So, angels have always played a very critical role. In the last seven or eight years, it feels like a lot of the angels actually have professionalized, and when they do that, they renamed themselves — angel investors, to seed investors — because angel kind of implies an individual, whereas seed kind of represents a sort of investable asset class. And so a lot of the best angels have now actually raised funds, instead of just investing out of their own pocket, and they actually run these seed firms.

And so, actually, we see kind of a restructuring happen in the industry where a lot of companies — companies used to just raise venture capital as their first round. They’d just go straight and raise a series A. And you could either raise a series A or you couldn’t. But it was only a very small percentage of founders who could raise an A round right out of the gate. You know, these days, it’s much more common to raise the seed round, you know, raise $500,000, or $1 million, or even $2 million as a seed round, and then go for a year or 2 or 3 — well, before you actually have to raise full venture capital. In fact, the seed phenomenon has now gotten so widespread that, now, the seed investors are trying to differentiate against each other. So now, there’s seed. There’s also pre-seed. There’s also seed extensions. There’s post-seed. There’s early A. And then, actually, below all of that, there’s incubator, accelerator kind of phenomenon. And so, we’ll actually sometimes meet companies that have raised, like, five rounds of seed capital in different forms. And so there’s just a lot more support in the infrastructure for a much larger number of new companies.

I think that maps to what’s happened in the industry over the last seven or eight years, which I think is really remarkable — either we’re just taking it for granted or we haven’t really wrapped our heads around it. Which is, the history of the Valley for 50 years, from the 1960s through the mid-2000s — the Valley was kind of the best place in the world building, literally, computers — so chips, and then computers, and then software that runs on computers, but fundamentally building tools, right. Computers or software as tools. And then, you know, these giant companies, Oracle, and Sun, and Cisco, and so on, would build these great tools and then would sell them to customers. And the customer might be a consumer at home, but the customer, more often, was a big bank, right, or a big insurance company, or you know, a hotel chain, or somebody like that — or a car company.

In the last seven or eight years, post the financial crisis, something has changed. Either the Valley is about to grow to become a lot bigger and more important than the Valley has ever been, or we are completely smoking crack. Many Valley companies still build technology and sell the technology as tools, but a lot of the best new Valley companies build technology and use it as a wedge to enter an end market, right? And so, as an example, the predecessor company to Uber was not, you know, a ride-sharing service that failed. The predecessor company was a little boutique software company that built dispatch software that got sold to taxicab operators, right? And there actually were companies that were in that business, it’s just, it was a tiny little business, because it turns out taxicab operators actually aren’t that excited about adopting new technology, they don’t buy very much IT, they don’t buy very much software. If they did buy software, they wouldn’t know what to do with it. And so, that was just never a very big business. And so, Uber and Lyft just come in and basically say, “Let’s just do it. Let’s just provide the ride. Let’s take complete responsibility for the customer service.”

Elon Musk, of course, has pushed this to its logical conclusion, which is, you know, why not just build the car. I think that Elon gets tremendous credit, both for the car company and the rocket ship company, both of which are things that — nobody 10 years ago thought was possible to build either kind of thing as a new company, and it turns out that it is. It feels like the Valley is really expanding, basically — certainly expanding in ambition, and quite possibly, we believe expanding in capability, to be able to actually go directly into a lot of markets that historically you would have viewed as, you know, much more the province of existing banks, or existing car companies, or existing incumbents.

Balaji: I think a big part of that is actually the fact that, if you’re selling IT to somebody, versus actually using it yourself, you can just recognize the benefits, you know, more obviously. Like, oh, if you’ve got your entire thing in a database, well, you can push out, like, a report of all ride times, and so on, and so forth. And they can understand and think about data, but the customer wouldn’t necessarily do that. It’s a major efficiency.

Marc: If you’re selling technology to a company that’s then implementing it, it’s a layer of indirection. And there are companies — I mean, look, there’s, you know, Oracle got built to do this, and a lot of Oracle customers have gotten great results with Oracle. And salesforce.com just had a great quarter, and you know, they sell their stuff to lots of companies with big sales forces who do great with it. So, it works. But, yeah, we see this — we have this sort of, like — the term we use is full stack. Which is, you sort of see there’s a particular magic, exactly to Balaji’s point, there’s a magic that kicks in when you actually have complete responsibility for the end customer experience, and how the product or service is delivered. And then, especially these days, right, in the era of big data and machine learning, and all these things, there are things that you can do to optimize both experience, and then ultimately the economic model of the business. It’s become a very open question or a topic. Okay. So, how many industries are opening up where you could possibly do, you know, the equivalent of an Uber, Airbnb, or a Tesla, and these industries from the Valley?

Balaji: I guess, let’s start taking questions, yeah.

Audience Q&A

Man 1: Hi. So, for a first-time founder who’s bootstrapping a v1 product, when do you think is the most appropriate time to first approach investors, and at what level? Is having a business plan and a team reasonable, a prototype to show potential, or demonstrable customer traction? Thank you.

Marc: Yeah. So, it’s hard to give general advice because it really depends, but unquestionably, it’s better to have something working. Coming in with something working is a gigantic edge over coming in with nothing working, like, a huge edge. Even, by the way, for people who have done it before, people who have successfully run companies before, coming in with something working is a really big deal. And then it is, like, absolute magic. I mean, it’s like catnip to VCs if you can walk in and you’ve already got both the product and customers. Just rub it on us and it’ll drive us crazy. And this is another thing. Probably what’s overestimated right now is just raising lots of money — to be able to say you’ve raised lots of money. Probably, what’s underestimated is the bootstrapping process of getting in position with the core thing that you’re doing, and both the product itself and its value to customers, before you start raising a lot of money.

Man 1: And with that customer traction and MVP all ready, like, what level angel seed A?

Marc: If you’re a first-time founder, first-time founder, it’s almost always better to start with angels or with the early seed investors. It’s, again, contrary to myth and archetype. It’s very hard for the first-time founder to raise a straight A round. It’s almost always the case that they’re coming up through a seed. I mean, as an example, you know, Mark Zuckerberg raised literally angel money from Peter Thiel. That’s how he got started. He didn’t go and raise an A out of the gate. Sergey and Larry, same thing, they raised angel money. And so I think that that’s almost always the best thing for a first-time founder.

Man 1: Thank you.

Marc: Yeah.

Man 2: You mentioned all the progress in AI, a new input, output, and all the language process. So, I have a very — if you have to pick, in 30 years, what’s the chance that we have a bot that does a better job in picking companies than Andreessen Horowitz?

Marc: I hope to God we invest in it, because it’ll be the last investment we ever make. So, I mean, this idea is out there, right? And so, there are actually people literally trying to do this, and there’s actually a venture firm called Correlation Ventures that literally is trying to do this, or a version of this. And then, you know, there are people who are, like, data mining angel lists, and trying to figure out how to do this. And there are other people who are going about this. The computer scientist in me, the engineer in me, would like to believe this is possible, and I would like to be able to figure it out, and I’d frankly like us to figure it out. The thing I keep running up against — the cognitive dissonance in my head that I struggle with — is what I just see in practice — talk about in theory versus in practice. Like, in theory, you should be able to get the signals, like, you know, founder backgrounds, and this, and that, progress against goals, or whatever, customer satisfaction, you should be able to measure all these things. We just find, what we deal with every day is not numbers, right? There’s nothing to be quantified.

What we deal with every day is idiosyncrasies of people. And under the pressure of a startup, like, idiosyncrasies of people get magnified out to, like, a thousand-fold. Like, people become, like, the most extreme version of themselves under the kind of pressure they get under in a startup, and then that’s either to the good or to the bad, or both. But people have their own issues, and then the interpersonal conflicts between people. So, the day job is so much dealing with people, that you’d have to have the AI bot that can, like, sit down and do founder therapy. Maybe.

Balaji: Yeah. I mean, like…

Marc: My guess would be we’re still a ways off.

Balaji: Yeah. Like, just add to Marc’s point on that. I mean, the fundamental issue from, like, a machine learning standpoint is, you have very few events that are mostly returns, which are, like, these Facebook-like outcomes, right? And so it’s, like, almost like a rare event detector, like the Large Hadron Collider, right? You’ve got all these particles coming through, and you have to be able to predict, “Okay, which one of them is actually gonna make a lot of money?” That’s number one. Number two is, especially at the very earliest stages, you don’t have features in the traditional sense. Like, you don’t have a lot of really good data to work with, in terms of prediction. So, the later it gets, probably like series C or thereabout, you have enough, you know, systematic data to work with, but early on, it’s actually pretty challenging.

Marc: Yeah.

Man 3: Hi. Thank you. How are you guys thinking through your fund structure and the types of investments that you have to make as you raise more money? And can VC be, like, a winner-take-all market?

Marc: There are a bunch of challenges to it. The central challenge is, any top-end venture capital firm that has a reputation that it wants to maintain, which is I think very important, can only invest in one company in a category. You can’t, practically speaking, invest in competitors. The company you’ve already invested in will feel it’s [a] betrayal if you invest in the new one, and then the new one will think, if you’re willing to invest in them, you must be very, like, dishonorable that you’re willing to betray your previous one. So, just — it doesn’t work. And so, like, the minimal number of venture capital firms has to be the number of firms required to fund the number of competitors, right, in each new market. And then we can debate — is that 3, or 5, or 20, or 40, or 100? And you know, certainly, we have too many venture capital firms. Like, we’ve got like 500 venture capital firms in the U.S., and certainly, there aren’t 500 competitors in every market, at least. There need to be at least a half dozen, dozen, you know, 15, you know, good firms to fund the competitors. We would love to make venture capital a winner-take-all.

Man 4: I have a question with regards to blockchain and, like, the financial services industry. So it seems like there’s a lot of low-hanging fruit and a lot of far-fetched ideas that one could foresee using blockchain. So, I’m wondering, what advice would you give for someone who’s trying to see what is the best, I guess, niche area to target when you’re given such a wide array of potential use cases for the blockchain?

Marc: Yeah. So, we actually shy away from giving advice like that. So, there’s two reasons for it. So, one is there is a concept called product-market fit right, which has become very fairly publicized now, you know, right product in the right market. There’s another concept we call founder-market fit, which is — is the founder of a company — is that the person who’s born to do that idea? And so, that question we tend to defer to the founders, because we figured the really great founders are gonna figure that — like, part of what makes a founder great is they’re gonna figure that out. The other thing we found is that it’s very hard — we have ideas for companies we’d like to fund, but we try not to talk about them too much, because we don’t want somebody — we don’t want founders to pick up somebody else’s idea. And it goes back to what Balaji said, which is, it is so hard to make a startup work. You have to be so irrationally committed to it. I mean, this is another thing. Like, startups are over-glorified in the sense of, like, people think they’re fun. Like, they’re not fun. Like, they’re not even remotely fun. Like, they’re punishing as hell.

Balaji: I think it’s Bill Lee <inaudible>, it’s like chewing broken glass and staring into the abyss. That’s right.

Marc: He said starting a company is like chewing broken glass. It’s, like, after a while, you start to like the taste of your own blood. Very vivid quote. But, like, it’s so hard, and it’s so hard because people are saying no to you all the time. It’s just no, no, no, no, constantly being told no. And you know, “Your idea is stupid, and, like, I would never do that, or why would anyone do that. This other company is gonna kick your butt.” And like, then your lead engineer quits. It’s just, like, endless. It’s got to be an idea that they feel so deeply about. It goes to, like, Balaji’s term, ideological mission. It’s got to be something where people feel so deeply that they have to do it, that they’re willing to tolerate that level of pain. And in our experience, most people aren’t willing to tolerate that level of pain for somebody else’s idea. And so, I respectfully decline to answer the question.

Man 4: Okay, I see. No, it just seems like, for blockchain, there’s so many use cases, and for many of them, the timing could be completely off. Whereas, for example, for remittance payments, one could easily see how that’s a very easily applicable use case of blockchain, so.

Marc: Yeah.

Balaji: I’ll comment on this briefly. Basically, I think that remittances are to Bitcoin what VoIP was to the internet, in the sense of — it’ll work at some point. In the first 5 years or 10 years of the thing, it’s not high enough quality with the obvious alternative, namely VoIP versus landlines, or remittances versus legacy remittance systems to win. I think that, you know, Bitcoin, like Bitcoin as opposed to blockchain, but Bitcoin is good for transactions that are very large, very small, very fast, very international, or very automated. And you have to try to envision transactions that are, like, two, three, four, or even more of these kinds of things to think of things that cannot be done with the current system. If you think of things that cannot be done with the current system that are still useful, well, then, that’s 10x, right? So, that’s one way to think about it. 

The other great thing about it is, like, Evan Williams’ thing, which is sort of vague, but it’s actually very useful. So, on the one hand, oh, a new technology, 10x, something that people haven’t done before. On the other hand, Evan Williams’ thing is, take a behavior that humans want to do and allow them to do it faster, better, cheaper, over and over. Take something that was once a rich man’s thing and make it accessible to the middle class, or take it from the middle class and make it accessible to everyone, right? And so, if you kind of combine those two things, the technology allows you to go, and in a way that was not possible. So I’d — you know, hunt in that general area. That might be something.

Man 4: Thank you.

Man 5: I co-founded two companies that faded into obscurity too quickly. You identified problems, and issues, and opportunities [that] it might take a startup, you know, weeks, if not months, if not years to identify. I’m kind of curious why Andreessen Horowitz and others don’t explicitly identify opportunities and problems, or even issue challenges or competitions. Then, so — I wanna delve in a little bit deeper — one of the things you’ve been talking about, Balaji, more specifically, is, like, the cloud versus the land and, you know, “software eating the world,” like, the divergence of the cloud. And I’m kind of wondering, in that world, where ownership seems to be more centralized, there could be some risk associated with that. I’m wondering if you could speculate about ownership in the future. I’d be interested, especially, talking from a blockchain perspective on asset management.

Balaji: So kind of, there’s two separate questions there, and I think the first one is, why doesn’t VC pursue, like, an XPRIZE style model? That’s one. And then number two is, what happens with, like, the future of ownership, right? Kind of interrelated. So, the first one, I actually think would be a very interesting model for a fund. The reason I think that’s interesting is, one of the points Marc made is, and it’s one the most counterintuitive points about VC — no matter how innovative it is, an idea that comes across your doorstep today, there’ll be two more like it. My best example of that is Hyperloop, right? Like, so Hyperloop company comes across our doorstep, and like, a few weeks later, we have, like, two more that come in there. And so, what it means is that VC is all about filtering winner-take-all. So, the more that you can kind of push the tournament to inception, the more you can push the tournament earlier and earlier before you invest, the better. So, prize model, I think, could work. The problem is, of course, grading the prizes, judging the prizes, all of that type of stuff. That’s one.

Number two, in terms of the future of ownership, I do think that, basically, the interface to every physical object will be ultimately digitized, in the sense that you won’t own a car, you won’t have — we already don’t have a book, you have Kindle, right, and you don’t have a house, you have an Airbnb, and so on, and so forth. And all of it becomes extremely mainstream. And what that means is that, actually, your mobility is vastly increased. And right now, we think of mobile as, “Oh, I can just go to Starbucks, and I can work from there, and it’s as much as I could work at home.” But I think, in the next 5, 10 years, it’s gonna be as easy to just jump up and move to another country as it is to just go down the street. What that means is the more internationally flexible you are — so, one of the big aspects of that, by the way, is the bank accounts. That impacts the blockchain aspect. One of the big things that’s a pain moving between countries — your Gmail works, your Facebook works, all your internet services work, those are IP-based, right — but all the nation-state-based things, like your bank account, are not quite as portable or as easy. And so, those kinds of things, I think it’s useful to identify all the prerequisites. So, as a thesis for, kind of, startups to look at, chop the things that anchor people to land, and I think you’ll have some interesting things there.

Man 6: Hi. This question is for Balaji. I’m a freshman studying physics at the University of Illinois, and I was just wondering, what convinced you to continue on to do a Ph.D, and what were the skills that helped you on, like, in regards to entrepreneurship and whatever you’re doing today?

Balaji: So, I would not do a Ph.D today. That’s my quick answer.

Marc: So, why did you do a Ph.D.

Balaji: Why did I do a Ph.D? Because I wanted freedom, in the sense of I wanted to do math and, you know, computer science, and so on, on my own time, right? But what I would have done instead is — I think the single most important metric for you guys to measure is your personal runway. In Silicon Valley, people think a lot about, you know, “Okay, how do I get an exit and get the money on top?” But they think much less about, “How do I minimize my personal burn?” So today, in the world, it is possible to just find a jurisdiction that is amenable to your preferences — that is warm, that is safe, that has good internet, and it’s really, really cheap. And so, you know, what I would do instead of getting a Ph.D, if I was just doing it today. First, I’d worked for a year at Google, or Facebook, or GitHub. I would have a job that permitted remote work. I would sacrifice the advancement to be able to work remote for the next three years or so on, and I would just save enormous amounts of money and live very, very cheaply. Every year that you work, you’ve got three years of runway. And so, that’s actually freedom. Once you have the ability to have, like, 3, 4, 10 years of runway, and you have the discipline of the grad student but the earnings of an engineer, right — so that’s what I would have done instead. So I wouldn’t do the Ph.D. I think you can learn and self-learn faster on the internet than you can, you know, in grad school. I think a bachelor’s degree is fine, like, you know. Like, I’m not saying drop out, or what have you, right now, at least. But I think you can do better than a Ph.D today.

Marc: So I’ve got a question for Balaji.

Balaji: Yeah.

Marc: So Balaji is, for those of you who know him by reputation or know him — and tonight he’s done this — very big advocate for entrepreneurship outside the Valley, very big advocate for developing world entrepreneurship, very big advocate…

Balaji: Why am I still here?

Marc: …in this case, for, actually, literally, moving computers someplace else. I can’t help but point out that Balaji lives — where do you live?

Balaji: Yeah, no. Unfortunately, I’m in San Francisco. But, but, but.

Marc: Interesting. Interesting. Interesting. Literally, if you drew a circle around San Francisco, he’s right in the middle.

Balaji: Let’s say that sometimes you have a goal that you have, it takes a while to get to because there are a bunch of prerequisites that have to be met.

Marc: Right. He keeps saying he’s thinking about it. We’ll see. We’ll see.

Balaji: I keep saying, no, no, I’m working on it. All right.

Marc: He now is married, and he has a lovely little baby.

Balaji: I do, I do. Those are all anchors.

Marc: And the two of them are gonna have, I think they get votes.

Balaji: They get votes.

Marc: It’s my understanding of how this works. They get to contribute to the experience.

Woman: Thanks, Marc. I’m from China. I work for Google China and [am] now a current student in the Stanford GSB. I’m really inspired by the entrepreneurship here, but I know there’s a lot of challenges for the immigrant entrepreneurs to start a company here. So, I’ve been wanting your advice for the immigrant entrepreneurs, especially for the first time.

Marc: Right. I’m gonna turn that question over.

Balaji: Sure. Okay. Yeah, sure.

Marc: To the immigrant entrepreneur on the stage.

Balaji: Yeah, yeah, sure. So, I thought about this a lot, and I’ve discussed this with Marc and Ben a lot. What comes after the dorm room entrepreneur is the developing world entrepreneur and the immigrant entrepreneur, but especially in the developing world. And I think, you know, one thing, you know, depending on what country one is coming from, and so on, obviously, there’s a wide range, but for someone coming out of India, for example, frequently making $100,000 is like making $1 million, in the sense of, like, the impact on quality of life, and so on, right? And there’s actually much lower-risk ways to make $100,000 than to do a startup, which is just extremely stressful, and you’re going for infinity, and so on, right? 

And so I think that we’re gonna see new kinds of things, particularly as you get another billion, two billion people with cellphones, right? Like, then we’re gonna see new kinds of business models that are based on knowledge that folks outside of the U.S. and in the developing world have about their local economies, and also have maybe less than we have upside, more predictable returns, and they’re not quite as much of a, you know, roll-the-dice kind of thing. In some ways, if you start at zero, it’s easier to get to infinity, because you just have nothing to lose.

Marc: Good, good, good. Thanks, everybody, for coming. Thank you. Thank you.

Balaji: That’s good.

• 

  Marc Andreessen is a cofounder and general partner at a16z. Marc co-created the highly influential Mosaic internet browser and cofounded Netscape.

  Follow Twitter
• 

  Balaji Srinivasan

One of the most important lessons of the internet age is what happens when we give people — including companies, developers, engineers, hobbyists, and yes, even a few bad (or dumb) actors — a new platform, along with the freedom to innovate on top of it. For example, who could have predicted how profoundly the internet would change our economy, given how it started off as a research project — one where commercial applications were actually frowned upon in the early days?

Now, the U.S. is on the cusp of opening up another such platform for commercial and social innovation: airspace (think drones, the non-military kind). There’s so many use cases for drones that we already know about, but what about new business use cases? And then, on the policy front, how do we calculate the risk of innovation on a platform made up of atoms (drones) vs. bits (the internet)? What are the pros and cons of registration? Because even though drones are like flying smartphones controlled by software, they’re also hard objects that could fall out of the sky … or go places where no one could go before, for better or worse.

The guests on this episode of the a16z Podcast — continuing our D.C. and tech/innovation/policy theme — share their thoughts on safety, privacy, paper airplanes, and what they think are some of the most exciting things now possible in airspace. Joining the conversation are Washington, D.C.-based Mercatus Center tech policy lead Eli Dourado, along with graduate research fellow Samuel Hammond; Airware founder and CEO Jonathan Downey; and SkySafe CEO and co-founder Grant Jordan.

Show Notes

• The current state of drone technology and potential use cases [0:57]
• Creative applications for using drones [10:04] and barriers to innovation [14:00]
• Issues with safety [19:23] and privacy [28:55]
• Discussion of new potential developments on the horizon [33:07]

Transcript

Sonal: Hi, everyone. Welcome to the “a16z Podcast.” I’m Sonal. And today’s topic, continuing our D.C. theme, is drones, and more broadly, policy and airspace as a platform for innovation. Joining us for that conversation, just to really quickly do the intros, we have Eli Dourado. He directs the technology policy program at George Mason University’s Mercatus Center. And he’s here with Samuel Hammond, who is a graduate research fellow in that program. And they’ve done a lot of research in policy reports around drones. And then we have Jonathan Downey, who is the founder and CEO of Airware, which provides operating systems for commercial drones, so enterprises can take advantage of aerial data for business applications. And we also have Grant Jordan, who is the CEO and co-founder of SkySafe, which is the company that provides security for airspace — for example, by taking control of and safely landing rogue drones. 

And, by the way, those are both a16z companies, full disclosure. We have another drone’s investment, Skydio — which focuses on onboard intelligence by giving drones the same visual awareness and agility of human pilots — but they couldn’t join us today. So those are the intros. And now let’s just get started. 

Innovation in drone technology

I think the place to start off is, you know — Eli, you wrote an op-ed for me a few years ago, where we talked about airspace as the next platform for innovation. And I thought that was a really eye-opening concept for me, and I think we need to just break down what each of those terms mean. Like, what is airspace, why a platform, and why is it the next platform for innovation?

Eli: Well, yeah, when I wrote that op-ed — I think it must’ve been about three years ago when I was first starting to get interested in drones. And the thing that I noticed immediately was that drones are completely legal to use for hobbyist purposes, and completely illegal to use for commercial purposes. And it reminded me of the internet back in the 1980s, right? The internet in the 1980s, it was this research program that the government had, but there were guidelines. You know, MIT had an AI lab, and the guidelines for students were, like, “You may not use this for any commercial purpose. We can get in trouble with the government.” It’s just illegal. 

And I thought about the people running this program. They didn’t mean to, like, hold back the internet. They were, I’m sure, very well-intentioned. They wanted this to succeed, and they weren’t thinking about how much holding something back from commercial operation was likely to affect the success, right? If they knew where we were today, they would’ve allowed commercial use from the inception, right? And so to me, I was just thinking about — what would people do with airspace if they could use it for commercial purposes if it wasn’t — you know, we’ve had remote-controlled airplanes for decades. Hobbyists use them, so in that sense, consumer drones are nothing new. But you can’t do anything commercially. And so, what would people do if you could use it commercially?

Sonal: And Jonathan, I think this is something you can weigh in on, because you are the founder of a company that is doing software for commercial drones.

Jonathan: Yeah, happy to. And, you know, I totally agree. Three years ago, around the time you wrote that article, we looked at the global landscape for commercial drones. And most everyone at that time was thinking about consumer and hobbyist drones, and they were thinking about large military aircraft, but this idea of using drones for commercial applications was relatively new. And when we looked at the landscape, most of the companies who were doing anything meaningful in the space were international. They were in Australia, they were in France, they were in the UK. There were only a few countries where they either said, “Hey, we’re not going to regulate this at all. We’re going to allow it.” Or they said, “We’re going to have a process by which you can be a commercial operator of drones.” 

And the UK and France were really leading the way there, with hundreds of commercial operations years ago. And many of the companies in the space said, “Well, we’re just going to ignore the U.S., or if we are in the U.S. we’re going to start all of our sales, and operations, and research and development outside of the country.” And only now with this Section 333 exemption process — and, hopefully, soon with Part 107 — do we see a lot of those companies that either started internationally expanding their operations to the U.S., or a lot more of these companies that are in the U.S. starting to get some commercial traction.

Sonal: What is the Section… What was the regulation that you just cited?

Eli: So, Part 107 is the new proposed small UAS rule and it will, sort of…

Sonal: Buy UAS, you mean unmanned systems?

Eli: Unmanned Aerial Systems.

Sonal: Right, Unmanned Aerial Systems, yeah.

Eli: So, this is the FAA term. But we call them…

Sonal: Which, by the way…

Eli: Cool people just call them drones.

Sonal: Right, that includes drones, which just for a quick definitional thing on drones — because I think people actually are still confused by this sometimes. I noticed this in the early days of this community, an online community about drones — which is that drones are different than RC copters in that they can fly waypoints. And so they can follow, like, a preprogrammed path, and that you can — they’re not, like, remote-controlled, basically, in that context.

Eli: So, the terms do get, like, blended, and so on. But yeah, I would agree that what we’re interested in here is some partial or full autonomy at some point. That you’re going to eventually be able to just tell it what you want it to do and it does it.

Sonal: Okay, and so back to section whatever.

Eli: So, the Part 107 rule is — back in 2012, Congress, actually with some foresight…

Jonathan: It was actually with a lot of lobbying from large commercial industries. 

Eli: A lot of lobbying, but good for Congress, they passed a provision of FAA reauthorization bill at the time to require that FAA come out with commercial drone rules by September 30th, 2015, which was, you know, several months ago.

Jonathan: Came and went.

Eli: It came and went. They’re still not out. But they’re coming. It’ll probably be sometime next month. There will be some permanent commercial drone rules. And in the meantime, as Jonathan said, we have these 333 exemptions that basically allows people to, sort of, negotiate or apply with the FAA to say, “I wanna operate commercially for this purpose.” And the FAA might allow it, or might not.

Sonal: And what are some of the purposes and use cases that you guys are seeing people put to — I mean, what are the — why would they want those exemptions, basically?

Jonathan: Yes, so the Section 333 exemption process was created last fall. And now just, you know, 9 months later or so, there’s over 4,000 granted exemptions in the U.S. for commercial use of drones across a variety of different industries. But we’re seeing a lot of exemptions in our own customers using drones in the insurance industry, in agriculture, in utilities, in all types of industrial inspections — oil and gas, land management, forestry, wildlife conservation. You know, that’s one of the things that I think is so interesting about this industry, is just the wide variety of applications and use cases. They’re really endless. We’re hearing about new ones all of the time. And I really think, you know, it makes that analogy to the early internet, you know, very, very real, in that it was designed and developed with several really important use cases in mind. But ultimately, when it was, kind of, released into the wild and, you know, not just people “with authority to develop it,” but really all kinds of people everywhere were adding capabilities to software and to the internet — is when we really saw this, you know, bloom in all the different uses for it.

Grant: It’s totally the case that we don’t even know what the real killer apps for drones are yet. You know, I think there’s a couple of spaces that are, kind of, obvious that we’ve been thinking about so far of, you know — delivery, and inspection, and all these sorts of things. But there’s so much potential there, it’s, kind of crazy. You know, I think part of — a lot of our focus is on thinking a little bit further forward about the airspace management aspect. You know, I think using commercial drones early on in the process is pretty easy when they’re still very expensive. When the sorts of companies that can use them are still very limited. 

You know, much like security in that early internet, right? When the internet is just a connection of a bunch of research centers and government, then it’s not — security is not really that big of a problem. Everybody on there is trusted. Everybody on there is doing the right thing. But as those barriers to entry start coming down — as, you know, anybody with 500 bucks can be flying a drone — suddenly it becomes, kind of, a different story. And suddenly, you know, people who don’t really know the rules of the road, who aren’t really sure what they’re doing, or have actual malicious intent — they, kind of, start coming to the party as well.

Sonal: So you guys are, kind of, getting ahead of that. I want to take a step back for a moment and think about again — we’re all really reinforcing this concept of airspace as a platform for innovation. I think it’s actually, kind of, shocking to think about what it means to be able to do things in the sky. And I just want to take, like, a moment to pause on that.

Eli: Yeah, and I think people don’t realize that, like, cell phone tower inspections is, like, one of the most dangerous jobs in America. OSHA called it…

Grant: OSHA declared it the number one most dangerous job.

Eli: …the most dangerous job. And that’s going to be…

Sonal: Really?

Eli: Yes.

Sonal: Why?

Jonathan: Yeah.

Sonal: Just because it’s so high up?

Jonathan: There were 14 deaths in 2013 alone from tower climbing.

Eli: So, and there’s not that many…

Sonal: Wow.

Eli: …people who climb towers, right? So this is a fairly high percentage of the people who do this for a living. These are high towers and people fall, or something goes wrong.

Jonathan: And that’s one that’s gotten a lot of publicity recently. There’s a lot of other jobs that are really dangerous as well. For two story steep rooftop inspections of residential properties, whether it’s during the underwriting phase, whether it’s during the claims phase — can also be a pretty dangerous job as well. And many companies — it’s, kind of, an opt-in job. They don’t just assign it to you, you have to, kind of, ask to be that person climbing up on the roof.

Sonal: Oh, really? Because it’s that dangerous?

Grant: It’s that dangerous, yeah.

Sonal: Wow, so there’s clearly a lot of dangerous cases. What are some of the — I mean, oil and gas, that’s what I hear about all the time as an industry that needs drones. Like, why is that?

Jonathan: Flare stack inspections for both onshore and offshore infrastructure in oil and gas. Also, oil derricks need to be inspected, and oil platforms need to be inspected for corrosion and damage on a regular basis. And a lot of these inspections, similarly, are very dangerous to do. Or, with the case of flair stacks, often the infrastructure needs to be shut down so that people can actually climb up on it. But with a drone in an aerial perspective, it gives you a completely different way to assess the status of the infrastructure without having to shut down critical equipment.

Practical uses for drones

Sonal: So we’ve, kind of, outlined some of the more dangerous jobs and use cases that drones can help address. Now let’s think about some of the things that there are opportunities you wouldn’t have had if it weren’t for drones. Because when we talk about all those application use cases — insurance, agriculture, etc. — in a lot of ways, we’re talking about disintermediating existing alternate approaches that are expensive, or prohibitive, or difficult, or unsafe. Like, having to do ladders, or oil and gas inspection. Things that are just impossible. I think it’s especially interesting on the creative side, like, what you can do with photography, aerial Hollywood filmmaking, and some of the really creative aspects of this. Because to me, the internet wasn’t just a commercial platform, it was a creativity platform. And I’m curious to hear what you guys are seeing on that front as well.

Samuel: I’d just say, first of all, that that qualitative part gets missed in, like, FAA cost-benefit analyses, you know? So, how do you put a price on a vista that you haven’t seen before, right?

Sonal: Exactly. Especially when you don’t know what it’s going to look like, because that’s the whole point of — again, not to be platitudish, but that’s the whole point of innovation. Like, it’s supposed to surprise you in terms of what’s possible. We can list all the use cases we want in this room, but we truly have no concept until we see companies and people start really inventing things around it.

Grant: Well, yeah. I was actually going to say on the creative side, one of the things that I think is some of the most interesting stuff is on filming — just being able to do shots that used to require helicopters, that used to require, you know, tremendous amounts of coordination, and time, and money. And now it’s just, like, you know, a drone just allows that so easily. So, kind of, the entry-level for what used to be a helicopter shot is, like, nothing now.

Jonathan: Well, and whether it’s in Hollywood and taking shots that were previously from manned helicopters, or whether it’s the utilities industry, you know, getting high-resolution photos of power lines from what used to be manned helicopters — that’s the starting point. But then it gets really interesting when you start understanding what wasn’t even possible with manned helicopters, and now is becoming possible with small aircraft — including, you know, new shots in Hollywood that previously — you know, you can’t get a helicopter within 5 and 10 feet of a person and, kind of, circle that person. But you can do that with a drone, especially as these things are becoming smaller, safer, lighter-weight.

Grant: Right, or even shots that transition from inside to outside.

Samuel: Passing through windows and so on. I’m no Hollywood expert, but I imagine in the past that they’d do a shot that pulls into the window, and then there’d be some hidden cut, and then reset the shot from inside. Now you could presumably just open the window.

Jonathan: The corollary for a lot of these industrial inspections is the ability to do things like fly underneath a bridge, fly underneath an oil derrick. The top side of some of this infrastructure was always accessible with manned helicopters, albeit at a very expensive price tag. But now you can go inside of buildings, inside of — I mean, we’re seeing companies do inspections of the insides of everything from, you know, large oil containers to even large gas turbines.

Eli: What’s interesting to me about the Hollywood application is that they actually were some of the earliest adopters, and they adopted it even before it was legal for them to do so. So Hollywood is, like, the…

Sonal: Follows its own rules, goddamnit.

Eli: Hollywood was, like, the Uber of drones and, sort of, like, they were just going to do it and ask for forgiveness and not permission. So, that was one thing I think, yeah, Hollywood helped with the drone policy.

Grant: Yeah, well, and also Hollywood actually helped a lot on the 333 exemptions. Right, because they, kind of, had some of the biggest immediate incentive to get commercial use approved. And also, what I thought was, kind of, cool is, you know, they paved the way for the use of drones in filming, because they already had all of the safety procedures, and flight manuals, and things that they had been using for manned helicopter shoots for years. So, they literally just took those and shifted them over and, you know, called it, “This is the manual for shooting with drones.” And that’s what, kind of, lets you get the 333 exemptions for shooting so easily now.

Samuel: So, we discussed some of the innovation arbitrage around drones. When it comes to Canada, some of that was also film, because there’s film industries in Toronto and Vancouver. And they could make shots and not have to worry about asking for forgiveness.

Sonal: So it’s interesting you reference innovation arbitrage, because the example you shared is basically — and this goes to what Jonathan was saying earlier about some of the innovation happening around the world — is that when certain places have more regulatory flexibility, it then draws that industry correspondingly. And I think the reason this causes U.S. lawmakers to freak out a little bit when they hear, you know, Amazon saying, “Hey, we might start developing drones in Brazil or another country,” it actually becomes correlated to a direct loss of the economic opportunities that are provided as a result of this. 

I mean, I’m thinking of the internet example. It’d literally be like saying, “You know what? We’re not gonna let commercial applications happen on the internet. So let’s develop them in China, and India, and Brazil, and South Africa.” Yeah, I’m just listing the bricks right there. But that is, kind of, the risk, to me, at stake here when we talk about this. Because I think, again, people are really underestimating how much is possible in the air. I don’t mean to be cheesy, but drones excite the fuck out of me because — and no, they really do — because it’s insane to me that, you know, we talk about men, women, humankind wanting to, like, fly. And now we’re talking about a whole new level of excitement, to be able to reach into the air and do things. I mean, don’t people get that, like, this is a really big deal for God’s sake?

Samuel: Adam Thierer from Mercatus Institute and I have a paper forthcoming called Global Innovation Arbitrage, and drones are a major case study that we look at.

Sonal: Why’d you guys pick drones?

Samuel: Well, we picked, you know, drones, we picked genetics. So, there’s these big things where, first of all, they’re major emerging tech, and they’re very much in the news. I picked drones personally, because, as a Canadian, seeing drones in the industry take off in my home country. And so, we actually focused on Canada and Switzerland. And the Swiss government has taken a very risk-based approach to drone regulation.

Sonal: So, is that a bad thing or a good thing?

Samuel: A good thing.

Sonal: What do you mean by risk-based?

Samuel: A good thing.

Eli: They’re actually evaluating the risks and moving forward accordingly, as opposed to just blanket bans or…

Samuel: And it’s flexible for that reason. So there are very few — for example, in Switzerland, there are very few bright-line restrictions on what you can or cannot do. They’ll have guidelines about, for example, going beyond line of sight. But those aren’t written in stone, such that if new technology makes that safer, that they can’t be revised, sort of, on the spot.

Sonal: Right, because the current law, if I’m not mistaken — at least locally, I know that you do have to keep drones within your visual line of sight. And that, sort of, seems to defeat the purpose, that the very purpose in certain use cases, like, if you’re a farmer mapping your fields is to be able to go beyond the line of sight.

Samuel: So when I reached out to the Swiss government on this — and this is another analogy of the early internet. I was really asking them for an estimate of how much commercial operation is going on in their country. And the reply I got back was, “We’re not the United States. We don’t keep a tab on every single commercial entity.” So…

Sonal: That’s fascinating.

Samuel: And, sort of, like, the early internet, you could look up every website in, like, a phone book, right? That’s, sort of, the mentality that still exists in the U.S. with drones and the Swiss and the Canadians. They’re comfortable not knowing exactly how many commercial operators there are.

Sonal: Right, and in fact, isn’t part of the problem — I mean, in the current state — that the regulation process can be prohibitive? I mean, the registration could be really expensive for small operators. I don’t know enough about it — like, what’s…

Eli: Well, there is a consumer registry now. And that registry, it’s not expensive — it’s five dollars — but it could potentially stop people from taking that step. There’s, I think, a lot of lawbreaking going on right now, because I think there’s something, like, a million-plus consumer drones, and only 400,000 have registered.

Sonal: Are registered, right?

Eli: So, we’ve created a new law that has turned us all into a nation of lawbreakers. But then the other thing is this isn’t — the registry is not very well-tailored to the actual risks that we face. You have to register a drone if it’s more than 250 grams, which the FAA helpfully put that…

Sonal: Which, by the way, how much is that in pounds?

Grant: Well, the FAA…

Samuel: 0.55.

Eli: This is two sticks of butter, is what the FAA says.

Sonal: Oh, really?

Eli: That’s the way, yes.

Sonal: Oh my God.

Eli: So two sticks of butter or bigger you have to — so it’s, like, half a pound, well, 0.55 pounds.

Sonal: It’s insane.

Eli: If it’s that big, you have to register it, right? And one of the things that we’re looking at is how dangerous is it, you know, to have a more flexible standard that goes up to, like, 2 kilograms, which is what’s used in a lot of other countries.

Sonal: How much of that in pounds, 2 kilograms?

Eli: 4.4 pounds.

Sonal: And so this is not even — this is just the weight of the drones themselves.

Eli: Of the drones, yeah.

Safety concerns

Sonal: It’s not including any, kind of, payload, like, for commercial applications when you’re doing delivery. <crosstalk> It counts the payload, okay. So, let’s take a step back for a moment and just talk about the safety implications of drones. Because what’s different, obviously, with the internet, and drones, and any airspace objects is that they are flying — we call them flying smartphones or flying computers, we think of them that way. But they are flying objects that can fall out of the sky, like, hit your head, they can get in your tree, they could kill your cat. I mean, I don’t mean to be frivolous about it, but these are realities. So, let’s talk about the safety implications of drones. And what are some of the concerns that people have, and that you guys who are really involved in this space have heard?

Eli: Well, I think there’s two things that people are worried about, and one is, as you say, falling out of the sky, hitting people on the head. And to me, that is something that we can properly deal with through the tort system. Just in the same way — if you hit somebody with your car, like, they sue you. Maybe the insurance company…

Jonathan: It’s a great example. It’s an example, though, that requires liability insurance to drive on public roads, and that requires registration of your car. So, I might be the odd person out here, but in the same way that you more or less have to have a MAC address to get on the internet, there should be some mechanisms by which we identify the other people who are flying drones near us, or flying drones. I was having dinner about a year ago, actually, in Berkeley, and a drone flew right into the side of the restaurant, crashed, and then just about fell on top of the head of this girl who was standing there. So, you know, I think it is an…

Eli: There’s a lot of people out there doing dumb things with drones.

Jonathan: There are some people doing dumb things. And I think we can keep the registration requirements and things like this very, very basic, and easy…

Sonal: Lightweight, yeah.

Jonathan: …and lightweight, but structured in a way where, you know, the person has an incentive. If there’s some identifying marking on that drone that’s going to say whose drone it is, people are going to be incentivized and maybe think twice before they, you know, fly their drone into the side of the building or, you know, above New York City and into a skyscraper.

Sonal: So, you’re saying associate some, sort of, identity or location.

Jonathan: That’s the idea behind registration, is that if you’re flying this drone, especially in a public space, and something goes wrong, people will be able to identify whose drone it is.

Eli: But then is it really 250 grams that’s the right threshold for that? I mean, I’m not worried about a 250-gram drone.

Sonal: Two sticks of butter falling on your head, I hate to tell you, nothing’s gonna happen.

Eli: Two sticks of butter falling on my head, I’m not too worried about that, personally.

Jonathan: I think that that 250-gram allowance could be called the paper airplane allowance.

Eli: Right. Well, because actually under federal statute, paper airplanes are aircraft.

Sonal: Wait, are you serious?

Eli: Yes, yes.

Jonathan: Yes.

Eli: So paper…

Sonal: What?

Eli: …airplanes are… The FAA…

Sonal: So every time a kid, like, folds up a paper airplane in the classroom in kindergarten, they’re, like, breaking the law? Or they’re not registered?

Eli: The FAA is forbearing on enforcing standards…

Samuel: Sort of discretion.

Eli: …the registration standards on paper airplanes.

Sonal: Oh my God.

Eli: So, the FAA would say that they have the right to regulate that. I mean, they would be embarrassed to say it, but they would say that they do.

Sonal: So, Grant, when you guys, you know, started thinking about this, like, thinking far ahead about, like, okay, the same way the internet needed, like, security — and airspace will need security, in this way where you can essentially enforce, so to speak, the anti-drone —what were the scenarios that were coming in you guys’ minds that you came up with this?

Grant: Yeah, well, I mean, it’s kind of interesting, too, because when we first started working on this and thinking about this, you know, it was still pretty early in the space, and we weren’t really seeing drone incidents occurring. You know, whereas now it’s, like, literally one a week, if not more. But I think to me, the big difference is that there’s, kind of, a gap in airspace enforcement, right? Like, if you’re talking about commercial aviation, civil aviation — at some point your enforcement of airspace restrictions — there’s, kind of, two things that come into play. One, is just the barrier to entry to be involved in aviation at all, right? You know, the amount of training required, the amount of money required upfront with planes, and fuel, and all of that. But then, you know, in addition to that, it’s a question of, at the end of the day, there is an enforcement mechanism up there. You have your plane registered with who owns it. The FAA can come and just cite you, can take away your licenses, things like that. You know, they can track you down. They have transponder requirements. And also, you know, at some point, if you fly into restricted airspace, the National Guard will literally, you know, fly an F-16 up next to you and tell you to land.

Sonal: They shoot you down like, “Top Gun.” Sorry, I don’t mean to get all dramatic. It’s just…

Grant: Yeah, but, you know, the F-16 and the National Guard doesn’t really help when you have, you know, just a quadcopter flying somewhere, that’s not really an appropriate level of response. But, you know, the spread of drones on the consumer side really, kind of, brings in this level of — it, kind of, changes the rules in thinking about airspace security, facility security, things like that. You know, when we talk about how it redefines various ways that we think about things, you know, if you’re thinking about perimeter security of something like a power plant, like, a nuclear power plant or something — or a prison, for example — there’s a lot of assumptions we make in building a security perimeter about fences, right? But now that you have drones, you know, an 8-foot fence versus a 20-foot fence versus a 2-foot fence is pretty much equivalent. It doesn’t really matter. You know, you can fly your drone over and deliver your contraband, kind of, regardless of the height of the fence. So it, kind of, just breaks down a lot of our, like, traditional security models.

Sonal: And so you’re thinking about it more in the sense of how to — because I know there’s geofencing, where you can actually, like, fence in a region that a drone is, sort of, contained to fly, but you’re talking about when you can’t control the perimeter, so to speak, in military parlance.

Grant: Yeah. I mean, you know, the problem at the end of the day with something, like, geofences, right —which, obviously, totally necessary, totally a step in the right direction. But you’re trusting the device…

Sonal: The operator, right?

Grant: …you know, to essentially police itself. You know, you’re saying, “Drone, don’t fly here.” And it’s going to abide by your rules. And that’s really good for eliminating, kind of, the initial low-hanging fruit of the people that are going to follow the rules. But, you know, a really good case in point is literally anyone flying a drone right now within a 30-mile radius of Washington D.C., right? Right now, in order to do that, you need to essentially override those controls, because that 30-mile radius is a no-drone zone — which also, incidentally, is, kind of, confusing if you are an RC hobbyist who’s been living near D.C. for decades flying RC planes, but now because it’s a quad rotor instead of a traditional RC plane, now it’s not okay to fly there. It’s, kind of, a confusing set of rules.

Sonal: Right, so you’re thinking about the enforcement aspect. Well, I want to think about the other safety issue, or is that the — you were mentioning there’s another one.

Eli: So, the other one I was thinking of was collisions with planes in the air. It’s the one that — the manned planes. So, this is something Sam and I have done some research on, and we use as a, sort of, parallel phenomenon — planes hitting birds. So, there’s actually many orders of magnitudes more birds in the airspace than there are drones. And so, what sort of conclusions can we have? And the FAA actually has 25 years of data on, sort of, voluntarily-reported bird strikes. And so we looked through that.

Sonal: What’d you guys find?

Eli: Birds are pretty safe.

Sonal: Okay. Thank God, I love birds.

Eli: So, we do hit birds all the time. Sometimes they cause damage, sometimes they do cause injury or fatalities. But in the context of — just the massive, massive number of birds and all of the flights — manned flights that we have, it’s actually a very low rate at which they cause any threat to humans.

Sonal: And how does this play out with drones?

Eli: So what we think is that the evidence seems to show that for small, you know — one of the things we do is we look at individual drones versus swarms of drones, because birds, they fly in flocks. So we look at the subset of bird strikes where it’s just a single bird. And then we look at the species of the bird and assign it a weight based on the average mass of the species. And so, what we found is — we looked at the 2-kilogram thresholds, and so that’s what’s used in a number of countries.

Sonal: More than a stick of butter. Two sticks of butter.

Eli: More than a stick of butter. And that’s what’s used in a number of countries for the threshold for what can be unregulated. And for a 2-kilogram drone, I think we found that there might be a human injury once every 187 million years — continuous flight hours…

Jonathan: Continuous flight hours…

Sonal: Wow.

Jonathan: …of the drone.

Eli: …of the drone. And then if you look at commercial jets, it’s even smaller. I mean, I think I got — I think the last number that we got was 41 billion years of continuous operation. Which is, you know, 3 times the age of the universe. So pretty safe. I’m not very worried about, like, a 2-kilogram drone taking out a 737, or anything like that.

Sonal: Right, well, I think one of the funniest videos I’ve seen on the internet, and I’m sure you guys have all seen it, it’s this one that — it goes every viral every now and then, and there’s always a different version of it. Like, of an eagle battling a drone.

Eli: Yeah.

Jonathan: Eagle versus drone, kangaroo versus drone.

Eli: Yeah, spoiler alert.

Sonal: Oh, was it…what were some of the other animals?

Jonathan: Kangaroo wins.

Eli: Yeah, yeah, it’s a kangaroo.

Sonal: That’s great.

Eli: Yeah, I mean, we’ve learned about birds that they’re more territorial than a lot of us knew. I guess scientists knew this, but the rest of us didn’t realize how territorial birds are.

Privacy issues

Sonal: Right. Well, then the third category of safety that I think has come top of mind for a lot of people is privacy. So, one example this weekend I thought was really interesting is — the New York Times decided to fly and use drones to see the mass graves that were being unearthed, and they weren’t allowed to look at them. And I thought it was great. I saw that — oh, I think it was someone on Twitter, Jenna Wortham or someone said, “Hey, we sent in drones.” And I’m, like, “Yeah, that’s great.” And then I was thinking of the counterexample about it — an individual level, where you might have a star who wants her privacy, and she doesn’t need to be spied on by paparazzi. And they’re going to use drones, just like anything else. And so, what are some of your thoughts on that use case, and that concern?

Jonathan: I think this is a case where, you know, technology can really be a significant enabler. You know, 10, 15 years ago, you know, if you had asked everyone whether they were willing to carry around what’s essentially a GPS tracking device in their pocket, people would’ve thrown their arms in the air and said, “Absolutely not.”

Sonal: Totally.

Grant: But with a couple, you know, technology additions to your cell phone, and the allowing you to turn the GPS on and off, and delegate which programs have access to it, and when they have access to it, and opt in to allow 911 to have access to your GPS position.

Sonal: You feel safer.

Jonathan: People are quite comfortable. They actually are happy to have that on them. And now there’s a variety of different applications around, you know, enabling people to run by themselves and alert someone if their GPS position ever stops for five minutes or more.

Sonal: Parents use it to track their kids for safety.

Jonathan: Parents are using it to track their kids. So, all of that, I intend to just be, kind of, an example of — I think the same thing is playing out here with drone use. Which is to say, yes, this is a technology that could be used to invade people’s privacy. But with some basically, you know, technology controls on it, it can also add a tremendous amount of value to society, without invading people’s privacy. And so there are technology mechanisms to — if you have permission to fly over Property A, and Property A abuts Property B, it’s relatively easy to make sure that photos that are taken over Property B are immediately deleted, or are never taken at all. Or, photography is only taken over the property that you have permission to fly over.

Sonal: You can do a lot through technology.

Jonathan: You can, and you can have, you know, geofences that allow for you to only fly within the bounds of properties that you have permissions to fly, or Section 333 exemptions to fly over, with the permission of the property owner.

Sonal: Right, and besides technology, there [are] also existing laws that cover so much of this, like, the Peeping Tom case. Like, why do we need new law, when there’s already…

Eli: No, that’s right. I’ve heard policymakers say, “Well, what if this, like, goes up right next to my bathroom window and looks in?” And the answer is, there’s already a law against this. It’s probably a state or local law, and it would just be enforced in exactly the same way. So, there doesn’t need to be, I think, a drone-specific…

Sonal: Preemptive.

Eli: …rule for that.

Jonathan: And these types of laws should be technology-agnostic. It shouldn’t matter whether it’s binoculars used, or whether it’s a drone used, or a ladder.

Eli: That’s right.

Sonal: I like that idea — that it should be technology-agnostic.

Samuel: But in terms of how the law should evolve, as long as we’re treating drones as any other kind of airplane — I wouldn’t recommend this, but if a drone is trespassing over your property and you choose to shoot it down, it’s, like, treated as shooting down an airplane.

Eli: Shooting an airplane, which is, like, 25 years in jail, or something like that. I don’t actually know the penalty. But this is — so, don’t shoot down a drone right now. Don’t be the test case.

Sonal: Right.

Grant: Well, I guess, yeah. As far as [a] test case, I mean, that’s part of the question here — is that regulatory-wise, we don’t really know where we stand on that piece, right? You know, so far we have one piece of case law of — in Kentucky, apparently, you can shoot down a drone with a shotgun and you’re fine.

Eli: Well, the FAA I think has issued — they’ve preempted that, and they say, “This is a federal offense, and it is shooting down an airplane. And it’s the same.”

Jonathan: Yeah, it’s pretty clear, both from all of the, you know, past laws that exist and then right now, there’s language just to reinforce it in FAA Reauthorization Act of 2016 — just further clarifying the Federal Government and the FAA’s ability to preempt all state and local laws as it pertains to the national airspace.

New developments on the horizon

Sonal: Great. I’m just surprised it wasn’t Texas and it was Kentucky. So, what excites you guys? I mean, we’re talking about some of the — just to switch gears from the safety topic again, and go back to, like, what’s new and exciting. So talk to me about what’s interesting to you guys. You guys are on the forefront of watching the trends in this space. I want to hear what’s new and interesting.

Jonathan: I guess from our perspective, right now in the United States, we’re still at this stage where the military has been using drones. They’re using them every single day. You have millions of consumers, literally, who are using drones in many cases every single day. And the major commercial companies have yet to really step into the fray and move from testing of the technology, which is where most all of them are at today, to actually using it as the way they do things. For whether it’s, you know, the insurance industry underwriting claims, catastrophe response, or utility inspections, replacing climbing up towers, and manual flights with helicopters with drones. So, I think that’s the thing that’s most exciting to me, and something that I also expect we’re gonna see in the next 18 months — the commercial companies actually move forward with, you know, commercial drones and aerial data as a way of doing business.

Grant: I’m super excited about airspace integration, and about us getting to the point where we can actually have, you know, large quantities of commercial drones in the airspace, you know, kind of, interacting with commercial traffic. Routing correctly, and things like that. You know, and it’s the kind of thing where it’s going to take a concerted effort by a lot of different groups coming together. You know, you can’t have Amazon using drones, plotting their own paths, and Google using their drones and plotting their own independent paths, with no interchange of information between them. 

Like, the NASA UTM program is working on a lot of that stuff. And I’m super excited about that. You know, super excited about transponders on aircraft, you know, things being registered properly — you know, actually having accountability. And once we get to that point, where we’ve got integration, where we’ve got accountability — then that just, like, opens up the door to all these different uses. And being able to have a point where a company, you know, can sit down and say, “How can we use drones? Okay, this is a thing that would help us.” And then there’s just [an] unknown path to actually use them correctly. You know, right now we just have so much gray area left in that system. But once we get that cleared out, I think it’s going to be great.

Samuel: Yeah, I’m excited about the entertainment side. So, I follow some of the hobbyist goings-on around first-person view drone racing. And these are people who put on goggles and fly drones around tracks. And I think it’s just — I think we’re only a year or two out before this is broadcast on ESPN, because just — it’s some of the most exciting things to watch.

Grant: Oh my God, yeah.

Samuel: Oh, yeah.

Grant: Just, the few groups out there that have started doing drone racing and have really, kind of, tried to address how to make watching that exciting — they’ve done an amazing job. Like, watching Drone Racing League and a few of the others, like — it’s going to be super sweet. I also, I’ve been a little disappointed in drone racing, just because now that I’ve started watching these guys that are doing it now — these guys and girls — they’re already so good. I’ve already had to, like…

Sonal: You can’t keep up, Grant.

Grant: I can’t keep up. I’ve already dashed my dreams of becoming a professional drone racer, because clearly I’m not good enough even for this early stage of the sport.

Eli: And the other thing, going along with the drone racing, which is all first-person view — I mean, just imagine putting on, like, an Oculus Rift headset, and just taking a drone up, and looking out of the drone’s camera and just being…

Sonal: And being in the sky.

Eli: So, you get to, like — it’s, like, being a warg on “Game Of Thrones,” right? It’s, like, you get to experience flight as if you were a bird or, you know, I guess as if you were a drone. I think it’ll be, like, really fun.

Sonal: I love that. It’s like, the e-sports version, and now we have, like, drone sports — kind of, like, a different version of, like, digital sports, essentially.

Samuel: Right, and I’d love to participate, and I’d love to get in on that, but I live too close to the White House.

Eli: And then I think, pushing it forward a few years, why do these all have to be unmanned systems, right? Why can’t you have an autonomously piloted aircraft that carries a human, right? Why shouldn’t we get human pilots out of the cockpits, just as we’re getting them out of, you know, away from behind the steering wheel in cars? You know, you can imagine a world where robots are us flying around, and that’s way cheaper, and you don’t have to carry a pilot. You don’t have to pay a pilot. Maybe we could have air taxi systems that are economical again. Maybe just huge safety benefits. I think in general, aviation — something like three-quarters of all accidents are pilot error. In commercial aviation, I think it’s about half. And so, you know, safer. What can we do to redesign airspace? The FAA is working on [the] NextGen airspace system, where there is more machine-to-machine communication. And so, you can have better routing. Well, what does that look like when that gets adopted? And how does that improve?

Sonal: That was actually one of the interesting quick sidebars about Jonathan’s notion of identity, to urge some kind of registration tied to, like, an entity — what was interesting to me — what I immediately thought of — just, like, IP address and the internet. Like, you essentially can have all these drone nodes communicate with each other and route information as a result of that.

Eli: Yeah, and some of the plans are actually very similar to how the internet is structured, in terms of public key infrastructure and, sort of, using…

Sonal: Right, exactly, pack it.

Eli: …like, SSL certificates.

Sonal: Right, exactly. It’s super fascinating.

Jonathan: There’s just no need to reinvent all of this technology just because it’s being used with drones.

Eli: Yeah. And I would say the last thing that I’m excited about that’s in aviation generally, but it’s perhaps somewhat unrelated, is supersonic. Because we haven’t — we’ve had a complete ban on supersonic in the United States, over land anyway, since 1973. We haven’t had a commercial supersonic jet since the Concorde.

Sonal: And what does supersonic do for us besides make a loud boom?

Eli: You could go cross-country in two hours, right? So I could come from DC, fly in, record a podcast with you, and then fly home. Like, that day.

Sonal: It’s like the hyperloop of the sky.

Eli: Yeah, hyperloop of the sky.

Sonal: I love that. I’m, kind of, excited by the art aspects. And when I think about this in the context — when people think of swarming drones — I love drone swarms. I think it’s amazing to see this orchestration of multiple drones in the air. And people view it as a very menacing thing, but I think there’s something very artistic, and elegant, and beautiful about it. But the other thing that really excites me — when I think of movies. Like, you know, when they redid the first three “Star Wars” movies — which were just awful, for the record, as everyone probably in this room agrees — I love the visual, though, of the fact that you had all these aircraft in the sky, and that people could jump from one aircraft to another. And even in this movie, “The Fifth Element,” which is this really lame, fun movie, there’s this amazing scene of people literally doing the same kind of thing. Like, they’re in the air and there’s layers — not just, like, one layer, but there’s layers of aircraft in the air. And it just gives a sense of actually living your life in the clouds, you know, where you can actually have, like, cafés in the air. You can do things in the air. I know that sounds a little crazy, but to me, when I think of airspace, I just think it’s amazing to me that we can now build upwards in ways that we couldn’t before. Well, thank you, guys, for joining the “a16z Podcast.”

Jonathan: Thank you.

Eli: Thanks to you.

Grant: Thanks for having us.

• 

  Sonal Chokshi is the editor in chief as well as podcast network showrunner. Prior to joining a16z 2014 to build the editorial operation, Sonal was a senior editor at WIRED, and before that in content at Xerox PARC.

  Follow Twitter
• 

  Eli Dourado
• 

  Grant Jordan
• 

  Jonathan Downey
• 

  Samuel Hammond

From Aaron Sorkin to Steve Jobs to Meredith Perry and Elon Musk, “original” thinkers — such as entrepreneurs — do a lot of different things to move the world to their visions. And many of those things (and traits) are counterintuitive, such as … Embracing procrastination. But there’s a catch: It’s about being the just-right amount of procrastinator, expert, or confidant. There’s a curvilinear relationship between too much and too little.

There’s also some surprising findings about why NOT to “start with the why” but with the how. Because sometimes the how is much more believable than the why. Especially when it comes to getting people to engineer things from ubeam to SpaceX. Or to really being able to tell the difference between communication vs. confidence vs. competence.

Ultimately, it’s all about being flexible, argues top Wharton management professor and New York Times columnist Adam Grant in his new book Originals. So how do we strike the just-right balance — whether making an entrepreneur or just trying to raise more creative, productive kids? Is the answer perhaps to immerse them in sci-fi books and video games? Well, J.K. Rowling could be the most influential “original” alive, argues Grant in this podcast… but not for the reasons you think.

Show Notes

• Discussion of what a non-conformist is, and different types of procrastination [0:00]
• Steve Jobs, Meredith Perry, and other originals [7:38] and the idea of the “skeptical optimist” [12:52]
• The importance of depth of experience [14:41], what it means to be an expert [19:49], and implications for entrepreneurs [21:12]
• How originals were influenced in childhood [23:42]
• Applying this research in organizations [31:34]

Transcript

Sonal: Hi, everyone. Welcome to the “a16z Podcast.” I’m Sonal, and today I’m here with Adam Grant, who is a Wharton professor and a New York Times columnist who covers the topics of work and psychology. And he has a new book out: “Originals: How Non-Conformists Move the World.” And we thought it’d be really interesting to talk to Adam about this, because there’s a lot of overlap between non-conformists and entrepreneurs. Welcome, Adam.

What is a non-conformist?

Adam: Thank you.

Sonal: The first thing I want to start with is how you actually define a non-conformist, because I think there’s a whole spectrum — and I use spectrum in both the psychological sense and just a descriptive sense. How do you know if someone’s a non-conformist, and not, like, a good rebel versus a bad rebel — like, someone who’s actually detrimental to society?

Adam: When I think about non-conformist, I’m thinking about people who don’t just reject the status quo for the sake of being different, or for disagreeing, but actually care about making things better for other people. So, I think about, you know, non-conformity as being — you can think about, you know, creative rebels who say, “Look, you know, there’s a standard way of doing things that isn’t right. And I think I can improve it.” Or, you can think about being a moral rebel and say, “You know, there’s a rule, a law, a policy, that doesn’t make sense, and it’s hurting a particular group. And I want to try to do something about that.”

Sonal: So the constructive form of non-conformism, essentially.

Adam: Exactly.

Sonal: One funny anecdote that you mentioned, which made me laugh out loud, is that a person’s choice of browser indicates where they fit. So, if you’re like a Chrome user, for example, or a Firefox Mozilla user, you’re more likely to be along those lines than not — it feels kind of obvious in hindsight, but it’s a really funny thing to come across. Like, how did you sort of come up with that?

Adam: Well, I wish I could take credit for it. I was sitting at a conference one day and an economist, Michael Housman, presented this study showing that he could predict your job performance and how long you stay in your job just by knowing what browser you use. And a lot of people don’t like the results of this study, obviously. And if you don’t know what browser you use, you should check Ask Jeeves right away. Basically, what he found was, Chrome and Firefox users were on average getting, in call center jobs, to customer satisfaction rates in 90 days that took Internet Explorer and Safari users 120 days to reach. And the Chrome and Firefox users also stuck around 15% longer in their jobs. So, the first instinct for me was, this has got to be a technical advantage, right? The people who are more computer savvy are the ones who are using Chrome and Firefox.

Sonal: That’s actually what I would think too.

Adam: But Mike ran the data, and there was no difference in typing speed or computer knowledge between the different browser groups. And what I eventually realized was, it’s about how you get the browser. Because if you’re an Internet Explorer or Safari user, those came pre-installed with your computer. They’re the default, and you just accepted the status quo that’s handed to you. Whereas, if you wanted Chrome or Firefox, you have to take a tiny bit of initiative and upgrade, or figure out, you know, maybe there’s something different out there. Let me try it out. And that turns out to be a signal of being the kind of person who doesn’t just conform and accept the defaults that are given to you.

Sonal: What are some of the other counterintuitive characteristics? I think, by the way, procrastination was super interesting to me.

Adam: I think we’ve all procrastinated at some point in our lives. And I will say though, I’m pretty much the opposite of a procrastinator. There’s a term for me — I’m a precrastinator. So I’m one of those people who, when I have a presentation to give in six months, I will wake up tomorrow morning feeling this tremendous sense of urgency to get it done now, so that I don’t wait until the last minute and it’s not hanging over my head. And as I studied originals, I found that many of them resisted that temptation and use procrastination productively. 

So, look at Leonardo da Vinci, for example, who spent roughly 15 years trying to finish up “The Last Supper,” and kept putting it off and working on these little optics experiments. And he felt like he was spinning his wheels, and wrote in this notebook over and over again, “Tell me if anything ever was done.” But ultimately, those diversions led him to make these discoveries in how to display light that dramatically improved his painting and made him the Renaissance man. And there’s a lot of research coming out suggesting that when we procrastinate, we give ourselves more time to incubate ideas. We do more divergent thinking. We’re less likely to be stuck in linear structured patterns of thought. And that can be useful if you want to come up with new ideas.

Sonal: I think I would probably nuancify the description of procrastination then, a little bit, because I think you either do it or you don’t. And I think what you’re actually saying is actually that there’s something more in between. Because when I think of how I procrastinate as well, you know, it’s not like you’re not working on it. You’re working on the back of your head, or you’re putting on the back burner, or you’re exploring ideas that are related to X topic. And then, when they suddenly kind of come together, you’re like, “Okay. Now is the time to actually put it together on paper, and this is the right time to put it out.” There’s another form of not procrastinating, where you just stick your head in the sand, and you just completely avoid it. Like, the kind I remember doing, like, in high school — you don’t even show up. I feel like there’s different flavors to those kinds of procrastination in what you’re describing.

Adam: Yeah. I think productive procrastination is intentionally delaying the start or finish of a task to make sure you have all the creative ideas that you might develop at your disposal. And that’s very different from just not engaging with the task at all, right? So I had a Ph.D student, Jihae Shin, who studied this, and she found that there’s a curvilinear relationship between how often you procrastinate and how creative your supervisors rate you in multiple organizations. So, if you put things off until the very last minute, you’re screwed, and you just have to rush forward with the easiest ideas. But there’s a sweet spot, where you put things off a little bit, you’re delaying, but you are kind of doing some unconscious thoughts, some incubating. And that allows you, then, to come forward with more interesting ideas and more unusual possibilities, because the first ideas that you generate are usually the most conventional and obvious.

Sonal: Right, exactly.

Adam: And if you just marched forward with those, then you’re limiting your field of vision. But it is key that you are, sort of, processing the task. So one of Jihae’s experiments randomly assigned people to procrastinate before developing business plan ideas by playing “Minesweeper.” And they were, after doing that, 28% more creative than people who jumped right into the task. “Minesweeper’s” awesome, but it’s not the reason to become more creative. The effect only held if they were told about the task before they played “Minesweeper.” So that, you know, while they’re working in the games, they’re kind of thinking about different business ideas. And that’s where the creativity came in.

Sonal: So what’s interesting about what you’re describing is a type of behavior. But in the real world, people have deadlines, and constraints that they have to follow, and things that they have to deal with. Like, if you’re in a company, or in school, or, you know, in just paying your bills on time. How does this sort of behavior play out in those scenarios? Like, can you be an original in one aspect of your life and then suddenly be very punctual about paying your bills? The psychological traits we have — it’s not like you get to pick and choose what arenas of your life you get to be a certain way.

Adam: Yeah. I think it’s hard to be an original without being flexible. In fact, that might be the most central defining characteristic of original people, is that they’re willing to bring different ways of solving problems to different situations. So, of course, there’s some tasks — actually, take any task where creativity is not important. If you’re paying bills, you don’t have to come up with novel solutions. And in fact, if you do, the IRS might come calling. That’s a task where you want to be structured, conscientious, focused, and, you know, sort of punch it in as quickly and efficiently as possible. I think that where originals end up procrastinating is when they know they’re working on a hard problem. One of my favorite examples of this is Aaron Sorkin, the screenwriter who’s known for the “West Wing” and the Steve Jobs movie. And he was interviewed once by Katie Couric, who said, “You know, basically, you drive your staff crazy, because sometimes you’re literally about to shoot a scene, and there’s still no script. Like, how do you put up with this procrastinating?” And Sorkin said, “You call it procrastinating. I call it thinking.”

Case studies

Sonal: It’s an interesting way of reframing that. I know you were talking about Aaron Sorkin writing the script for that movie, but I do think it’s interesting to talk about Steve Jobs as an example here because — and I definitely don’t want to be one of those people who elevates to this cult of Steve Jobs all the time. I think we need to be both critical and mindful of what he did and didn’t do. But one thing that struck me when I was reading his biography, the one that Walter Isaacson wrote, was this concept of his reality distortion field. And I think it’s very closely tied — like, you have to have some sort of flexibility of reality, a view of reality, in order to distort it for a better world and be able to envision a better possibility. But then there were times when it became just straight-up delusion. How do people navigate that balance? And what are your views on how this sort of thing played out with an example like Steve Jobs?

Adam: It’s a really interesting question. I feel like Steve Jobs is a Rorschach test, where you put him out there, and then whatever response you hear is much more revealing of the person answering than it is of Steve Jobs.

Sonal: Oh, that’s so interesting. That’s actually a really good point.

Adam: So, what does it mean that that was my reaction? I don’t know.

Sonal: That’s fair.

Adam: No. I mean, okay. So I’ll put my biases on the table here. I think that, for me, successful originals are not distorting reality as much as they are choosing when to present different realities. My favorite example of this is Meredith Perry. So Meredith has this amazing startup called uBeam, which is doing wireless power, right?

Sonal: Full disclosure, we’re actually investors in that.

Adam: When Meredith came up with the idea for wireless power, she went to some physicists and engineers, and they all told her it was impossible and she was insane. And she was in this, sort of, chicken and egg, catch-22 scenario, where she needed to build a prototype to prove it. But she couldn’t get anyone to work for her, because they all told her she was out of her mind. And at some point, she realized that instead of going to engineers and saying, “You know, I’m trying to build wireless power. Can you create this kind of transducer for me that I think will help me convert vibrations in the air into energy?” and having them say, “No, that can’t be done,” she started hiding her purpose and telling the engineers she was trying to recruit about the means that she wanted but not the ends. So, instead of saying, “You know, I’m trying to build wireless power, you know, I needed a transducer that will help me convert vibrations in the air into energy,” she just said, “Do you think you could build me a transducer with these properties?” And all of a sudden, instead of, “Hell no,” the answer was, “Yeah, I could probably figure out a way to make that work.”

And I think this is such a good example of timing which realities to present, as opposed to distorting them. It’s not like she’s lying to them. She’s not saying, you know, “I’m trying to build a transducer in order to tie my shoes faster.” She’s just choosing to reveal this information after she has more of the technology available, and people will be much more likely to believe her. And I think this is, to me, a fascinating strategy for originals, because we’re always told, especially if you watch Simon Sinek’s TED Talk, start with why.

Sonal: Oh, I love that TED Talk.

Adam: I do, too. And I think the point is largely right, that you have to — in order to motivate people to come on board with most ideas and visions, you have to explain your purpose. But if you have a really original idea, that’s terrible advice. Because your “why” sounds insane to other people. And so, if you’re Meredith, or if you’re Steve Jobs, for that matter, sometimes the “how” is much more believable than the “why.” And I think that that’s a skill that we could all work on, right, knowing when to say, “This is my ultimate goal,” and when to say, “You know, I’m kind of working toward this mid-level objective. Do you think you could help me with that?”

Sonal: The most fascinating aspect of the anecdote you just shared, to me, is that she essentially had a very original idea, and had to take a very non-original approach. She had to use people’s non-originality in order to get them to deliver what she needed incrementally to get her to the next goal. So it’s almost like, you have this interesting interplay in an organization between the people who have these characteristics and people who don’t, and then how you, sort of, interact with each other and how you switch contacts based on that.

Adam: Bingo. And we see this with lots of great entrepreneurs. This is exactly what Elon Musk did with SpaceX. He didn’t recruit his team by saying, “Let’s go to Mars.” He said, “Let’s see if we can get a rocket into orbit, and then back.” And once they saw that was possible, it’s a little bit more acceptable to start talking about whether we can colonize another planet.

Sonal: But don’t you want people who believe in your vision? Because when you say, like, Elon Musk recruited those — some of those folks, I mean, okay, clearly, he needed people who aren’t just so pie in the sky that they actually need to build what he’s envisioning. But at the same time, I feel like one of the defining characteristics of startups, and sometimes for really good leaders of startups, is this collective of people who believe in a similar vision. And not to sound, like, cult-like or like it’s a mission, but more like — it’s a way to really align people around like you’re doing something. Like, I know I would not want to work for someplace where I don’t believe in the product, for example.

Adam: Of course. I think, though, that the people that I would want to hire, and that I think Elon wants to hire are skeptical optimists.

Sonal: Let’s break that down a little bit.

Adam: Yes. So the optimism part is, you believe that the future can be better than the present and the past. And when you consider possibilities, you’re willing to have hope and see upsides. The skepticism is saying, “I’m not going to be convinced by every Pollyanna idea that somebody throws at me. I want to see the hard evidence, you know. Show me that this is doable.” And so, you know, I think actually recruiting people who think that they could do, for a fraction of NASA’s budget, something that NASA has never been able to do — those are people who are willing to believe in a vision. That buys you the optimism. But I don’t necessarily want to recruit people who believe on day one that Mars is a realistic destination in the next decade or two.

Sonal: I love that you said that phrase, of skeptical optimism. Chris Anderson, the former editor in chief of WIRED, used to always kind of describe our mission when we were at WIRED as being informed optimists. Like, the bar for a story getting through was, like — it’s not only optimistic about the future of technology, but that there’s a level that it’s informed. It’s not just conspiracy theory. I love that. And I have to say, by the way, though, Adam — I think that you’re not the only one doing this, but a lot of people use Pollyanna as this way of saying Pollyanna-ish ideas. Like these, sort of, fluffy, idealistic, feel-good, rose-colored glasses, view of the world. And in reality, Pollyanna is actually a story about overcoming hardship. The reality of the movie is — and I hope I’m not giving spoilers, because that movie is like 50 years old, but — is that she has a very optimistic view of the world, but then she gets paralyzed. And she has to then overcome her own fear of her own limitations in order to hopefully overcome that disease.

Adam: The movie is very much about the importance of a certain kind of optimism for overcoming tragedy. I think that, obviously, the modern use of the term has evolved to, sort of, focus on “your glasses are so rose-colored that you might be a little bit unrealistic” or “too easily duped — like, you’re gullible.”

The meaning of “expert”

Sonal: And that’s totally fair. I’m just, like, putting out my own personal agenda to defend the movie. Let’s talk a little bit more about some of the more counterintuitive characteristics. What are some of the surprising things that define originals, based on your research in this book?

Adam: I think one of the things that really caught me off guard is that they tend to have less expertise than a lot of their peers.

Sonal: Ooh, interesting.

Adam: Yeah. So there’s this curvilinear relationship between expertise and originality, where when you’re trying to come up with new ideas, you obviously need to know a field or a domain well enough to have an understanding of what’s possible and what’s been done in the past. Right? So, Einstein couldn’t have come up with his theory of relativity without knowing something about physics beforehand and studying Newton. But it’s not a coincidence that he was relatively new to the field, because the longer you learn a particular domain of knowledge and the more you internalize it, the easier it is to become entrenched, where you basically take for granted assumptions that need to get questioned.

And what you see with a lot of successful originals is they have this great combination of breadth and depth. Where, yes, they know the domain reasonably well, but they’ve also immersed themselves in ideas outside that domain to make sure that they’re seeing things from a fresh perspective. One of the ways you see this is — actually, if you look at Nobel Prize-winning scientists. One of the things that differentiates them from their peers is they’re much more likely to have artistic hobbies. So, on average, Nobel Prize winners are twice as likely to play a musical instrument. They’re seven times as likely to paint and do other kinds of art. They’re 12 times as likely to write creative fiction, poetry — and they are 22 times as likely to act, dance, or perform magic.

Sonal: Oh, my God, that’s actually really funny.

Adam: Yeah, as a former magician, I love that stat. But…

Sonal: I think Aaron Levie, the CEO of Box, is actually a magician too.

Adam: That is right. And I think that, you know, obviously, this is not all causal. The same curiosity that draws people to be creative also tends to pique their interests in these kinds of artistic hobbies. But sometimes, engagement with these hobbies helps with the discovery of original ideas. Einstein said that the theory of relativity was a musical thought, and that it came to him because of all the time he spent playing the violin. And Galileo — one of his greatest discoveries was being able to spot mountains on the moon for the first time. What’s remarkable about Galileo is, he was looking through his telescope at an image that other astronomers had seen, but he was the only one who recognized that the shading he was observing was mountains. And the reason for that was, he had specialized in a drawing technique that used a very similar kind of shading. And so he knew that that was how you represent a mountain. And in that case, if he had not been an artist, he never would have made that discovery.

Sonal: So, that’s a case where the art actually influenced how they viewed certain things. So, I mean, you’re describing two things. One is this, sort of, co-occurrence of this creativity in a field or experience in a field. But you’re also describing something where you’re talking about the exact right amount of experience and expertise, and you’d mentioned it as being curvilinear. And that’s the second time you’ve mentioned the curvilinear as an example. So, it sounds like there’s always, like, a sweet spot — where there’s not too much, not too little, but there’s just this one — just right amount. How do you know what that sweet spot is? Like, where do you sort of fall off into one side of the curve or not?

Adam: I think that part is much more art than science. As much as it pains me as a social scientist to admit it. I wrote a paper about this a couple years ago with Barry Schwartz, where we argued that everything in life is an inverted U. And that, you know, if you take any strength, or virtue, or positive experience, you can find too much of a good thing — where, you know, like, okay, if you’re too confident, we get narcissism. If you’re too generous, we get altruistic self-sacrifice. And you can play this out for any trait that you probably see in a positive light. I think the only thing that we really know at this stage about how to find the sweet spot is that good things satiate and bad things escalate. So, the further that you move down the positive end of something, the more likely the costs are to start outweighing the benefits. And I think you can only usually see that by looking at the results. So, you know, in the case of expertise, right, the question is — okay, when you start to generate ideas, are you finding yourself trapped by what you already know in the field? As you’re, you know, evaluating different kinds of ideas, do you consistently gravitate toward what’s already accepted and proven?

Sonal: It’s really funny that you guys tried overfitting the U-shaped curve to all these different things as well. But I think that there’s a gender or racial background, or other background effect that can play out here differently. I’m thinking of cases where a lot of women, myself included, will sometimes underplay their expertise. Because, you know, I’ve seen a lot of my former male colleagues — like, they would be experts in things that they necessarily weren’t. But they have the confidence to say that they were. And to me, that wasn’t a sign of confidence. I’d actually get really irritated when people said, like, “Be confident and say you’re an expert in that.” And I’d be like, “I’m not going to frickin claim to be an expert on something I’m not. Like, I don’t think that’s confidence. I think that’s just being full of crap.” I think it’s really interesting, because I think some of these things also play out where there’s an interaction effect between people’s background, whether it’s gender, race, or privilege, or other things that have influenced how they grow up. Like, how did you see that play out in thinking about originals?

Adam: Yeah, it’s interesting even how you set up that comment, right? Because a man would have said, “It’s a fundamental fact,” as opposed to, “Here’s what I think and I’ve kind of noticed…”

Sonal: Oh, God, you’re right. It’s funny because when I hear myself on the podcast, I’m always like, I really got to take out some of those caveating words I use, like kind of, maybe, what do you think, vocal fry, whatever. All the stuff that I think I tend to do sometimes, and I hear it and I cringe. Other times, I’m like, “Fuck it. It’s who I am, like, take it.” You know what I mean? Like, anyway, you’re totally right though. So…

Adam: Well, let’s take that a step further, though.

Sonal: Yeah, let’s talk about it.

Adam: Do you really want to take that out? I would say maybe not. So, Zak Tormala at Stanford has these studies showing that experts are believed more when they express uncertainty.

Sonal: I like that, because I actually think that is what a true expert is. It’s hubris to claim to be an expert in something you’re not. At the same time, it will say, coming full circle to your point, I have observed that the people who actually go out and start companies — and this takes us to entrepreneurs — are people who have such belief in an alternative view of the world — even if they’re not experts in X, Y, or Z, that they’re the ones who go out and do it. And I admire that. I do think it takes a certain amount of knowing that you can do that. So, like, what’s the difference? Is that a confidence thing? Is that an experience thing? Is that an original’s mindset? I mean, where do we figure out, like, what’s making an entrepreneur tick there?

Adam: There’s a huge debate about, you know, how much does confidence really drive success. And, like everything else, is curvilinear. Right? So, if you have too little confidence, you never act. And if you have too much confidence, then you end up getting complacent and missing out on threats and opportunities that you underestimate. I think that where I would come down on this is — Susan Cain is fond of saying that there’s zero correlation between who’s the best talker and who has the best ideas. And that’s true empirically. The sad thing is though, a lot of us take confidence as a signal of competence. And I think we need to stop doing that. I think if we stop doing that, we’ll see many more women and minorities rise into leadership positions, because we’ll see that oftentimes they are better prepared, but communicating in such a way that didn’t always signal, you know, the confidence behind the idea.

I would also say that the self-esteem movement has been disastrous for entrepreneurs, in the sense that, like, becoming a successful entrepreneur is not about thinking that you’re special. It’s about believing that, you know, somebody else could do this. Maybe I could too. And I think that confidence should come as a consequence of competence. Right? So instead of saying, “Well, I need to build my confidence, and then I’ll be successful,” no, let’s develop grit. Let’s have a growth mindset. Let’s work as hard as we can to achieve success. And then confidence will be the natural product of that.

Sonal: When you know something really well, or you feel very passionate about it, it feels true to you. The way I feel and, say, personally about editing — you feel incredibly confident in that, because it is a consequence of competence. You quoted Susan Cain’s work, and she’s the author of “Introverts.” That — there not necessarily being a correlation between how one communicates and that confidence, and the competence. I think it’s interesting, though, because — when it comes to leadership, as you know, and I’ve seen this with entrepreneurs as well — you are motivating people by being able to communicate your vision as well. And it has implications for hiring, for everything else. So, it actually really does matter. I mean, I don’t think we can easily dismiss that out of hand either.

Adam: Let’s be careful not to overrate confidence. But, yeah, it plays a role in our lives. I think for most people, grounded confidence comes from accomplishments, not the other way around.

Sonal: Are there any other takeaways? Like, I’m actually curious, not just for entrepreneurship, but like, for education, for raising kids. Like, you had an article in the New York Times this past weekend that talked about, like, the mistakes a lot of parents commonly make — like, you know, over-programming their kids. I mean, what are some of the implications of your research?

Non-conformists in childhood

Adan: One of the things that surprised me the most is, when you study originals and look back at their childhood histories, you see that their parents often focused on creating a really strong moral compass. So, there are these brilliant studies of creative architects, where you look at the people who are nominated consistently by the most respected people in their field as truly original. And then you compare them to their peers, who are technically skilled, but haven’t necessarily done anything creative. And in the original studies, there were extensive interviews — not only with the architects but their family members, lots of observations, assessments. One of the things that came out was that the parents of the creative architects tended to focus less on rules and more on values. But that they gave their children a lot of freedom to actually determine their own values. And what happened was, the architects, then, develop their own values, which, you know, were grounded in a moral framework — that when other people said your idea was ridiculous, they were much more comfortable standing their ground and saying, “Well, this is who I am. And I’m going to try it anyway.”

They also, you know, in addition to just being comfortable with nonconformity, they were much more likely to be concerned about, you know, what is my contribution to the world, right? When I’m constantly asked to think about what are the consequences of my actions for other people, I want to leave the world better than I found it. And I think we could probably all do a better job. I know I can as a parent — you know, really having that conversation about, you know, here’s some broad values that we think matter. How do you want to live by those?

Sonal: Are those things you think communicated verbally or through modeling? Because one thing that comes to mind when you describe that is, that probably explains a lot of immigrant children’s success in the first, second, and third generations. I know the effect tends to disappear after the third generation in past studies. But I wonder, with immigrants, it’s kind of, like, this epic that you get, because you just watch it. I mean, I wouldn’t speak blandly for every ethnicity out there. But a lot of immigrant groups — you’re not having those conversations with your parents in any kind of articulated way. Like, it’s not a — it’s a very nonverbal type of culture in that way. And so, you can actually learn those things just by watching them work hard and try to contribute something to the world.

Adam: Yeah. Modeling effects are often stronger than conversation effects. In part because, you know, role models — when you see the behavior, it teaches you how to do it. It tends to raise your expectations of what’s possible. Conversations don’t always have that impact. They also sometimes, you know, create this reverse psychology reaction of, “Don’t tell me what to do, I will now do the opposite.”

Sonal: Exactly. You also mentioned role models, the existence of. Because one thing that comes to mind as well — and I’m thinking of classic developmental psychology studies of resilience in orphaned children [who] were orphans in previous world wars. And one of the consistent findings that came through over and over again is, no matter what else those children did not have, one of the greatest predictors of resilience was having a person they could look to as a mentor or as a role model. How does that play out with — taking it a step further — like, beyond survival, to becoming a productive non-conformist?

Adam: I think that parents don’t necessarily have to be role models. I think that everyone needs a role model in order to have some kind of vision for what it looks like to make a mark. But we can find role models in some pretty unexpected places. There’s some classic research looking at patent rates and innovation trajectories of entire economies. And my favorite finding out of this body of work is that you can predict the spikes and falls in U.S. patent rates by coding themes of original accomplishment in children’s books.

Sonal: Really?

Adam: Yeah. So if you look at the children’s literature of a particular era, when that literature starts to include examples of people accomplishing things and succeeding in ways that are new and innovative, patent rates actually spiked 20 to 40 years later.

Sonal: Was the era where Dr. Seuss, like, published “Cat in the Hat,” like, super high on patents? Like, I mean, those kids were fully defying their parents with, like, the cat mouse.

Adam: Yeah, I think there’s a case to be made there, and more recently “Oh, The Places You’ll Go,” I think was the most popular children’s book in the ’90s, which was all about choosing your own path. And what’s fascinating about this is, you know, in part is just a reflection of the culture. Right? So, you know, when innovation becomes more important, we tend to write, and buy, and read children’s books that are innovative. But there’s, I think, a story to be told about how these books actually shape originals. What you will find is that if you talk to some of the great originals of our time, they are constantly saying their favorite books as kids were stories of, actually, other kids who were, you know, inventing things, or accomplishing things that were impossible previously. If you ask Peter Thiel and Elon Musk to name their favorite childhood books, they both pick “Lord of the Rings.” Jeff Bezos and Sheryl Sandberg both said “A Wrinkle in Time.”

Sonal: Yeah. I used to love Madeleine L’​Engle.

Adam: Yeah, we all did. Right? And what is that story about? It’s about a young girl bending the laws of physics and traveling through time.

Sonal: Yes.

Adam: Like, if that doesn’t get you thinking about making an original contribution to the world, I don’t know what does.

Sonal: Well, they’re both sci-fi books. They all fall in the sci-fi genre, which is interesting in and of itself.

Adam: They do. And it’s not a coincidence, by the way, that you can trace a lot of modern inventions to, you know, the writings of Jules Verne and the technology that we watched on “Star Trek.” And, you know, I think in a very real way, these fictional role models give children the freedom to define their own niches, and imagine doing things that don’t exist or aren’t currently considered possible. 

I really think “Harry Potter” is going to have this impact. I would probably put my money on J.K. Rowling as the most influential original alive, because “Harry Potter” sold more books than any other series except maybe the Bible. So it’s reached a lot of people. You have kids saving the world and inventing spells in ways that spark lots of creative thinking. And there’s also academic research now showing that, after kids read “Harry Potter,” they become less prejudiced. So, they learn not to stereotype people in the way that wizards look down on muggles. And so, you know, that’s a pretty good trifecta, right, reaching hundreds of millions of people, getting them to think in original ways, and making sure that they don’t have these strong in-group, out-group boundaries.

Sonal: I’m fascinated by that. I mean, it’s one of my all-time favorite series. So, I’m personally incredibly motivated by “Harry Potter.” But to hear that it can have that effect on people is incredible. And to your point, it does indicate how culture does shape the sort of thinking that comes out in each generation. What’s more fascinating to me is the recent uptick in the last 5 to 10 years of young adult literature that is — really strong female characters. You know, like, obvious examples include “The Hunger Games” with Katniss Everdeen, the “Divergent” series by Veronica Roth. And there’s, like, countless others. I mean, I read, like, one a month. They’re just amazing.

Adam: I’m glad I’m not alone.

Sonal: No, and you’re not alone. And what’s actually refreshing is when we were growing up, do you remember even having that many female strong characters out there? Because I don’t remember that. I used to read stuff like — I mean, I remember reading like David Eddings, like, “The Belgariad,” or, like, other things that — they were male characters. There weren’t, like, strong female characters, or if there were, they were, like, adjuncts to the male character instead of the main character.

Adam: Yeah. As I think about it, like, the female protagonists or heroes, like, in my childhood were Nancy Drew, Wonder Woman, and maybe Penny from “Inspector Gadget.”

Sonal: Exactly. And they’re great strong women. But it’s very different than today, where you have Katniss Everdeen, like, murdering people for survival in “The Hunger Games” arena. I mean, they’re strong, hard characters. And I think that’s incredible. One other interesting thing, though, is I think we’re only talking about books as an influence on literature. I think it’s important to mention other forms of narrative, like TV shows. Things like “Game of Thrones” where there’s really no narrative arc. It’s this endless story that keeps developing. Or even, like, video games. Not all video games have a fixed content arc if they’re not a content-based video game. So, I think that’ll be interesting to see how that plays out in what you’re describing here, because there might not be as many examples in the future of that sort of thing.

Adam: I think so, too. I’d love to see the evidence.

Implications for organizations

Sonal: So far we’ve been talking about a lot of interesting themes, and research, and anecdotes that are really centered on outliers as cases, or as individuals. One thing I’m interested in, particularly because you’re a professor at Wharton, and you studied management science and things connected to this. How does this all play out systematically, like in the organization, for example? Because we all live — like, we probably spend more of our day inside a firm than we do in our own families. And so, I’m really interested in hearing how these dynamics play out in groups, and culturally across organizations as, sort of, containers have that sort of culture.

Adam: Part of “Originals” is about how individuals can champion new ideas, and then how parents can try to nurture kids to think differently. But I think it’s just as important, if not more so, for us to understand how leaders build cultures that welcome original thoughts and that fight groupthink. And there are a couple of things that most leaders do wrong, if you look at the data. First thing is, hiring on cultural fit. So, one of my favorite studies looks at over 200 Silicon Valley startups and tracks them before and after the dot-com bust.

Sonal: I like where this is going. This is going to be interesting.

Adam: Yeah, this is fun. So, you see that there are three prototypical ways that founders hire when they’re looking at talent. Some hire on skills, so they’re looking for people who have a certain set of competencies now. Some hire on potential. So, it doesn’t matter what you know today and — how much do we think you can learn? And then a third group hire on cultural fit. Do you share our values around here? And when you track the founders’ firms, what you see is that the founders who hire on cultural fit are less likely to see their startups fail, and they’re more likely to make it to IPO. And then after IPO, their firms grow at a much slower rate than the ones that hire on skill or potential. So, cultural fit helps you grow and take off, and then it causes you to stagnate, and maybe increases your risk of failure. Why is that?

The basic explanation is that cultural fit is a great way to get groupthink. You have these founders who are incredibly original at the outset, and then they hire a bunch of people who see the world exactly the same way they do, and they end up cloning themselves, and getting homogeneity of thought instead of diversity. So, I think the solution to this is, at some point, as your organization grows and you need to start questioning the very values that made you successful in the first place, you want to stop hiring on cultural fit and start hiring on cultural contribution.

Sonal: Ah, so what’s the difference there? Because I mean, I think people would — I would conflate that.

Adam: Yeah, so cultural fit is, basically, “Who are we, and how do we bring in people who are just like that?” Cultural contribution is asking, “What’s missing from the culture, and how can we bring that to the table?” So, trying to figure out who’s going to enrich the culture and add diversity of thought to it, as opposed to just replicating it. Of course, you can also overcome some of these problems if diversity is one of your core values.

Sonal: Right. I agree. I don’t think you can just tack it on, you know, as, like, a sidebar silo thing, it has to start from the leadership. It has to start from the top, like, you have to believe in something. You have to believe — whatever values leaders believe in is essentially what the company is going to believe in.

Adam: Yeah. And I think, you know, you can screen on this in really interesting ways. So, one of my favorite interview questions is to ask people, “Tell me what’s wrong with our interview process and how you would improve it.” Or, you know, more broadly, “Based on what you know about the culture so far, if you were in charge here, what are the three biggest changes that you would make?” People aren’t willing to give that kind of constructive criticism or bring in dissenting opinions in the hiring process — I’m a little worried about whether they’re going to do that moving forward.

Sonal: One thing I’m really fascinated in — because you brought this up a number of times in the book — is power differentials. And that’s power differentials between people, like, speaking truth to power, people have less powers, people who aren’t in management but who are contributing original thinking — even power differentials with people who are not represented, like, whether you’re underrepresented in the organization. How do those play out in this scenario?

Adam: Unfortunately, the evidence suggests that lots of people who come from non-dominant groups are less likely to get heard when they speak up. So, Sheryl Sandberg and I wrote an op-ed last year called “Speaking While Female,” where we covered a lot of evidence that when a woman and a man make the same point, the man gets a big pat on the back and people start to rally around him. And the woman is either not heard, or punished for being too aggressive. And you see the same effect with different kinds of minority groups, because this is not fundamentally a gender effect. It’s a power effect. Right? So the groups that are perceived as, you know, as not occupying high-status positions in society get stereotyped as, you know, sort of needing to find their place. And, you know, they’re often perceived as stepping out of bounds when they’re just trying to make suggestions or get their opinions heard.

You know, I think, from an organizational standpoint, we need to be especially careful to welcome dissenting voices when they come from people who don’t look like everyone else, and who don’t come from the most common backgrounds in the organization. But, I think, from an individual perspective, one of the opportunities to overcome these biases is to make sure that you earn status before you exercise power.

Sonal: That’s a good way of putting it.

Adam: Yeah. Once you’re recognized as an expert, an authority, as, you know, having made valuable contributions to the organization, it’s much easier for people to look at your suggestion and say, “Yeah, you know, like, you’ve given a lot here, so you have license to deviate from the majority’s preferences.” Or, you know, “You’ve shown that you really care about the group, and you’re committed to making the team successful. And now we’re going to interpret your idea as, you know, an effort to help us get better, as opposed to a threat or a challenge.”

Sonal: Of course, that does, by the way, assume a very meritocratic organization, because there are plenty of cases where there’s a very nepotistic flavor of earning status. Like, you happen to work with someone before and you believe in them, or you’re just friends and buddies, and you guys party together, or — you know what I mean, like, not everything necessarily plays out in a meritocratic way. There is that effect as well, I think.

Adam: So true and so sad.

Sonal: Yeah, it is.

Adam: You know, I think one of the other opportunities for leaders on this is, when making decisions, almost every leader I’ve ever worked with has made a point of assigning a devil’s advocate, and said, “You know, look, we need to make sure we have, you know, divergent thinking in the room. We want to hear all the dissenting opinions. So, you know, I’m going to ask a few people to represent the opposite.” The sad thing is, if you look at 40 years of research by Charlan Nemeth at Berkeley, she shows that devil’s advocates rarely work. You know, when you’re given the devil’s advocate role, you don’t argue as forcefully as you should. You’re like, “All right. So I’m going to take the opposite perspective. Okay. Now, let me go right back to what I really…”

Sonal: Or you can have something like the New York Times, where you have like devil’s advocate. I mean, I don’t — I know you work there, so you may or may not be able to comment on this — but where you have, like, someone like Margaret Sullivan, who plays the role of public editor, and I love reading her stuff. It’s fascinating. But even though it’s public, and it’s a criticism of the New York Times, it’s, like, a siloed thing. Like, do people actually then do anything with those takeaways?

Adam: I have no idea. But, you know, the evidence would suggest that a lot of people don’t, because the other side of this is — just as you don’t take the role seriously enough, your audience doesn’t either. They’re like, “All right, so we know you’re just playing the role. So, we’re just going to let you give your lip service to it and stick with what we already believed.” So what do you do? What you do is, instead of assigning a devil’s advocate, you unearth a devil’s advocate. It’s only authentic dissent that has the best chance of working. That means you need to find people who genuinely disagree and invite them into the conversation. And guess what? That’s more likely to be minority group members, right, who come from a different perspective and a different background. And, you know, this is one of the things that’s easy to talk about and hard to do, which is — you actually have to know what people think. Right? You have to go out to meet the silent minority and say, “Look, you know, we really value your input. Let’s find out what your reaction is to this idea.”

Sonal: That’s actually a good way of putting it. There’s so much more we could talk about, but I think people should just go ahead and read your book, “Originals: How Non-Conformists Move the World,” which is out now. And thank you for joining the “a16z Podcast.”

Adam: Thank you for having me.

• 

  Adam Grant
• 

  Sonal Chokshi is the editor in chief as well as podcast network showrunner. Prior to joining a16z 2014 to build the editorial operation, Sonal was a senior editor at WIRED, and before that in content at Xerox PARC.

  Follow Twitter

We often hear stats like “more people have mobile phones than toilets” about places like Africa, but what does that actually mean for people? “It is b.s.,” (no pun intended!), argues one of the guests on this episode.

Then there are statistical predictions about mobile penetration and usage — for example, that there will be one mobile phone per African within just three years. But how do we make sense of such stats, in context? It may make more sense to measure household per device(s) not just per person … and whether women, too, truly have such access given power structures. Finally, since access is mainly about affordability, what good is a smartphone with an internet connection if data plans are prohibitively expensive? Watching just one 7-second YouTube video could cost a low-income African family an entire month of groceries. Yet mobile and wi-fi may be collapsing and renegotiating Maslow’s famous hierarchy of needs. Perhaps the truth is at the heart of these contradictions…

With three experts from various backgrounds and regions of Africa — Nkiru Balonwu of international music media company Spinlet; Alan Knott-Craig of free wi-fi non-profit Project Isiszwe; and Nanjira Sambuli of Kenyan startup incubator iHub — we explore the nuances of what connectivity in Africa really means. How does this change app design? What does it mean for doing business with Africa? What does it mean for businesses in Africa trying to compete with Silicon Valley (is there really local advantage)? All this and more in this episode of the a16z Podcast.

photo: David Mutua

Show Notes

• Discussion of Africa as a collection of countries and regions, not as a monolith [1:40]
• The importance of mobile phones and issues with internet access [5:42], including shared devices and types of content that are popular [14:09]
• Africa’s growing middle class [20:06], issues with apps and data usage [25:21], and the rise of WhatsApp [29:30]
• Discussion of local advantage when doing business in Africa [34:58] and women in technology [40:46]

Transcript

Sonal: Hi, everyone. Welcome to the “a16z Podcast,” and today this podcast is all about technology and Africa. Now, that’s a really huge topic, so we have experts from various backgrounds and regions of Africa to help us cover a lot of interesting nuances behind the stats that we typically hear, as well as a lot of the buzzwords that we commonly hear.

Our three guests to help us do this today are Alan Knott-Craig who runs Project Isizwe, an NGO based in South Africa that helps African governments get free Wi-Fi in poor communities. Their goal is to create internet access as a utility. Prior to founding Project Isizwe, Alan was an entrepreneur who ran Mxit, which is one of the largest mobile social networks created in Africa. Nkiru is our next guest, and she is the CEO of Spinlet, a digital media company that focuses on African-centric music and has music available on iOS, Android, and also via web browser, and we’ll talk more about why — about that later. Spinlet is headquartered in Nigeria, but they’re also elsewhere in Africa and also in Europe, and in the United States. Fun fact, Nkiru was actually an IPO lawyer who joined the company as general counsel before becoming CEO, and she actually didn’t think she was ready to be CEO until she read Sheryl Sandberg’s book, “Lean In,” which actually inspired her to go ahead and take the CEO job, and now she’s been in it for the past year and a half. 

And, finally, we have Nanjira, the research lead at iHub. iHub is an incubator, not an accelerator, although it does lead to acceleration of startups. It’s based in Kenya and provides a physical space for connecting entrepreneurs, a way to test their ideas, provide info to them, and much more. And they do a lot of interesting research as well, and some of that will come up in this podcast. 

Understanding the diversity of Africa

Okay, so that’s our guest on today’s pod, and we’re honored to have you all join all the way from various parts of Africa. I think the first thing I wanna start with is, sort of, just this notion of — when we talk about people talk about technology in Africa, the first notion I wanna just jump right into is actually de-homogenizing what people mean when they say Africa, because Africa is clearly a huge continent, but we have a tendency to refer to Africa as, like, one big place. And, you know, I don’t have any personal experience in Africa. My mom was born and raised in Uganda, and I studied African literature in — particularly Nigerian authors, like, Flora Nwapa.

Nkiru: Yeah.

Sonal: Yeah, Chinua Achebe and others, Wole Soyinka. But, anyway, what I’d love to hear from you guys first is how would you, sort of, define, sort of, what you consider some of the universals when people refer to Africa, and then what are some of the more, you know, things that people need to pay attention not to clump in together when they’re referring to Africa the continent versus, like, different countries.

Nanjira: The notion seems to be, especially in business, that, you know, buy one get 53 free. So, you know, if you get your business working in South Africa, the model will work in Kenya and Nigeria. And so a lot of business mistakes are being made from that notion, and it’s that sort of actually homogenizing of it. But, yes, there are notions where — I mean, the little cultural nuances that make it seem like a country, but they’re very contextual and very — they’re still emerging. We’re all still learning who we are as a collective. And so, yeah, I hope maybe by 2017, we still won’t have to say Africa — you know, to remind people Africa is not a country.

Nkiru: I definitely agree with Nanjira and her experience. It’s sort of very similar to mine when people say Africa, I sort of, like, “Oh, there’s 53 countries, and very different experiences.” And particularly when people say, “I’m an expert in Africa.” Or “in Africa.” And I’m thinking, “What does that mean?” I live in Lagos, and I’m not even an expert in Lagos. On Lagos, anything that has to do with Lagos, there’s different people, different kinds of people living here. We’re really, really extremely different. Of course, there’s similarities, like everybody else, but the differences are really, really — they’re quite harsh, the differences. And so when you say you backpacked across the continent for three weeks, and then you then say you’re an Africa expert, it’s a bit irritating from my perspective. But just generally speaking, one of the more — maybe interesting things is when, you know, Americans are doing African voices in film, it’s always a South African accent trying to be Nigerian, or a South African trying to be, you know, a Kenyan.

Sonal: Right. Well, Alan, what’s your take on this? You were born and raised in Africa, and in South Africa specifically and, you know, interestingly, given that your company focuses on providing free Wi-Fi and working with different governments and communities, you must actually see more of the commonalities across the regions.

Alan: The way I look at it is, you’ve got, kind of, Arabic Africa, which is very much the Northern part from Sahara up. And then you’ve got sub-Saharan, which is a little bit more homogeneous, and then sub-Saharan you kind of divide it between the West African trading bloc and the East African trading bloc and the Southern African trading bloc, which is — you know, Southern Africa is about, you know, SADC, really — South African Development Community. South Africa, Mozambique, Zimbabwe, Botswana, Namibia. And then you’ve got English-speaking and French-speaking Africa, which are vastly different, kind of, communities. So, it’s impossible to, kind of, lump it all together in one big <inaudible>. And, you know, from our perspective, we kind of — very much, you know, at least from an English-speaking perspective — sub-Saharan Africa is as close to a homogeneous market as you can get.

Sonal: Oh, okay, that’s actually interesting.

Alan: But I just wanna say also, you know, there are some stereotypes that the people should know. So, for instance, South Africa is a bit like Germany in Europe, you know. No one in Europe likes Germany. Everyone drives a German car. So, you know, South African product is respected, but South Africans aren’t always the most welcome. You know, Kenyans are well known for talking a hell of a good game, and the guys from Nigeria, they know how to make money.

Nkiru: Oh, no.

Mobile phones and internet access

Sonal: We have stereotypes like that in regions of India as well. Okay, great. So, you guys — so then, what is the major commonality then, because, you know, based on what a lot of the reports that I read about technology in Africa, a big focus is talking about mobile as a sort of this great — I don’t wanna say the word equalizer, but at the same time, like, it does have that sort of a power.

Nanjira: It’s true. Mobile has really been that technology that has been perhaps most disruptive, and in terms of one of the, sort of, development speak terms being leapfrogging. It’s helped them leapfrogging so many things — communication, access to finances. And this is where we go — again, cue M-Pesa.

Sonal: Right. M-Pesa as in, the company that came out of — the payment system that came out of Kenya, and it’s basically mobile payments.

Nanjira: Essentially. Now, the thing is — I think I’d love us to not restrict ourselves to thinking that mobile is the only, sort of, tool through which advancements and developments can happen as far as ICTs go. There’s so many other things that come with the idea of a mobile phone. So, yes, it’s been that technology that has advanced so many things, but now we’re talking about the infrastructure that needs to be underlying access to [the] internet, or access to certain services that are offered via the internet. And so, that also goes beyond just providing through the narrow spectrum of mobile, though that has obviously been the, you know, the front leader, the trailblazer, so to speak.

Alan: And in terms of commonality, it seems to me that most countries in Africa just are not connected. So, that’s whether it’s roads, or trade, or financial services, or actual, you know, bandwidth. You know, having, personally, you know, tried and failed a number of times. I’ve done okay here and there, but for the most part, any internet startup in Africa is really struggling to, kind of, get any traction on its local market, because there’s just not enough people on the internet. And whilst 3G is pretty ubiquitous, and LTE is becoming even more ubiquitous, the question is not so much about accessibility. It’s a question of affordability. And the vast majority of people in Africa can’t really afford the kind of going rates for 3G. Just to put things in perspective, if you spend — if you watch a 7-second YouTube video on 3G at South African data rates, it’ll cost you about $20.

Sonal: Oh, my God. That’s a lot.

Alan: Yeah, that’s a lot of money. It’s, you know, a monthly grocery bill for your average — your low-end household. So, where we’re coming from, you know, after trial and fail here and there, around trying to get more people on to apps and internet applications, decided to go back to square one, and kind of get people on the internet before you try — start selling them things on the internet. And, you know, from my perspective, it just kind of dawned on me that the only way it’s really gonna happen in our environment is if the government starts getting very much involved in infrastructure, subsidizing it, and making it free. And in a country like South Africa, water and electricity is constitutionally a right for everybody. So everyone in the country gets a basic quota of water and electricity, and the government pays for it from tax money. And we’re trying to, kind of, push a model, whereby public Wi-Fi is provided by the government, as well as subsidized by taxpayers, and everybody is entitled to a daily quota.

Nkiru: The commonalities are that it’s a difficult terrain to navigate, but I think also that’s one of the strengths that we have — is that there’s, like, immense opportunities here in terms of finding, you know, you can sort of create magic out of nothing here, because there’s just not a lot to work with. And so, that’s what I find very interesting, you know, here. And I do agree to — I totally am, sort of, a believer in the internet being — I mean, I sort of — I’m not saying it should be a human right. I’m not that, you know, advanced in saying that, but I still think it should be a right — access to it, anyway, because, from my perspective, what the internet affords is access to the world, access to education, you know, the things that maybe government can’t really provide because, you know, government doesn’t have as much money as it should do, and because of mismanagement. But if people had access to the internet, perhaps, I think that it creates, like, a whole opportunity for people to access information that they then use in, you know, daily life, daily business, health, education whatever.

Nanjira: I mean, or maybe we should just have it as a right. I mean, maybe we should dare to dream, because I think the internet and access to it and affordability, as Alan rightfully pointed out, are the two key things we’re looking at. But also the fact that participating in society and economies going forward is actually going to be facilitated primarily by the internet. And so, we need to maybe dream and have a bit of a more lofty goal, other than just accepting from our government that there isn’t enough money. The issue, as you’ve rightfully pointed out, Nkiru, is actually that, you know, we’re not poor. It’s actually just mismanagement. So, it would be actually nice to probably dream for that and actually push for that.

Sonal: I think you guys are completely right. It’s worth actually taking a pause for a moment to reflect on those words, because we tend to throw them around very loosely here in the U.S. — like, access, affordability — but you’re actually talking about true access as a right, affordability as one of the mechanisms for making that happen. And then what you’re really saying, Nanjira, is that there’s an element of inclusion as well — that it’s about getting people the ability to be included in this larger, broader, global movement, where they do have access to this knowledge, information, etc., etc. Well, just to put things in perspective — so, I read a statistic that estimated that the number of smartphones in Africa would be about 930 million by 2019, which is just 3 years away. And that’s basically about one mobile phone per African. How do you guys see that on the ground there? And that, by the way, as we’ve just learned, does not equate with actual internet access. So, what’s the, sort of, differential between having a mobile phone and then the penetration of access to bandwidth?

Nkiru: So, I can generally just talk about smartphones. Oh, actually, smartphones — I think about 100 million on the continent, considering we’re about 1.1, 1.2 billion now. I don’t know. I’m not quite sure, maybe about a billion. The penetration hasn’t been as fast as people have, sort of, forecasted. Well, almost everybody has a mobile phone, but they’re not smart-enabled, so there’s a difference here.

Sonal: Right. I’m glad you’re reminding us to make that distinction.

Nkiru: So, smartphones are not penetrating as fast as we, you know, all the predictions have said. I think a lot of businesses have been predicated on the, you know, the different — what do you call it? The different forecasts about how fast smartphones will penetrate the markets, but they haven’t come as quickly, because they’re really expensive. Of course, with the new — with Androids now becoming much cheaper, we’re seeing penetration is becoming higher, but it’s still not catching up as quickly as we’re hoping.

Sonal: Okay. So, you think one of the reasons things haven’t penetrated as fast as forecasted — and businesses are being built on these assumptions — is because of the fact that the phones are not quite as cheap yet necessarily for everybody.

Nkiru: Exactly.

Nanjira: Yeah. And I think it’s, again, context, you know. Again, just back to the point I made earlier, about the ecosystem and looking beyond the mobile phone as the only tool through which, you know, access should be facilitated. I mean, because — there’s also the question we are asking ourselves now is — in studies that we’re conducting with people at, sort of, the base of the pyramid, or those who are about to be first-time internet users, to better understand their needs — and their understanding of this oncoming device or space that is the internet, just to better understand that. There are needs that go beyond the mobile phone. And so there are those that maybe, you know — we have cyber cafes, for instance, that still exist. And so one, you know, cybercafe could serve maybe 100 people in a locality, and maybe they don’t have mobile phones at home, but they’re still getting access, right? Or, other centers like that. So, there’s a need to start taking all statistics beyond the rush to quantify Africa, as I call it, to take statistics in context and bring them all together to a holistic picture. So, maybe not mobile phones. There may be a plateau, obviously, because of the affordability aspect or that last-mile connectivity element, but what about other ways people are accessing the internet?

Sonal: So, that’s actually really interesting. Do you have some of those statistics? Or, can you give us a little bit more flavor about what that looks like?

Nanjira: Right. So, what we’ve found, and it’s really mostly studies that we do, user experience studies — just better understanding the target user for any mobile phone application or service is that, you know, these are very — and especially in a country like Kenya — and I’ll use Kenya as an example here — is much famed for elements like M-Pesa, but there are other cultural factors that also have been hindrances to start that last-mile connectivity. So, you’ll find, for instance, it would be traditionally that women may not have access to a mobile phone, because of how structures of property ownership exist beyond, you know, or predating technology. And so, how to overcome those? And if that lady cannot own a phone at home, maybe she can access the same services she needs on the internet via a center she can go to during the day, when nobody’s bothering her. We don’t know who’s holding what to your head when you’re accessing a mobile phone.

Sonal: That’s a really good point.

Nanjira: So it’s — this is important statistics. They’re good for estimations. But we also need to bring in the qualitative insights.

Nkiru: The single most instrumental fact on people not owning smartphones is actually the price of smartphones. Well, yes, there are obviously qualitative reasons why people don’t have, you know, different people — but the majority of people don’t own smartphones because they’re expensive. And that’s what’s cool about Android compared to iOS, which is that, you know, you can now get an Android phone for, like, $17. And I’m sure in the next year or two, there will be Android phones for less than $15. And I think the cheaper the phones get, the more people will have them.

Sonal: Alan, do you wanna jump in here and maybe share a perspective on what you’re seeing from a more systemic level, working with governments across Africa? Especially because I think what Nanjira brought up about the last mile is a really, really interesting point. What are some of the obstacles that you’re seeing culturally as well as technically?

Alan: I’ll just give you an example. In a South African township, in a place like Soshanguve or Mamelodi, 70% of people, if they want to apply for a job, they need to get in a taxi. They need to go to the local community center, they look on a board, they look for the jobs posted on the board, they write down the contact details for the job, they get on a taxi, they go to a printing shop, they print their CV, they get in a taxi, they go to the offices of the employer, and they hand in their CV, and then they wait for somebody to get back to them. And that’s, you know, between $4 and $10, kind of, cost and takes you a whole day. And that’s something that, you know, a lot of people in the world haven’t done for 20 years. So just to — you really have to go back to basics. You know, there’s no ubiquity of the internet for most people in Africa. 

We’re now involved with the deployment of the largest public free Wi-Fi network on the continent. And, you know, we find there’s some interesting behavioral stuff. First thing is that there’s not a very big penetration of smart devices. So, Wi-Fi-enabled devices, maybe 50% of households in low-income communities — but there’s a big sharing behavior. So, a lot of people are sharing devices. So, it’s not so much about how many people have a device, it’s about how many households have a device. And if there’s between 5 and 10 people living in a household, you know, then your device penetration goes up quite a lot, because those 5 or 10 people are sharing one device to get on the web. Secondly, a lot of low-income communities who are not accustomed to the internet are still measuring data in minutes.

So, if it takes you five minutes to download a video, they think it’s more expensive when it takes you one minute, even if the video happens to actually be more data, and you’re just on a faster network. So, there’s a lot of education that has to go around there. Old people — basically, anybody over there doesn’t really know what to do with the internet when you give it to them. So, it’s a bit like being an American in 1995. You need to have, like, a landing page from AOL, Yahoo saying, “This is the internet. Click here for, you know, something that’s useful.” So, you can’t give people a clean Google page. It’s absolutely meaningless for most of the communities we deal with. It’s like, “What is this? What are we gonna do with this?” And some of the communities aren’t literate either, which brings me to the content. And the content is — I mean, music videos are just a killer category, which is fun. That’s gonna make lots of money. 

Christianity is a massive — faith-based content, particularly Christian faith, is massive, and European football. I mean, in South Africa, the Premier Soccer League is quite big. But just generally, English Premier League and European football, particularly clubs like Chelsea with lots of African players, it’s a massive content category. You don’t understand how many people are following that kind of stuff. And we see in our networks that Blackberry is a massive chunk of the market. I mean, I know in America, Blackberry is dead. You know, people laugh about Blackberry, but around here, it’s still pretty big, and Huawei — Huawei Android devices are pretty big. So, I think it’s a little bit different from what’s happening in the States.

Sonal: That’s actually super helpful for fleshing out some of the statistics we’re talking about. Together you guys are, sort of, sharing both different ways of putting those numbers that we typically hear in context. I think what you pointed out about measuring things in terms of people versus devices in the household is really important — to talk about that penetration rate, and even the misconceptions people have around minutes versus data size is also really interesting. And then we also talk about some of the more qualitative behaviors that, you know, go with the power structures that you mentioned, Nanjira — where it’s, sort of, like, about, like, you know, is the woman the one who has a device? And she may be in the same household with, like, 10 other adults, but is her power and access equal to, say, the men in the house? I mean, it sounds like there’s a lot of different things to think about.

Alan, I think it’s also really interesting that you talk about entertainment-based content being very popular, because you would almost think that with this being, sort of, this utility model that we’re all sort of arguing for, it’s interesting that the things that people find most popular are actually still down to entertainment, music, sports. For some people, they would consider faith an entertainment. I mean, it’s sort of, like, how people pass their free time in a lot of different regions. That’s actually kind of counterintuitive to me. And how does that play out with your observations, Nkiru, given that you run a music company?

Africa’s growing middle class

Nkiru: It’s a big continent, and there’s definitely a lot of poverty. But in saying that, there’s a growing middle class. And, you know, you can’t solve Africa’s problem or the world’s problem in one, you know, fell swoop. So, is the growing middle class big enough for us to, sort of, like, look at? I think so. I think that, you know, the whole picture of, you know — when you see on CNN when there’s, you know, black kids swatting flies, and sort of, that’s not — I mean, yes, there’s poverty, but there’s great stuff happening here as well. So, entertainment is huge because people are — people have, you know, more income. Well, maybe not in Nigeria. Right now, we have a recession. But people have more income and they want us — you know, they want to spend money, you know, doing chill, cool things. And, for example, going back to mobile phones, they’re aspirational. 

So, you know, in the way that you want to, sort of, move to Lagos, because you’re in the town — so, you’re in the rural area and you want to move to Lagos. Things are aspirational here. People want to be cool. People want to associate. Being Nigerian is cool now. Nigerian music is cool compared to American music. You know, those kinds of things are things that we consider. So, there is a huge growing middle class, we think, and that’s why you see that. Well, some people are poor, and some people are poorer, and some people are less poor, and some people are really rich. And there’s I think — there’s a spectrum of people that are all-African, and we can, sort of, like, look at different perspectives rather than just focusing on all the bad stuff that goes on here, as everywhere else.

Sonal: I think it’s interesting actually because you guys are painting a whole range, which is sort of the fact that the mobile phone can be everything from utility — which we’ve heard that common statistic, that more people have mobile phones than they do toilets, in a lot of developing countries. And then at the other end of the spectrum, you have this notion of the mobile phone as a very aspirational device, ecosystem, and all these things that come with it. I think the point you’re drawing, about focusing on the growing middle class, is incredibly important, because that is on a continuum. I mean, we can’t be focusing only on one extreme, the very rich or the very poor either. It’s certainly true in thinking about this.

Nanjira: Yeah. I’d say another angle is also — let’s even imagine a utopia where every African has a mobile phone. Are they actually able to create? Are they able to code on that mobile phone and contribute to another <crosstalk>, or they’re just going to be consumers? And this is where we need to start asking these questions around what access to the internet, and how we’re facilitating — so that this fixation on a particular device does not, rather, inhibit us from seeing a bigger picture on how people could actually contribute and benefit from the global economy that’s also gonna be a very digital one.

Sonal: There’s this famous quote about Steve Jobs. I don’t know if it’s true or not, if it’s just anecdote — saying that, you know, getting this criticism that the iPad was only intended to be, you know, for consumption, and him feeling very, you know, affected by that and trying to, you know, work on that, and clearly this huge ecosystem of apps have grown up around — since the launch of the iPad, where people can actually produce versus just consume on mobile devices. That said, we tend to take a very academic approach, I think, to that debate. It’s almost like frosting on the cake, versus what you’re describing is actually, again, incredibly important to the notion of inclusion, and being included in this larger global movement. Like, can you create code? Can you create art? Can you not just be a consumer who’s taking yet another technology from elsewhere, or even locally, and just providing money to it, versus actually being able to do something with it?

Nkiru: So, when Nanjira is talking about, “Oh, can you code on your mobile phone? Can you do this?” I’m thinking, “I don’t even have water.” I need to have generally — I need to have, you know, like, lights. You know, I need to have power 24/7. So, it’s really, sort of, like — I don’t then say, “Oh, there’s no power here.” Because we have to produce our own power. Like, at Spinlet here, we really have to run a generator 24/7. You know what I mean? So, these are crazy things that happen here, but yet, you know, we’re still thinking about, you know, how we wanna code. So, I think it’s just an interesting place to be, where you can do both.

Sonal: Yeah. No, I think it’s fascinating. It’s, like, collapsing Maslow’s hierarchy of needs. It’s sort of, like, saying — there’s, like, this funny diagram that periodically makes rounds on the internet where it’s, like, Maslow’s hierarchy of needs — where you have basic things for survival and at the very bottom, it’s Wi-Fi, exactly.

Nkiru: Yeah. I mean, I must say, it’s really interesting to see how even in development considerations and development agendas, what we’re starting to see is whether it’s a zero-sum game. So, do we invest in water or internet? It’s starting to become a very short-sighted question. You know, does it mean, then, we should not focus — should we get everyone access to water so that we then get people internet? So the hierarchy of needs is being renegotiated, even from a state perspective or a donor perspective, you know. And each, you know, there’s so many nuances to each aspect, because either way, even if we waited and divested money into bringing in water first, we’re still going to be missing out for those who already have access or those who could have access. So, it’s a really interesting development question.

Alan: So, the main thing is governance. You know, if you’re living in South Africa or Nigeria or Kenya or Zimbabwe, or any part of Africa — actually any part of the world, come to think of it — we’re all kind of complaining about the same thing. You know, our leaders aren’t doing a good job, etc. So, when it comes to votes, and deciding who gets them to power, we’re never gonna have better leadership if people don’t know what’s going on.

Sonal: Oh, that’s a good point. So, like, sort of knowledge as a way to, sort of, break down some of that discussion.

Alan: Yeah.

The popularity of messaging

Sonal: I wanna revisit a theme that you guys brought up earlier, which is, you know — you mentioned, Alan, that people get a little taken aback when they see a clean Google page, because they, sort of, need some instructions on how to navigate for first-time internet users what to do next. One of the interesting insights that I heard on our podcast about China and India is how people actually get taken aback with a very clean, uncluttered page, because they’re so used to having limited bandwidth that they want high information density. Like, they’d much rather have a design where there’s just a bunch of links for the sake of being parsimonious about that use of bandwidth. What are some of the interesting things happening on the design side of all these apps and things that you guys are seeing and using in your various regions, or that you’re working on?

Alan: You know, we deal with some U.S. companies that try to come into, you know, to Africa. I find that maybe some “first world” kind of companies underestimate the consumer. You know, so they think something pretty B-grade, they can get away with it. But, you know, the internet is here. People can look around and see for themselves what’s good and what’s not. So, provided it’s not consuming too much data, I think user experience — consumers seem to be demanding as good of user experience as anywhere outside of Africa. But you just have to make it mobile-centric. You just have to make it mobile-centric. You know, there’s just — it’s absolutely irrelevant whether it looks nice on a laptop, you know, it’s just not relevant here.

Nanjira: For us, one way we try to understand that — and also not homogenize users, end-users — is encourage, especially startups that we house here and other clients really, to consider user experience designing there, you know, in their iterations of either applications, mobile websites, or any offering they’re giving to the internet. Because, yeah, it’s usually assumed, you know, there’s this standard typical user who just needs to get this and that and that, but they’re very — they’re different nuances. So, if it’s a woman, again, who’s trying to use a mobile health app, there are gonna be things that she sees to have value, maybe at the landing page — just seeing all the different ways they could go to one place, other than just landing on a single page, as you mentioned. So, there’s so many nuances.

Nkiru: You know, everyone’s talking about mobile phones and how people are using mobile phones, but we found that — so, we’ve had to recently redo our browser, you know, website, because we found that people who have advanced-feature phones actually, you know, prefer to sort of, like — or can only sort of access us via browser, as opposed to downloading an app. And, in any case, when you download an app, it sucks up your data. So that’s really — you know, it’s a very curious thing, where we are still doing a lot of browser work as opposed to just concentrating on apps, you know — going the app way because of the mobile phone filtration. So, we find that the browser is equally important, if not more important, because of the fact that data is so expensive. So, when you’re accessing the internet via a browser, you don’t have to download an app that could, sort of, chop up all your data.

Alan: Yeah.

Sonal: It also puts to rest some of the academic debates people have about whether mobile-centric isn’t just about, like, being on the mobile phone. You’re saying it’s down to the nuance of whether you’re designing for a browser versus the app itself.

Nkiru: Exactly, yeah. Just design — mobile-friendly designs — as opposed to, you know, when, you know — the whole new concentration on apps. Like, they’re expensive to run. They chop up all your data, and then that means that you can’t actually — data is one of the biggest problems here. It’s expensive. So, people are counting every second they spend, you know, on the internet.

Nanjira: And apps are not that popular, really, actually. A lot of studies we’ve done especially for, you know, applications for governance issues. They’re not popular. People are happy to either get their information, again, from the popular app. So if you go to Facebook, you’ll probably have links to this page that also has information, and that kind of thing. So, yeah, before you invest in an app, you have to really check whether it’s actually the appropriate technology to use for advancing your agenda.

Sonal: Right. I would actually say, anecdotally, that’s also very true in India. I just came back from a trip a couple of weeks ago, and I noticed how very few apps anyone around me had on their phones. I mean, what I am curious about is whether you see people doing a lot more stuff over messaging, the way we’ve talked about, like, what’s happening with WeChat in China, or WhatsApp in India, and elsewhere. Do you have any perspectives on where messaging plays out into all this?

Nkiru: I mean, WhatsApp is popular everywhere.

Nanjira: WhatsApp is huge here, but, you know, it’s a millennial thing, isn’t it?

Nkiru: Yeah.

Nanjira: You know, millennials react the same way everywhere in the world, and so WhatsApp is really huge, too.

Sonal: Oh, is it really a millennial thing? I was gonna say, because — I was gonna say, in India what shocked me when I went a few couple of years ago was that grandmothers and aunts were using WhatsApp, and that’s what was surprising to me.

Nanjira: This is true. But then you’re, like, just generally all the apps. So there’s Snapchat now. There’s all kinds of new things that I can’t keep up with, and I keep getting told about them from, you know, my team. So WhatsApp is huge, huge, but other things are becoming equally huge. I think it’s a trending — you know, like, the trend — something is in, and then something is out, and then they like, “Have you heard of the new thing coming in?” So I think it’s general — worldwide things that are trending everywhere. But WhatsApp is huge because you can make free calls from WhatsApp, if you have internet, of course, but then that means that you’re then — if you’re, I think, maybe middle class where you have internet access, and then you can then make cheaper calls from WhatsApp compared to using the usual Telco line.

Nkiru: You’re mentioning very U.S.-centric apps, and I actually wanted to direct this to Alan to talk a bit about Mxit and its heyday.

Alan: Well, so, you know, I have the questionable honor of residing over probably the biggest tech failure in Africa.

Nanjira: Oh, no.

Nkiru: It was a success at some point.

Nanjira: This isn’t what we want.

Alan: But at the same time, it was probably the biggest <inaudible>, so there were some good things out of it.

Sonal: What were some of the learnings?

Alan: Mxit’s success came from the fact that it made it really, really cheap for people to text one another. That was it. So SMS or texting was — it was too expensive, and Mxit was first moving and built this massive network effect. So, it’s a killer app. You know, one of the lessons is, never ever go up against American companies. You’re dead. Silicon Valley, they’re gonna crush you. So, you know, as soon as the smartphone, kind of, wave started breaking, you know, WhatsApp just won the race. It has won the race. It will win the race. I like WeChat. I mean, I don’t think — likes for WeChat aren’t to be underestimated — but just aspirationally, in our markets, we can see that WhatsApp is winning. It’s at a youth level. It’s at a middle-age level, and it’s [at] an old-people level. So, it doesn’t matter what it is. And I’m quite excited, actually, to see WhatsApp opening itself up to be a bit of a platform, because if you can plug that community into whatever application you’ve got, you know, boom, you don’t have to reinvent the wheel. And everyone’s trying to reinvent the instant messaging wheel. So, I think that that ship has left, and WhatsApp has won the race.

Sonal: Well, it’s interesting that you talk about it opening itself up into a platform, because that’s exactly right to see what comes next. But what I think is really fascinating is the use cases where I’ve seen abroad, where people are actually using it less for messaging, in a typical way, and, for example, they’re using it as a news source. They’re using it to vote in elections, in certain places, or to, like, do certain things. I’ve seen relatives use it exactly the way you would use a social network, but instead of using an actual social network, they’re just using messaging to do all those things. Like, share status updates, share photos, etc. How is the use of messaging — I guess I’m trying to get a little bit more understanding of how people are using some of the messaging apps in Africa in this context, and anywhere in Africa — not just all over — but in this context of messaging as a broader trend.

Alan: In South Africa, we have a massive crime problem. So, security is a big concern for a lot of people in South Africa, and WhatsApp has become the de facto means of communities organizing against crime.

Sonal: Oh, wow. 

Alan: So, streets all have, you know — the block, or street, or whatever street has got, like, a WhatsApp group. Everyone participates in it, and anything that happens or any suspicious vehicles or anything, you know, people are always — the feedback group is WhatsApp. And it’s, like, far and away the most powerful tool for something like that. And the way I see this, kind of, evolving around here is, one of the new services we’re helping the government deploy is the equivalent of Uber, but for the police. So, if you see a crime in — especially in a community like Mamelodi or one of the townships, the street names aren’t very obvious. The addresses aren’t very obvious. But if you can use GPS to pin where you are, you can report it to your nearest police cars, and they can, kind of, respond without having to go through call centers and all the translations in between. You’ve solved a massive problem.

Sonal: Wow.

Alan: But chat is a massive component of that, you know. And you need the officers to be able to communicate with the citizen and vice versa, and just, kind of — just the more feedback there is, the faster people can respond and kind of get to the incident. And that’s where — I mean, chat is integral to everything that we do, that we see in the app space, and WhatsApp will — just, you know, if you can plug WhatsApp into it, you don’t have to reinvent the wheel.

Nanjira: Perhaps another way of looking at WhatsApp and, you know, similar chats, like WhatsApp is that it’s come — for me, it’s become so everyday that I don’t even think about it. You know, it’s how people interact, it’s how you get even, you know, work conferences on WhatsApp. Everyone, as you say, uses WhatsApp, but what I wanted to sort of, like, maybe, sort of, like, take us to is — when Alan talked about when WhatsApp came to the market, and what it did to Mxit — and I thought that that’s interesting — when the Americans come in. So, for example, Apple Music, Netflix is coming in. Netflix is already here. Apple Music is here. And so, what’s that gonna do to the, you know, existing African SMEs in the market? I think that would be interesting for me to, sort of, hear everybody’s perspective as well.

The importance of local advantage

Sonal: I guess what I really wanna get at here, as well, to build on your question is, is there a local advantage, or does it not apply when a global player comes in?

Alan: Do not ever take on America. So, that’s the golden rule. But there’s lots of things that in Silicon Valley, you know, people — some guy sitting in a cappuccino — having his cappuccino in San Francisco is just not gonna think about. You know, if you’ve got a problem that’s unique to Africa, and you’re trying to tackle that problem — and it must be unique to, not necessarily to Africa, but just not at all obvious in Silicon Valley — then you’ve got a shot. But, of course, if you get any kind of traction in any of the spaces, you’ve got to make sure that someone in the States ain’t gonna, like, just see your idea, copy it, and just come take you on. And the real capital advantage other than the, you know, the hugely talented, aggressive, you know, hard-working people. Just the capital advantage is so overwhelming. 

So, in my opinion, you know, we invest in businesses only if there’s a network effect, and if someone’s already dominated the market. So, something like e-commerce — you know, Jack Ma, or Baidu, or Tencent — they all benefited from, you know, kind of, closed markets to the U.S., where they could build up the network effects for e-commerce or for social. And by the time they, kind of, opened it up to foreign competition, you know, it’s kind of — the land grab had happened, and you’ve got this massive moat protecting you from Silicon Valley. And there are some things like that in South Africa, and I know in some places, the rest of Africa — but there’s not a lot, and you really have to have this buyer-seller or friend-friend network in place before. Or know that you can get it before anyone in America gets wind of it.

Sonal: Alan clearly has a very specific perspective. And do you guys agree, disagree?

Nkiru: No, I think I generally agree but, of course, there’s — I mean, local advantage is, I mean, important and necessary, but the thing about local advantage is that local advantage can be bought. So, for example. So, I train my staff, and clearly I’ve spent so much time training them. And then, you know, whoever — the American person is coming in, and they have a lot more money, like, than I do. They have the capacity. They have tech capacity. They have, you know, capital capacity. And it doesn’t take, you know, a huge, you know — what do you call it, increase — to just make them jump. Am I angry about that? Maybe, when I’m really sort of, like, “Oh, my God. I spent all this time training these people.” But then, I also don’t wanna have people working at Spinlet, for example, where nobody wants to hire them. So, it’s like, you know, what came first, catch-22. But I think, you know, an Apple or an Amazon or — I think local advantage is very important, and I’ve talked about, you know, you can’t be an Africa expert just by spending two weeks here. But also you can’t compete with the American power of, you know, capital and, you know, tech power as well, so that’s how I see it. So, I’m with Alan on this one.

Nanjira: I’ll give maybe two examples of where local advantage is giving a quick win for some startups here. One is very well-known. That was Ushahidi. I mean, it came up from a very contextual situation and occurrence here, but that was a specific use case. But for them to survive, they’ve also had to, you know, to embrace the Americans. So, Ushahidi has a team that works from the U.S., a team that works from Kenya, and a team that works from other countries — and that’s how they’ve been able to co-opt that expertise that’s ready in the market in the U.S., for instance, so that they’re not necessarily bought out. Another example that’s starting to come up now is the BRCK that’s being built as appropriate technology for facilitating Wi-Fi router access, right? The idea being that, you know, while I may buy a router for my house in the urban area in Nairobi, it may not necessarily be the most appropriate technology for beaming Wi-Fi access when I’m out in my grandmother’s village. You know, so it needs to be something rugged, something that could withstand, you know, falling and dropping and that kind of thing. Now, they have that local advantage by understanding that, but it also becomes a game of wits, in terms of — once you start gaining traction, as was mentioned, once you start gaining attention, and other people start to see what you’re on to, how do you make sure they don’t beat you to market in something that’s faster, cheaper — and there’s the second-to-market advantage. So, it becomes a game of smarts here, and it’s one we’re learning by — you know, it’s baptism by fire, if ever there was a case.

Sonal: Right. Well, it’s fascinating, too, because what you’re describing is competition everywhere, but there’s also a factor here where there’s a power differential between, as you mentioned, access to capital, access to — you know, it could be like Chinese investors coming into India, or Jeff Bezos going to India with Amazon, or it could be — right. I mean, no one can compete with that, sort of, deep pockets at a certain level. But at the same time, we do see success cases where they compete in different ways.

Nkiru: The one thing I can say about a local advantage, where it really is an advantage, is when, like, an international is coming in to — you know, is trying to establish footprint on the continent, and the general mistake that, you know, internationals seem to make is that they assume that by hiring someone out of, you know, business school from somewhere — I don’t know — they have — you know, they can do it. And that’s where you need the local — that’s where the local advantage will always, you know, sort of overpower the — you know, the — because you cannot come to, you know, Lagos, if you’ve just, you know, been here one week, and you cannot go to Kenya, and, you know, set up a company without having, you know, the local insight, the local, you know, local knowledge. Just, sort, of how to navigate the terrain. And so, that’s where local advantage would always, you know, beat, you know, the clout and the power. And so, I think it’s just finding a way to do things in collaboration.

Doing business in Africa

Sonal: Reflecting on this notion of competing with business around the world, and locally, any final reflections on things that people should know who are trying to build businesses in Africa?

Alan: Yeah, I once read a blog that Silicon Valley, you know, gives you 16 metrics to chase for a startup. And a lot of people in South Africa and Africa, and the rest of the world, read those blogs and think, “That’s the holy grail.” But there’s only one metric, and that’s cashflow. Advertising is not a business model. “One day we’ll sell to Google” is not a business model. You really need to be, kind of, thinking like a traditional business — like, how are you gonna make money pretty soon. And if you’re not, well, then you’re really up against some guy who’s funded out of Silicon Valley.

Sonal: So, one final question that’s come up a few times is — thoughts on, you know, technology and how women are using it at all ages, in various regions of Africa. You’ve brought up the notion of women doing things a little differently a couple of times. I wanted to make sure you guys could share some thoughts on that.

Nanjira: Sure. You know, the World Development Report by the World Bank just came out and reconfirmed the notion that, yes — much as there are many women who are participating online, the numbers still lag behind in comparison to men. And that could be seen, you know, as an indicator of existing inequality, so to speak, but the internet and access to it — for those who do get access to it, and who are able to bypass certain barriers that are socially, culturally, you know, income-wise, who get to participate — we are seeing such opportunities to raise agency. To have agency to, you know, speak up about issues, to address and to really mainstream the idea of gender in very various considerations.

So, for instance, you know, it’s no longer sufficient to just have numbers of women on a panel. It also goes now to, “We will criticize whether they actually get to be asked questions based on their expertise, and not just because they’re filling a diversity quota.” And so, you’re going to see these kinds of conversations taking place online. There’s so many campaigns we’re seeing that are helping mainstream the idea — taking on, you know, I would say misogynists who are both men and women. We’re seeing fantastic campaigns coming up about that. Sensitization on these issues in a way that’s not just preachy and traditional NGO-y, where it’s just driven by citizens who are engaging on platforms that facilitate free expression. So this is fantastic. That said, there’s still women who are keeping off because they’re attacked, and there’s a lot of digital safety considerations to have. And this, you know, it’s not to make women seem, like, some little eggs in a sense, it’s just — these are just indicators of existing patriarchal structures and so it’s really — the internet and access to it is really one of those things that could either perpetuate this patriarchy, or start deconstructing it as we all get connected.

Sonal: Or probably, more likely both, actually.

Nanjira: Yeah, or both. I mean, society does exist in contradictions. And in fact, as a parting shot, I also wanted to add that there’s a fact that we need to have more research supported to really unearth these qualitative insights, that then complement these numbers that we’re getting. I mean, I don’t know whether to be offended or amused by the statistic, you know, that they’re more mobile phones than toilets. I get that statistic, but then I’m, like, “To what end are we — what are we saying?” You know?

Sonal: Right. What does that mean exactly? Right. That’s a good point. We use it actually. I’m gonna confess, we use that statistic ourselves. I’ve linked to that or quoted that in various pieces, and you’re right to call BS on it a little bit. I mean, not BS necessarily, but to say, like, “What does that even mean?”

Nanjira: What does that actually mean?

Nkiru: That is actually BS.

Nanjira: It’s a statistic to be used <crosstalk> for a U.S. state, so there are more mobile phones than there are toilets — does that apply across the metric, if you’re using it as a comparative? Or is it just for Africa and the developing world? So, there’s a need for more nuanced research to strengthen and, sort of, flesh out those numbers being thrown out. So, let’s not just quantify Africa. Let’s bring other qualitative insights that bring out its diversity, and also make for wise investment decisions, actually.

Nkiru: I wanted to talk about like, you know, the idea of women in technology and how it’s been niggling at me for a while now. And I think that the phrase, in itself — instead of creating or, sort of, incubating, you know, a whole group of women who are get — then gonna have more low self-esteem, because or inferiority complex, because we then assume that you have to be great at math to do — to work in a tech company, for example. And, you know, and I feel like when we, sort of, create these things that will help women — or we need to have women do more math — these are, like, 9-year-olds we’re talking about, and 12-year-olds, and 14-year-olds. There’s a whole group of 25-year-olds and 22-year-olds who are long out of school, who don’t need to be told that. They just need to be told that they need to excel in whatever they’re doing, because they can do anything. And so, I’m sort of worried about the terminology that’s suddenly creeping up in the past two years about, oh, how women need to be encouraged to be, you know, in STEM and all that stuff. So, I hope that we can talk about these things more frequently and sort of, like, put light to them.

Sonal: Well, I actually agree with you personally, as well. I think that that discourse is important, but also troubling when it puts this hierarchy of skills that’s, sort of, not honoring some of the more nuanced qualitative sciences as well. Which is a meta-theme in this conversation as well. Alan, Nkiru, and Nanjira, I just wanna say thank you again for taking the time. This has been an incredible conversation. We’re gonna continue it, I’m sure, in the near future, and the first of many, and thank you.

Nanjira: Thank you.

Nkiru: Fabulous. Thank you so much.

Alan: Thanks very much.

• 

  Nkiru Balonwu
• 

  Alan Knott-Craig
• 

  Nanjira Sambuli
• 

  Sonal Chokshi is the editor in chief as well as podcast network showrunner. Prior to joining a16z 2014 to build the editorial operation, Sonal was a senior editor at WIRED, and before that in content at Xerox PARC.

  Follow Twitter

What is AI or artificial intelligence but the ‘space of possible minds’, argues Murray Shanahan, scientific advisor on the movie Ex Machina and Professor of Cognitive Robotics at Imperial College London.

In this special episode of the a16z Podcast brought to you on the ground from London, Shanahan — along with journalist-turned-entrepreneur Azeem Azhar (who also curates The Exponential View newsletter on AI and more) and The Economist Deputy Editor Tom Standage (the author of several tech history books) — we discuss the past, present, and future of AI … as well as how it fits (or doesn’t fit) with machine learning and deep learning.

But where are we now in the AI evolution? What players do we think will lead, if not win, the current race? And how should we think about issues such as ethics and automation of jobs without descending into obvious extremes? All this and more, including a surprise easter egg in Ex Machina shared by Shanahan, whose work influenced the movie.

Show Notes

• Distinguishing various type of AI and how they learn [0:57]
• Ethical concerns [11:13], and a discussion of how AI may develop going forward [22:02]
• Will academia or business succeed in developing AI [30:15], and how might it affect jobs? [34:51]
• Will AI ever become conscious? [36:18]

Transcript

Sonal: Hi, everyone. Welcome to the “a16z Podcast,” I’m Sonal. And, today, we have another episode of the “a16z Podcast” on the road, a special edition coming from the UK. We’re in the heart of London right now. I’m here with Murray Shanahan, who is a professor of cognitive robotics at Imperial College in London, and he also consulted on the movie “Ex Machina.” And so, if you didn’t like the way that movie turned out — we don’t wanna put any spoiler alerts — you can blame him. And then I’m here with Tom Standage, who’s the deputy editor at “The Economist” and also the author of a few books.

Tom: Oh, yeah, six books. The most recent one was “Writing on the Wall,” which was a history of social media going back to the Romans, and probably the best-known one in this context is “The Victorian Internet,” which is about telegraph networks in the 19th century being like the internet.

Sonal: That’s great. And I’m here with Azeem Azhar, who publishes an incredibly interesting and compelling newsletter that I’m subscribed to — the “Exponential View.” He used to be at “The Guardian” and “The Economist,” and then most recently founded and sold a company that used machine learning heavily. So, welcome, everyone.

Murray: Thank you.

Azeem: Hello.

Types of AI and learning techniques

Sonal: So, today, we’re gonna talk about a very grandiose theme, which is AI — artificial intelligence — and just, sort of, its impact and movements. This is really meant to be a conversation between the three of you guys, but, Murray, just to kick things off — like, you consulted in the movie “Ex Machina.” Like, what was that like?

Murray: Oh, it was tremendous fun, actually. So, I got an email out of the blue from Alex Garland, famous author — so, that was very exciting to get this email. And the email said, “Oh, my name’s Alex Garland. I’ve written a few books and stuff. And I read your book on consciousness, ‘Embodiment and the Inner Life,’ and I’m working on a film about artificial intelligence and consciousness. And would you like to, kind of, get together and talk about it?” So, of course I jumped at the chance, and we met and had lunch. And I read through the script, and he wanted a bit of feedback on the script, as well — where it hung together from the standpoint of somebody working in the field. And then we met up several times while the movie was being filmed, and I have a little Easter egg in the film.

Sonal: Oh, you do? What was your Easter egg? I’ve seen that movie three times in the theater, so I will remember it, I bet.

Murray: Oh, fantastic. So, there’s a point in the film where Caleb is typing into a screen to try and crack the security, and then some code flashes up on the screen at that point, and that code was actually written by me.

Sonal: Oh, yay.

Tom: So, it’s real code, it’s not the usual rubbish code.

Murray: And it just, sort of, flashes up. But what it actually does, if you actually type it into a Python interpreter, it will print out “ISBN equals,” and the ISBN of my book.

Sonal: Oh, that’s so great.

Tom: Oh, it’s Python as well? I’m even more thrilled.

Sonal: I think it’s fascinating that you say that the part that you didn’t go into detail about is the name of the second part of the title of your book, embodied consciousness?

Murray: Yeah, there’s a long subtitle which is “Cognition and Consciousness in the Space of Possible Minds.” Now, I very much like that phrase, the space of possible minds. I think if you were to kind of pin me down on what I think is my most fundamental, deepest interest, it’s this idea that what constitutes possible minds is much larger than just humans, or even the animals that we find on this earth, but also encompasses the AI that we might create in the future, either robots or disembodied AI.

Tom: So, it’s a Hamiltonian space of possible minds? That’s beautiful.

Murray: Yeah, a huge kind of space of possibilities.

Azeem: I mean, it’s a really interesting idea, and it’s something that comes across in a couple of your other books as well, which is this notion that we think of intelligence as — quite often the artificial intelligence — that plastic white mask that you see on the cover of many, many a film or book cover. But, of course, as we start to develop these new AI systems, they might take very, very different shape. They may be embodied in different ways, or they may be networked intelligence. So, one of the areas I think is interesting is what’s happening with Tesla, and the Tesla cars that learn from the road — but they all learn from each other. Now, where is that intelligence located, and what would it look like, and where will it sit in your space of possible minds?

Murray: Yeah. Absolutely. It’s a completely distributed intelligence, and it’s not embodied in quite the sense of — of course, a car is a kind of robot, in a way, if it’s a self-driving car, but it’s not really an embodied intelligence. It’s sort of disseminated or distributed throughout the internet, and it’s a kind of presence. So, I can imagine that within the future, rather than the AI necessarily being the stereotype of a robot standing in front of us, it’s going to be something that, sort of, is hidden away on the internet and is a kind of ambient presence that goes with us wherever we go.

Tom: Well, that’s another sci-fi stereotype there, isn’t it? That’s the UNIVAC or the Star Trek computer. But, as I understand it, your work starts with the presumption that embodiment is a crucial aspect of understanding intelligence, which is why you’re interested in both the robotic side and the intelligence side.

Murray: So, certainly, I would’ve taken a stance, you know — if you’d asked me 10 years ago — that cognition and intelligence is inherently embodied. Because what our brains are really for is to help us get around in this world of three-dimensional space and complex objects that move around in that three-dimensional space, and that everything else about our intelligence — our language, our problem-solving ability — is built on top of that. Now, I’m not totally sure that it’s not possible to build AI that is kind of disembodied. Maybe — in my latest book, I use the phrase vicarious embodiment. Or, I should say vicarious embodiment, for a U.S. audience.

Tom: So, it can kind of embody itself temporarily in a thing and then go somewhere else?

Murray: Oh, well, that’s another thing, you can have sort of avatars. But what I mean by vicarious embodiment is that it uses the embodiment of others to gather data. For example, the enormous repository of videos there are on the internet. There are zillions of videos of people picking up objects, and putting things down, or moving around in the world. And so, potentially, it can learn in that vicarious way everything that it needed to learn by being actually embodied in itself.

Tom: And this goes right back to the neurological basis, I think, of some of your — because you started off doing symbolic AI and then moved over as, kind of, the whole field has, to more of this neurological approach.

Sonal: And by neurological approach, Tom, you mean more like in a deep learning sense?

Tom: Well, exactly. As I understand it, part of your approach there was the idea that the brain itself can rehearse motor-neural, sort of, combinations, and that’s how we, kind of, predict how the world will behave. We kind of say, “What would happen if I did this,” which is very much like what the DeepMind AI is doing when it plays Breakout or whatever — those, kind of, Deep Q networks, which is all about feedback based on predicted actions and remembering how things worked out in the past.

Murray: Certainly. I’ve always thought that this idea of inner rehearsal is very, very important — our ability to imagine different possibilities and…

Tom: So, watching YouTube videos of people doing things can function as inner rehearsal, I think.

Murray: Or, if you have a system that can learn from that, the sort of dynamics of the world and the statistics of actions and their effects and so on — then it can use that — so, it sort of builds a model of how the world works, and then it can use that model to construct imaginary scenarios and rehearse imaginary scenarios. Actually, just going back very quickly to the DeepMind, DQN. So, in fact, for the bit of work that they actually published — I think one of its shortcomings, actually, is that, in fact, although it has done all of that learning about what the right actions are doing in the right circumstance is, it doesn’t actually do inner rehearsal. It doesn’t actually work through scenarios. It just…

Tom: Oh, it’s just remembering how things worked out in the past?

Murray: Yeah.

Sonal: Murray, actually, what exactly then is inner rehearsal? Because I think I’m actually confused. We’re describing three different things. There’s sort of a predictive aspect, there’s sort of this decision-making framework, and then there’s also, sort of, something that reacts to the world in a dynamic environment that’s constantly changing, and reacting to that information in a very proactive and intentional way. Those are all different qualities. So, what is inner rehearsal exactly?

Murray: So, I think that the, sort of, architecture of intelligence is putting all of those aspects together, really. So, inner rehearsal is when we close our eyes — of course, we’re not really necessarily gonna close our eyes, especially if we’re on the underground and the — but it’s when we close our eyes and imagine going through some particular scenario. Imagine doing an action and inwardly realizing that this will be a good, or this would have a bad outcome.

Sonal: It’s like a planning scenario. In model-based reasoning, it’s like planning…

Murray: Sort of planning, yeah. It’s model-based reasoning, yeah.

Tom: Some of the same bits of our brains light up. And if I imagine punching you, then, actually, the parts of my brain that will be involved in punching you are partly — fortunately, they’re not actually…

Sonal: Punching you. They’re just envisioning that similar — right.

Tom: But the point is there’s more to it than just, sort of, thinking of the scenario. In some sense, the brain does rehearse the scenario in other ways, doesn’t it?

Murray: Yeah. There’s quite a bit of evidence. The way the brain does it, as you say, is to actually use the very same bits of neurological apparatus that it uses to do things for real.

Sonal: So, the planning is almost interchangeable?

Murray: Yeah, it’s just kind of turning off the output and the input.

Tom: And I remember seeing a video of a cat, and the cat’s acting out its dreams because it had some part of its brain basically modified, so that the part that normally suppresses the intention to act things out that you’re rehearsing actually was taken away. And so, the cat was imagining swiping mice and this sort of thing while being asleep.

Sonal: So, that’s actually kind of fascinating, because it’s the reversal of how I’ve always thought of the human brain, which is — you’re basically saying, almost, that there’s always a bunch of scenarios and actions that can play out at any given moment in the brain, and that we’re actually already acting on, in essence, by the neurological impulses that are being fired in the brain. But in reality, what’s holding it back is some kind of control, that’s stopping something from happening, as opposed to saying, “I’m gonna do X, Y, or Z,” and then acting on something intentionally. So, it’s more of a negative space thing than a positive thing.

Murray: Yeah, or a kind of veto mechanism.

Sonal: Right.

Murray: Yeah. Oh, in fact, I think you’ve actually proposed [that] there two very good rival hypotheses for what’s going on. And I wouldn’t want to venture what I think is the answer there, and it’s the kind of thing that neuroscientists study.

Azeem: But it doesn’t feel like current AI — certainly, the stuff that’s implemented commercially, or even that’s published at a research level — is really bridging into this area that we’re talking about, these rehearsal mechanisms, for example.

Murray: Yeah. I think it’s actually one of the, potentially, hot topics to incorporate into machine learning in the not-too-distant future. So, one of the fundamental techniques in DeepMind’s work is reinforcement learning.

Sonal: Which is also very popular in developmental psychology.

Murray: It has its origins, really, in things like classical conditioning.

Sonal: That’s right, Pavlovian classic, bells, signals.

Murray: So, within the field of reinforcement learning, there’s a whole little subfield called model-based reinforcement learning, which is all about trying to do it by building a model, which you can then potentially use in this rehearsal sort of way. But although Rich Sutton, who is the sort of father of reinforcement learning — in his book, way back in 1997, he proposed his architectures in which these things are blended together very nicely. But I don’t think anybody’s really built that in a very satisfactory way quite yet.

Ethical concerns

Sonal: So, just to help us come along with this — concretely, where are we right now in this evolution? And there’s schools of thoughts that can disagree with this, but just to simplify things — machine learning, deep learning as a deeper evolution of machine learning, and then sort of like a full AI on a continuum. Is that sort of a fair way to start looking at it? And where do we kind of stand on that continuum?

Azeem: So, I have a model which says that, you know, AI and machine learning are really quite distinct things. You know, AI is all about building systems that can, in some way, replicate human intelligence or explore the spaces of possible minds, in Murray’s phrase — whereas machine learning is a very specific technique about building a system that can make predictions and learn from the data itself. So, there are AI efforts that have no machine learning in them. I mean, COIC, C-O-I-C, is a great example. You know, you try to catalog all of the knowledge in the world. I think, you know, it’s the mindset of the market to combine the two, because it might give something more attention.

Murray: Yeah. I mean, I very much agree with that. I see machine learning as a kind of subfield of artificial intelligence, and it’s a subfield that’s had a tremendous amount of success in recent years, and is gonna go very, very far. But, ultimately, the machine learning components have to be embedded in a larger architecture, as indeed they already are, you know, in some ways, in things like DeepMind’s…

Tom: We’ve had this sort of thing before though in the history of AI, haven’t we, where particular approaches have been flavor of the month, and you’ve got the expert systems for one. I mean, there was the early neural nets, which were much smaller neural nets, and now bigger neural nets and deep learning based on that, and, sort of, these systems that are sort of self-guided learning seem to be flavor of the month. But given that you’ve been in the field so long, do you see this as, you know, something that’s likely to run its course, and then will move on to something else?

Azeem: Is it the end of history?

Murray: So, I think there might be something special this time, and one of the indicators of that is the fact that there’s so much commercial and industrial interest in AI and in machine learning.

Tom: But that reflects the fact that it’s been making a lot more progress than any of those previous attempts.

Murray: Exactly, yeah.

Tom: Isn’t there a problem, though, with the expert-based systems, that you could ask them why they reach particular conclusions? And with a self-driving car based on an expert system and, you know, it decides — and the classic, you know, trolleyology dilemma of does it, you know, run over the…

Sonal: Oh, the school bus with the children?

Tom: Yeah, exactly. All of those sorts of things, I mean, which I think are very interesting, because even now, you have sort of implicit ethical standards in automatic braking systems. You know, is it small enough — if it’s that small, it’s probably a dog, if it’s this big, it’s probably a child.

Azeem: So, I think the trolley problem is definitely worth looking at and talking about, because we, as humans, don’t even agree on what the correct outcome should be.

Tom: So, if we’re thinking about the trolley problem, and one of these scenarios comes to pass, with an expert system, you know, rule-based system, you could say, “Why did you do this,” and the system will be able to say, “Well, basically, this rule followed,” and da, da, da. And with these more elaborate systems, where it’s more like gardening than engineering the way we built them — it’s much, much harder to get any of that kind of thing out of them. And it makes them much more capable, but isn’t that gonna be problematic, potentially?

Azeem: So, I think there are still objectives that we understand, right? So, the way that you build a system that predicts using machine learning is very utilitarian, right? You say there’s some cost function you wanna minimize, there’s some objective function we want to target, and then you train it. And you don’t really worry about the reasoning, because the ends, in a way, justify the means.

Tom: But the ends are gonna vary. I think we’re gonna see, you know, get into a car and you can, like, adjust the ethics dial.

Azeem: Right.

Tom: Because that recent research that suggested that people are totally fine about cars making utilitarian decisions as long as it’s not them that, you know, is in there.

Azeem: But in a way, we’ve lived in this world for a long time, we just haven’t had to ask the difficult questions.

Tom: Yeah.

Azeem: So, any time you pick up the phone, in the UK, it’s to your utilities provider, in the U.S., I understand it’s the Comcast clerk and customer service, you’re forced through an algorithm. You’re forced through a non-expert expert system, where the human at the other end has no discretion and has to just ply their way through a script, and we know how frustrating it is to live in that world.

Sonal: Very much so.

Azeem: Now, as we embed these AI-based systems, or machine learning-based systems, into our everyday lives, we’re gonna face exactly the same issues — which is, my car didn’t do what I wanted it to do, my toaster didn’t do what I wanted it to do — and I have no way of changing that. And so, this question about where is the utility function and what is the tradeoff has been designed in systems for 30, 50, 100 years or more.

Tom: It’s just becoming explicit now.

Azeem: It’s becoming much more explicit because it’s happening everywhere.

Murray: Also, I think there’s a big issue with these kinds of systems, which may work just the way we want them to work statistically. So, if you’re a company, then you know that it makes the right decision for 99% of the people who phoned up.

Sonal: Right, sort of an actuarial analysis.

Murray: If you are the 1% person who’s phoned up and got a decision which is not one that you like, then…

Tom: So, you don’t wanna be told, “Well, it was right statistically.”

Murray: Or, just “computer says no,” you know? You want to have reasons. Or more seriously, if you’re in government and you’re making some big decision about something, or in a company and making a big decision about something, you don’t want the computer to just say, “Just trust me. It’s statistics, man.”

Tom: Yeah.

Murray: You know, you want a chain of reasoning.

Tom: That brings up another aspect of this, which I find quite amusing, which is that there are quite a lot of sci-fi future — Sir Iain Banks’s future, and the Star Trek future — where you basically have a post-capitalist society, because you can have a perfect planned economy, because an AI can plan the economy perfectly. But, you know, there is a question of how plausible that is. But I wonder, you know, the extent to which you think AIs will start to be used in policy-making and those sorts of decisions.

Murray: Well, I suspect that they will be, and I think that’s why, in fact, this whole question that you’re raising — of trying to make the decision-making process more transparent, even though it’s based on statistics and so on — I think that’s a very important research area.

Azeem: And I think I would separate out the two areas of transparency. So, one is the black box nature, right? Can we look inside the box and see why it got to the conclusion it got to? The other side part that’s important is to actually say, “This is the conclusion we were aiming for.” And within policymaking, what becomes interesting, then, is forcing policymakers to go off and say, “That extra million pounds we could’ve put into heart research, we didn’t, even though it cost four lives. And we put it in something else because we needed to.”

Sonal: The kind of analysis we’re talking about — this actuarial analysis — we’re doing it every day already with insurance, which is just distributed risk.

Azeem: So, we don’t mind if humans do it, but we might mind if machines do it.

Murray: I think that’s the big issue is — will we be happy to hand over those decision-making processes to machines. Even if they made exactly the same decisions on exactly the same basis, you know, will society accept that being done in this automated way?.

Azeem: But, in a sense, we already have. It’s called Excel. It’s not even so much that we trust whether it works as a human works. What we’re doing is using Excel — we’re allowing ourselves to manipulate much larger data sets than we could’ve done just with pen and paper.

Sonal: Exactly. We’re sort of organizing the cells in our mind into cells in a spreadsheet.

Azeem: Into cells in a spreadsheet. And instead of having 100 data samples, you just look at 16 million and whatever, you know, that Excel can handle. So, we’ve already started to explore the space of decisions using these tools, right, to extend human reach.

Sonal: There’s some of the examples that kind of approximate where we can go. Because the examples that come to mind — I think historically of Doug Engelbart’s notion of augmented cognition, augmented intelligence. And then I’m even thinking of current examples, like Stephen Hawking. Helene Mialet wrote a beautiful book called “Hawking Incorporated” about how he’s essentially a collective. I mean, I don’t agree with this turn of phrase, but describing him almost as a brain in a vat, surrounded by a group of people who are anticipating his every need. And it’s not just, like, Obama’s crew who’s helping him get elected and his support team. It’s actually people who understand him so well that they know exactly how to help him interpret information.

Tom: Sort of like a group organism. In your space of possible minds, we have a whole bunch of minds that we could be, you know — and some people are trying to figure out already — which are animals, and then you’ve got the sort of social animals, the group minds there. And this, kind of, brings us to another ethical question from the previous one we were talking about which is, you know, the whole question of the evidence that octopuses are very — octopodes, we should say — are extremely intelligent, you know, has made some people change their mind about whether they want to eat octopus.

Azeem: Yeah, so I don’t eat octopus anymore.

Tom: You don’t eat octopus anymore. So, really…

Sonal: I’m vegetarian, so I don’t eat anything that…

Azeem: And as of today, I don’t eat crab either.

Tom: Anyway, so there’s the point of the extent to which a creature with a mind that we recognize is cleverer than we thought, whether it’s right for us to boss it around. But we’re gonna get this with AIs as well, aren’t we? Because the usual scenario people worry about is we are enslaved by the AIs, but I’m much more interested in the opposite scenario. Which is, if the AIs are smart enough to be useful, they will demand personhood and rights. At which point, we will be enslaving them.

Azeem: So, let me give you a practical example of that. There’s an AI assistant called Amy, which allows you to schedule calendar requests. And so, you know, I’ll send an email to you, Murray, and say, “I would like to meet you,” CC Amy. And then Amy will have a natural language conversation with you, and you think you’re dealing with my assistant. One of the things that I found was, I started to treat her very nicely. Because the way she’s been designed as a product, from a product manager perspective, is very thoughtful.

Tom: So, you didn’t say, “Organize lunch, slave.”

Azeem: Exactly. I didn’t do that, and I was quite nice to her. And then I had a couple of people who are incredibly busy write very long emails to her saying, “I could try this, or I could try this. If it’s not convenient, I could do this,” and I thought, “This is just not right. There is a misrepresentation on my part.” So, I then started to create a slightly apartheid system with Amy, which is — if you’re very important, and, Murray, you fell into that category — you’ll get an email directly from me, and other people will get an Amy invite. And it does start to raise some of the issues that are very present-day, right? They’re very present-day, because right now we have these systems.

So, I think one of the ethical considerations is, we need to think about our own attention as individuals and as people here. And as we start to interface with systems that are trying to be a bit like the Turk — the chess-playing device that pretended to be a human — we’re giving attention to something that can’t appreciate the fact that we’re giving it attention. And so, I’m now using a bit of computer code to impose a cost on you.

Tom: Well, actually, it’s like when I speak to an automated voice response system, you know, I speak in a much more precise way when it says, “Read out your policy number.” I know I’ve got to help the algorithm. I’m not…

Sonal: Right. We’re shaping our behaviors to, sort of, adapt to it.

Tom: Exactly. So, we already do it. When we type questions into Google, we miss out the stop words, and we know that we’re just basically helping the algorithm.

Sonal: We don’t ask questions anymore. We peck things out in keywords.

How AI may develop

Murray: I think Google will expect us to do that less and less as time goes by and expect the interactions to be more and more in natural language. So, I think between the two of you, you’ve raised the two, kind of, opposing sides of this deeply important ethical question about the relationship between consciousness and intelligence, and consciousness and artificial intelligence. Because, on the one hand, there’s the prospect of us failing to treat as conscious something that really is — that’s very intelligent — and that raises an ethical issue for how we treat them. Then, on the other side of the coin, there’s the possibility of us inappropriately treating as conscious something that is not conscious and is, you know, perhaps not as intelligent. So, both of those things are possible. We can go wrong in both of those ways.

And I think this is really one of the big questions we have to think about here is — and I think the first really important point to be made is that there’s a difference between consciousness and intelligence. And just because something is intelligent doesn’t necessarily mean that it’s conscious, in a sense of “capable of suffering.” And just because something is capable of suffering and conscious doesn’t necessarily mean that it’s terribly bright. So, we have to separate out those two things for a start before we kind of have this conversation.

Sonal: That’s a great point.

Azeem: And I think the thing we seem to care about is consciousness, from an ethical perspective, because we care a lot about the 28-week preterm baby, which is not very intelligent to…

Tom: And we care about dogs and cats as well.

Sonal: So, wait, where are we then when people have expressed fears? Because one of the things I think has compelled me to invite all three of you in this discussion is, none of you fall into one of these extremes of, like, completely, you know, cheerleading — like, “The future is dead,” and, you know, “We’re gonna be attacked and taken over.” Or the other extreme, which is, sort of, dismissive, like, “This will never happen, ever.” Where are we?

Tom: We’re all in the sensible middle, aren’t we?

Murray: I guess you’re asking where are we, you know, historically speaking now, right?

Sonal: In this evolution and this moment.

Murray: And I think the answer is we just don’t know. But, again, there’s a very, very important distinction to be met. This is the trouble with academics. We just wanna make distinctions, you know?

Azeem: Distinctions.

Tom: Journalists want to make generalizations.

Murray: Yeah, they all can be important, and this is a case where it’s really important to distinguish between the short-term specialist AI — the kind of tools and techniques that are becoming very, very useful and very economically significant — and general intelligence — artificial general intelligence, or human-level AI. And we really don’t know how to make that yet, and we don’t know when we’re gonna know how to make that.

Tom: You don’t sound like a believer in the, kind of, takeoff theory — that, you know, the AI is able to develop a better AI in, you know, less time, and so you get this sort of runaway. And I think that’s a very unconvincing argument. It assumes all sorts of things about how things scale.

Azeem: So, I think the takeoff argument — it has a sense of plausibility. It’s the timing that’s the issue. So, I can’t deny the possibility that we could build systems that could program better systems, and that could start program better systems.

Tom: But the point is that a system that’s twice as good, if it’s, say, you know, an order — it might scale non-linearly. So, it might be 256 times harder to build a system that’s twice as good. And so, every incremental improvement is going to take longer, and it’s going to take a lot longer. And improvements in other areas, like Moore’s law and so on, again, are not fast enough to allow each incremental generation of better intelligence to arrive sooner than the previous one. So, there’s a simple scaling argument that this need not be linear.

Sonal: There’s also a classic complexity brake argument. I mean, there are so many different arguments.

Azeem: There are lots. And, you know, as we start to peel apart the brain and our understanding of the neurological bases for how, kind of, cognition functions work, we learn more and more and we see more and more complexity as we dig into it. So, in a sense, it’s a case of, “we don’t know what we don’t know.” But we’ve been here before, before we’d understood this idea of there being a magnetic field, and needing to, you know, represent physical quantities with tensors, rather than with scalars or vectors. We didn’t see magnetic fields. We didn’t understand them. We didn’t have mechanisms for manipulating them, because we couldn’t measure them, and, therefore, we couldn’t affect them.

And there would have been this whole set of physical crystals and rocks that were useless, because we didn’t know that they had these magnetic properties, and we didn’t know we could use them. Silicon dioxide being a great example — totally useless in the 17th century, quite useful now. And so, at some point, we might say that the reason we think this looks very hard, or it’s not possible, is because we’re actually just not seeing these physical quantities. When we touch on this idea of consciousness, you know, there is this idea of integrated information theory, which is this theory that, you know, consciousness is actually an emergent property of the way in which systems integrate information, and it’s almost a physical property that we can measure.

Tom: Yeah, or we could be, like, I suppose, like Babbage saying, “I can’t imagine how you could ever build a general-purpose system using this architecture,” because he can’t imagine a non-mechanical architecture for computing.

Murray: Right.

Sonal: Murray, where do you fall in this singularity debate? And you’re not allowed to make any distinctions.

Murray: Well, without making any distinctions, I’m still gonna be boringly academic, because I wanna remain kind of neutral — because I think we just don’t know. I think these arguments in terms of recursive self-improvement — the idea that if you did build human-level AI, then it could self-improve — I think there’s a case to be answered there. I think it’s a very good argument, and, certainly, I do think that if we do build human-level AI, then that human-level AI will be able to improve itself. But I kind of agree with Tom’s argument, that it doesn’t necessarily entail it’s gonna be exponential. <crosstalk>

Sonal: Right. So, actually, to pause there for a moment, you started off very early on talking about some of the drivers for why you’re excited about this time — why this time might be different. What are some of those more specifically? Like, Moore’s law we’ve talked about, I mean, because that’s obviously one of the scalers that sort of helps.

Murray: Yeah, yeah. So, basically, what’s driving the whole machine learning revolution, if we can call it that, is — I mean, there are three things. And one is Moore’s law, so the availability of a huge amount of computation. And, in particular, the development of GPUs, or the application of GPUs to this whole space has been terrifically important, so that’s one. Two is big data, or just the availability of very, very large quantities of data, because we have found that algorithms that didn’t really work terribly well on what seemed like a lot of data — you know, 10,000 examples — actually work much better if you have 10 million examples. They work extremely well. So, the unreasonable effectiveness of data, as some Google researchers call it, so that’s two. And then the third one is some improvements in the algorithm. So, there have been quite a number of little tweaks and improvements to ways of using backpropagation and the kind of neural network architectures themselves.

Azeem: So, I add three more to that list. One is, in practical software architectures, we’re starting to see the rise of microservices. What’s nice about microservices — it’s a very, very cleanly defined system. So, you don’t need generalized intelligence, you just need very specialized optimizations. And as our software moves from these hideous spaghettis to these API-driven microservice architectures, you can apply machine learning or AI-based optimizations to improve those single interfaces. So, lots of reasons…

Sonal: Right. That’s actually closely tied to the containerization of code at the server level, and there are so many connected things with that.

Tom: So, it’s much easier to insert a bit of intelligence into a process.

Azeem: And then the other two are — so there’s this phrase, which I’m sure Andreessen Horowitz is familiar with — which is, “software is eating the world.” And as software eats the world, there are many more places where AI can actually be relevant and useful. So, you can start to use AI in a food delivery service, because it’s now a software coordination platform, not chefs in a kitchen, and, therefore, more places for it to play. And this is a commercial argument, and so Murray’s explained some of the technical reasons.

The third commercial argument is accelerating returns. So, as soon as you start within a particular industry category to use AI and get benefit from it, the increased profits you get, you reinvest into more AI, which means your competitors have to follow suit. So, you can’t now build an Xbox video game without tons of AI, and you can’t build a user interface without using natural language processing and natural language understanding. So, that forces the allocation of capital into these sectors, because that’s the only way that you can compete.

Business vs. academia

Sonal: So, given those six drivers, not three, who are the entities that are gonna win in this game? Like, is it startups, is it the big companies, is it government, universities?

Murray: Well, if you were to ask me to place a bet at the moment, I would place it on the big corporations like Google and Facebook.

Tom: Basically, they have access to the data, and everything else you can buy, but that you can’t, right?

Azeem: Yeah.

Murray: Right. And also, they have the resources to buy whoever they want.

Tom: Right.

Murray: An interesting phenomenon we’re seeing in academia these days is that it used to be the case that the people who, you know, were very interested in ideas and intellectual things — they wouldn’t necessarily be tempted away to the financial sector. But we’d still retain a good chunk of them in universities to do Ph.Ds. But now, companies like Google and Facebook can hoover up quite a few of those people as well, because they can offer intellectual satisfaction as well as a decent salary.

Tom: But, also, they are getting the — you know, the Silicon Valley is the new Wall Street argument. They are getting the people who used to go into financial services, which is a good thing. I remember the head of a Chinese sovereign wealth fund saying a few years ago, you know, “You Westerners are crazy. You educate your people in these fantastic universities, and then you take the best people and you send them into investment banks where they invent things that blow up your economy. I think you have to do something useful.”

Sonal: Right. We used to say that…

Tom: And the whole of, you know, the Chinese politburo is they’re all engineers, and, you know, they value sort of engineering culture and engineering skills, and they can’t believe that we’ve, sort of, wasted it this way. So, I think it’s fantastic that, you know, now there’s less money to be made at Wall Street than maybe there is in Silicon Valley, and people like going West. I think that’s only got to be a good thing.

Azeem: Coming back to who the winners might be, I mean, I think there is a strong argument to say that having the data makes a lot of the difference.

Tom: Yeah, no, I think that’s the crucial distinction.

Azeem: I think you’d be hard-pushed to say — look at voice interfaces, you know, between Apple, Microsoft, Google, Baidu, and Nuance. That’s quite a crowded field already, so it does feel like there are a lot of AI startups who are going to run up against this problem of both data and distribution. But, that said, there are particular niche applications where you can imagine a startup being able to compete, because it’s just not of interest to a large company now, and they may then be able to take a path to becoming, you know, independent.

Tom: Look at, say, Boston Dynamics. Because one of the ways you train machines to walk like animals is not to use a massive internet data set of how cats walk. So, in that case, not having access to that data is not an impediment, and you can develop amazing things, and they have done. They’ve been acquired by Google.

Murray: Actually, DeepMind are another example of the same thing. Because if you want to apply reinforcement learning to games — and that’s enabled them to make some quite fundamental sort of progress — you don’t need vast amounts of data, right? You just need to play the game loads, and loads, and loads…

Azeem: We’re just reinforcing your thesis there, Murray, which is that Google’s gonna buy all of these companies.

Murray: Well, yeah. Well, I ought to put in a little pitch for academia, yeah. Because the one thing that you do retain by staying in academia is a great deal of freedom, and the idea to disseminate your ideas to whoever you want — so you’re not in any kind of silo. And some of these companies are very generous in making stuff available.

Azeem: Right, with TensorFlow, yeah.

Murray: TensorFlow is a great example of that that we’ve just seen Google release. But, nevertheless, you know, all of these companies are ultimately driven by a profit motive, and they are gonna hold things back.

Tom: We’ve just seen, for example, Uber has snaffled the entire robotics department for Carnegie Mellon. Presumably, the motivation of the people there is that, you know, finally, the work that they’ve been doing on self-driving vehicles and so on…

Sonal: Right, you know, should get out into the world.

Tom: And you can actually make a difference. And, yeah, I’m sure they get much better pay, but, I mean, the main thing is that rather than doing all of this in a theoretical way, here is a company that’s prepared to fund you to do what you want to do.

Sonal: You can finally have impact.

Tom: In the real world, in the next decade — and that must be amazingly attractive.

Murray: It is incredibly attractive, and, of course, many, many people, you know, will go into industry in that way. But there’s also something attractive for a certain kind of mind in staying in academia, where also you can explore maybe some larger and deeper issues that you — I mean, for example, like, you know, Google aren’t gonna hire me to think about consciousness.

Sonal: Or they might. You never know. I mean…

Azeem: There’s also this question about the kind of questions that you will look at as an academic. So, the trolley problem being a good one. There are all sorts of ethical questions that don’t necessarily naturally play a part in your thinking when you think about your Wall Street <inaudible>.

Sonal: That’s right. And corporate entities aren’t set up to think about that. Like, Patrick Lin studies the ethics of robotics and AI, and that entire work is funded by government contracts and distributed through universities. So, okay, so the elephant in the room — AI and jobs, what are our thoughts on that?

Azeem: Well, I think look at where we are today, which is that we’re quite far away from a generalized intelligence. And, you know, McKinsey just looked at this question about the automation of the workforce, and they did something very interesting. They looked at every worker’s day, and they broke it down into the dozens of tasks they did and figured out which ones could be automated. And their conclusion was, we’ll be able to automate quite a bit, but by no means the entirety of any given worker’s job — which means the worker will have more time for those other bits, which were always the social, emotional, empathetic, and judgment-driven aspects of their job. Whether you’re a delivery person…

Sonal: Right, the creative…

Murray: Or creative, yeah.

Tom: Yeah, and I’ve read that and I thought, “Hang on a minute though,” because what they’re looking at is, they’re looking at the jobs of basically well-paid information workers and saying, “Well, you can’t automate their jobs away.” But the bits you can automate are the bits that are currently — many of them are bits that are currently done for them by other people. So, the typing pool, you know, we got rid of the typing pool because we all type for ourselves.

Sonal: Factory workers.

Tom: Exactly. So, you know, this means that the support workers for those people are potentially put out of business by AI.

Azeem: Or they’ve moved up.

Tom: Yeah, or they have to find something else to do. But I think just because the architects are safe doesn’t mean that the people who work for the architects are.

Azeem: If you walk down a British high street, the main street today, one of the things you’ll notice is a plethora of massage parlors, nail salons, and barbershops.

Tom: Service businesses.

Azeem: Because these are the things that you can’t do through Amazon. Everything else you can do through Amazon or Expedia.

Tom: Interior design, yoga, Zumba, whatever. That’s the future of employment.

Murray: And coffee shops.

How far will AI go?

Sonal: Okay. So, we’ve talked a lot about some of the abstract notions of this, and, you know, this is not a concrete answer, because we’re talking about a fiction film, but how possible in reality is the “Ex Machina” scenario? And a warning to all our listeners that spoiler alerts are about to follow, so if you’re really bitter about spoiler alerts, you should probably sign off now. The reality that the character — the main embodied AI, Ava — could essentially fight back to her enslavement. To me, the most fascinating part of the story — and we have no time to talk about it right now, but I do wanna explore this at some point in the future — is sort of the gendering of the AI, which I think is incredibly fascinating. How real is that scenario?

Murray: Yeah. So, the whole film is predicated on the idea — well, it seems to be predicated on the idea that Ava is not only a human-level AI, but is a very human-like AI.

Sonal: So, the humanoid aspect?

Murray: Human-like, of course — she looks like a human, but, I mean, human-like in her mind.

Tom: And her objectives.

Murray: And her objectives and her motives.

Sonal: Her needs, her emotions.

Murray: You know, so if you were a person in those circumstances, you would want to get out, right? And, in fact, very often, science fiction films that portray AI — that’s a fundamental premise that they use for how they work — is that they assume that we’re going to assume that the AI is very much like us, and has the same kinds of motives and drives for good or for ill. They can be good motives or bad motives. They could be evil, or they could be good, you know? But it’s not necessarily the case that AI will be like that. It all depends how we build it. And if you’re just gonna build something that is very, very good at making decisions, and solving problems, and optimizing…

Tom: It may just sit down and say, “I just wanna sit here and do math.” We really have no idea what their motivations will be.

Azeem: Yeah, I mean, if the AI had been modeled on a 45-year-old dad, it would’ve been perfectly happy being locked up in its shed at the bottom of the garden with an Xbox.

Sonal: And some of their magazines, right.

Murray: Well, but then just moving on a little bit from that, though, it is worth pointing out some of the arguments that people like Nick Bostrom and so on have advanced — that you shouldn’t anthropomorphize these creations. You shouldn’t think of them as too human-like.

Sonal: In the film — I saw it three times as I mentioned — on the third watching, I noticed that there is a scene where Nathan has, like, a photo of himself on his computer where he programs. Like, he’s on his computer all day, like, hacking the code — which I think is so fascinating because there’s almost this narcissistic notion, which kind of ties to your notion of the anthropomorphization of the AI.

Tom: You use the term anthropomorphism because it is — I’ve noticed you use the word creatures to refer to AI, and I think that’s really telling, because they are going to be more like aliens, or more like animals, than they are like humans. I mean, the chances of them being just like humans are very small.

Murray: We might try and, you know, architect their minds so that they are very human-like. But can I just come back to the Nick Bostrom kind of argument? Because he points out that although we shouldn’t anthropomorphize the AI, nevertheless, if we imagine this very, very powerful machine, capable of solving problems and answering questions, that there are what people who think about this refer to as convergent instrumental goals.

Sonal: You’ll have to break that down for us really quickly, yeah.

Murray: So, anything that’s really, really smart is gonna have a number of goals that anything is gonna share, and they are gonna be things like self-preservation and gathering resources. If it’s sufficiently powerful, then any goal that you can think of, if it’s really, really good at solving that goal, then it’s gonna want to preserve itself, first of all. Because how can it, you know, maximize the number paper clips in the world — to use Nick Bostrom’s argument — if it doesn’t preserve itself or if it doesn’t gather as many resources as it can? So, that’s their argument for why we have to be cautious about building something that is a very, very powerful AI, a very powerful optimizer. That’s the basis of the…

Sonal: Because it will always be optimizing for that.

Murray: So, I think the very important thing here is that the media tends to get the wrong end of the stick here, and think of this as some kind of evil Terminator-like thing. And so, we might think that those arguments are flawed — the arguments by Bostrom et al. Maybe we do, maybe we don’t, but I think there’s a very, very serious case to answer there, and in order to answer it, you have to read their arguments. You can’t just, kind of, assume what you think their arguments are.

Sonal: Right, the derivative. That’s the problem with a lot of technology discussion in general is to always revisit these in a very derivative way, versus viewing the original. But putting that exhortation aside, how do people make sense of this? Like, how do they make sense of what is possible?

Murray: So, how do we think about the future, really, when it comes to artificial intelligence? And I think the only way to do it is actually to, kind of, set out a whole tree of possibilities that we can imagine and try to, you know, not sort of fixate on one particular way that things might go — because we just don’t know where we’re gonna down that tree at the moment. So, there’s a whole tree of possibilities. Is AI gonna be human-like or not? Is it gonna be embodied or not? Is it gonna be a whole collection of these kinds of things? Is it gonna be a collective? Is it gonna be conscious or not? Is it gonna be self-improving in this exponential way or not? You know, I don’t think we really know, but we can lay out that huge range of possibilities, and we can, you know, try to analyze each possibility and think, you know, what would steer us down in that direction and what would the implications be.

Sonal: That’s a great way to approach it. Well, that’s another episode of the “a16z Podcast.” Thank you so much for joining, everyone.

Azeem: Thank you.

Murray: Thank you.

• 

  Azeem Azhar
• 

  Murray Shanahan
• 

  Tom Standage
• 

  Sonal Chokshi is the editor in chief as well as podcast network showrunner. Prior to joining a16z 2014 to build the editorial operation, Sonal was a senior editor at WIRED, and before that in content at Xerox PARC.

  Follow Twitter

What began as a scientific approach to creating and managing startups has now become a worldwide movement for companies of all sizes — and for creating (or rather rediscovering) entrepreneurs in all places. Not just inside startups, not just for software, and not just inside Silicon Valley. It’s about unlocking human creativity everywhere. Perhaps even reinventing the firm.

As utopian as that sounds, Eric Ries — who pioneered the lean startup movement and wrote the definitive book on it — argues the case in this episode of the a16z Podcast. But has it become too much of a religion? One where people apply the letter of, but not the spirit, behind lean startup principles?

Ries, who recently crowdsourced a leader’s guide for practitioners to test and evolve the very concepts he first published 5 years ago, shares lessons learned — as well as the true meaning of overused terms like ‘MVP’ and ‘pivot’. Ultimately, lean startups are about how to make decisions and build new products under conditions of high uncertainty. Without having to chisel the principles into stone tablets.

Show Notes

• Where the idea of “lean startup” came from and how it’s resonating with larger businesses [0:00]
• How innovation is often stifled in large organizations [10:22], and why it’s difficult to be entrepreneurial [20:23]
• The impact of software [27:31] and keys to unlocking people’s creativity [37:16]
• How to implement lean startup, and a discussion of the most popular terms [41:52]

Transcript

The origin of “lean startup”

Sonal: Hi, everyone, welcome to the “a16z Podcast.” I’m Sonal, and Michael and I are here together interviewing Eric Ries, who wrote the book on “The Lean Startup.” And it’s been, actually, now over five — is that right, Eric? About over five years since that book came out?

Michael: Yeah, it came out in 2011. So, next fall will be the five-year anniversary of its publication. And of course, the publication was the, like, the end of a process of blogging and writing and, kind of, building that community.

Sonal: I was about to say because I feel like your book came out well after lean startup was already on its way. Well, the other interesting thing that I think we’d like to talk to you about is how things have changed since then and now, but more importantly — we’d love to also — because one of the things that’s happened is that lean startup is now outside of Silicon Valley. And it’s gone also outside of startups, interestingly. So, we want to hear a little bit about your insights from that. And then kind of talk about not just the lean startup, but what it means for the future of the firm.

Michael: What is a lean startup? And if it’s not a startup, how does that apply here? 

Eric: So, the basic idea is to take a more scientific, more iterative, and more customer-centric approach to product development and customer building. It’s called lean startup, because we take ideas from lean manufacturing around cycle time and batch size and iteration, but we apply them not to the factory floor, but to the process of innovation itself — where the signal that pulls work from us is not the order from a customer, since in most startups, we don’t have a customer yet. We don’t know who the customer is going to be. The pull signal is our hypotheses, our beliefs, about what the customer will want in the future. Or another way of saying it is, we write these business plans that are full of assumptions and hypotheses and guesses about the future. And lean startup says, “Rather than take those things for granted and hope that they’re true, let’s test them scientifically — accepting the fact that every startup faces high uncertainty about the future.”

Sonal: So, how do you respond to critics who say, like — I would say, actually, thinking about it more ethnographically, that you’re essentially only then building for people who know what they — like, it’s based on what the market is getting as feedback versus, like, true internal creativity that’s an internal compass.

Eric: Yeah, it’s an interesting thing, because customers don’t know what they want. And everyone knows that to be true, or talks about it. And if you imagine I’m a scientist, I came in, “I’m gonna do chemistry,” and then I’m like, “Oh, shoot, the electrons don’t know what they want. And I can’t do a focus group with protons. Like, forget it, I guess I can’t do chemistry.” Like, that person is bonkers. <inaudible> mental problem. For some reason, we talk about customers, we get confused about this. Taking a scientific approach means having a very strong belief — some would say a visionary belief about what’s going to work — and finding out if that vision is right. And what we discover with — if you look at the stories of entrepreneurs, not the kind of movie version and the Hollywood version that you read in the magazines, but we get the real stories, you know, with actual entrepreneurs — what you discover is that even if the vision is right on, and they see the future, the specifics of the strategy often have fatal flaws.

And until you systematically figure out which elements of the strategy makes sense, right? Do we have the right business model and the right target customer and the right, you know, approach to disruptive innovation? Are we on the right technology platform? Have we gotten our timing right? You have to work to get all those details right. And the best entrepreneurs I know are extremely rigorous in their thinking. Because — it’s precisely because they care so much about the vision, they feel like they have to get the details right. And so all we’re saying is that we should approach that with the most, you know, scientific kind of rigor that we can.

Sonal: Right. Essentially, a business plan is a hypothesis.

Eric: That’s right.

Sonal: You can, sort of, test out not just whether it works or not really, but actually the details of how to execute on that because the vision is a large thing.

Michael: So, why is it then — we talked about this was the five-year anniversary, how is it that this has resonated so well outside of the Silicon Valley startup world? You know, what are the things that people were, kind of, grasping to, whether it’s aspirational or where you’re like, “I can roll up my sleeves and do this?”

Eric: Yeah, you know, it’s a question I think a lot about, and because no one is more surprised than I am about how this has grown beyond just a book and beyond just a set of like idiosyncratic ideas into a movement.

Sonal: It’s like a movement for a lot of people. Right.

Eric: You know, I feel like I have to disclaim periodically that lean startup is not a religion. Like, if you have to say that then something’s a little suspicious here. You know, I think that’s fair game. I think that, first of all, you gotta look at the timing of it. I started writing about this idea in 2008, 2009, right in the height of the financial crisis, when people had kind of lost confidence in, kind of, traditional ideas about how a company should be built. So, there was a — kind of a hunger for something new. I think having it called “lean startup” at a time when people were freaking out about money was probably good timing, good branding. And I think that as entrepreneurship has become more democratized…

Sonal: What do you mean by entrepreneurs have become democratized?

Eric: So, you know, I think there was a time when if you wanted to become an entrepreneur, you had to have the right connections, you had to have access to a lot of money and capital, you had to, kind of, look the part. And, you know, if you look at the, like, famous entrepreneurs in history, a lot of them came from very specific kinds of backgrounds. And it was a rare thing. It was considered a little bit crazy. You know, as the semiconductor revolution has, kind of, systematically, you know, eradicated barriers to entry for everything. I think about it — you go back to the old Karl Marx idea of who owns the means of production.

We are now living through a time where you can rent the means of production. Which means that if you want to test an idea, if you want to start something in your dorm room at Harvard, you know, just because you started small doesn’t actually mean it can’t scale to something, you know, massively large. And so, you have that phenomenon. Plus, the fact of the internet itself has made the idea of entrepreneurship as a possible career path accessible to people who, you know, used to have to, like, know somebody in Silicon Valley to, kind of, like, understand what was happening here. Now, Silicon Valley is like a reality TV show that everyone in the world watches.

Michael: I want to go back to Karl Marx, just for a minute. You say who owns the means of production? Sounds like Amazon does. But maybe owning the means of production isn’t worth as much as — well, it’s worth a lot to Amazon, but…

Eric: Yeah. Well, I think the kind of older industrial thinking was that if you own something, then the right thing to do was to control it in order to maximize the return you would get from it. And I think what we were having a recognition — probably partly through the just the realities of how technology works. But I think also partly due to, you know, like better management thinking, frankly — is that if you make those tools available to other people so that you don’t have complete control —  if you turn it into a platform, if you give people open access — the sum total of the creative power that you unleash when you do that means that you have a smaller piece of a much larger pie.

Sonal: So, speaking of that shift in management thinking, let’s talk about the surprise that you have and that we’re surprised by — that big companies are trying to be lean startup-like. I mean, because I think one of the things that’s interesting about the management point is that we’re living through an interesting time right now, where a lot of the old school theories about management — like, they may apply in certain ways but in other ways, there’s a real gap — a real need for, like, how do you really — the world has changed, and it’s very trite to say that, but it’s true. So, talk to us more about that.

Eric: I’ll tell you a funny story. So, when I was doing the research for “The Lean Startup,” since we’re talking about the five-year anniversary — doing the research for the book, I read everything I could get my hands on about management. Business books. I felt like I needed to really understand what came before. And I remember reading about the development of our modern accounting system, which I didn’t really know a lot — I’d never thought about, what — going back to the 1920s and Alfred Sloan and that whole movement. And I realized at a certain point that we had developed this accounting system not to keep track of the money, which is kind of how we think about it now — it was originally developed as a system of accountability, so you could figure out which managers are really doing a good job. Because if some manager makes more money this year than last year, they say, “I should be promoted.” But you’re like, “Hold on, if you had a forecast of how much money you were supposed to make this year and you fell short of the forecast, you don’t get promoted for that.”

We take that so for granted in our world today. But when I first read that, I almost fell out of my chair, because I’m, “Wait a minute, I’m sorry. You’re telling me there’s a part of the world — or a part of history or anywhere — where people make forecasts for things and then they come true? I never heard of that.” You know, as an entrepreneur, I had no idea why I was always asked to make forecasts. I thought the spreadsheet I put in my business plan when I raised VC was like a kabuki ritual I did to show how much pain I could endure — to show that I was tough. I didn’t think anyone would read the content of those spreadsheets and take them seriously. We just made those numbers up. We had no idea. But of course, if you, like, think about that — like, that is so specific to my experience as an entrepreneur. And I was like, “Okay, well, how is it that I’ve never seen an accurate forecast, but in the real world, in mainstream business, forecasts rule the world? Why?”

And so I, you know, studied that and learned about it. My realization was, like — a forecast is accurate only to the extent that it is an extrapolation from a long and stable operating history. So, any time you don’t have that — either because you’re doing something brand new, or because your long and stable operating history has just gone unstable on you, and, “Oh, my God” — then you’re in a situation of high uncertainty. And since my definition of entrepreneurship is trying to create something new under conditions of extreme uncertainty, then you are an entrepreneur, no matter what it says on your business card. Now, when I first started saying that out loud, you know, in talks and conferences, you know, a couple years ago, this weird thing happened to me where people would come up to me after the talk. And they would say, “Oh, hi, I’m a general manager at such-and-such large business, and I accept your challenge.” 

First time it happened. I was like, “What are you talking about?” <What challenge?> They’re like, “What you said because of the definition — that this can apply to companies of all sizes, all sectors, all industries.” I was like, “Yeah, I just said that.” So, they’re like, “I would like to go, you know, prove that that can work.” And [the] first time it happened to me, I was like, “Good luck. What does it have to do with me? Like, okay, that sounds great. This is just a deduction from a theory. That’s not my responsibility.” But then, luckily for me, really, some very visionary, you know, big company folks, kind of, dragged me kicking and screaming to the realization that there are real honest to God entrepreneurs — just as visionary, just as exciting to work with — inside some of these larger organizations as you walk down the street here in Silicon Valley.

Michael: And it is that lean startup, kind of, army being activated because of the conditions now facing these large companies, I mean, that you describe?

Eric: I think that’s right. I mean, look, big companies have always faced the forces of disruption. That’s an old phenomenon. But the rate of change and the, kind of, existential crisis that some of these companies find themselves in, I think, is more severe than ever. And there’s more of a recognition that they need entrepreneurial activity in order to survive. So, I think that has created a fertile ground for these ideas to come in, because so many companies want to act more entrepreneurially. You know, the idea that you have to act like a startup, or have internal startups, is, like, almost to the point of cliché here. And yet most of the companies that I meet that have that as a plan, there’s no plan. They don’t actually know how to make it happen.

Innovation and big business

Sonal: Right, exactly. But I do want to say that they do have long standing processes. I mean, R&D — deciding about where to invest your resources — that’s all about decision making under uncertainty. And there are actually entire schools of portfolio management for managing R&D around that, which — some of it is very not dissimilar to managing a portfolio as a VC, with a lot of startups. And so, I think I want to pause for a minute what you said about things happening faster and just make sure we reflect on what that really means. Because, that’s a phrase people use, like, “Rapid changes now. Things are dynamically evolving.”

What we’re really saying is that these big companies, which before could have quickly acquired startups to help them do some of these things — now those companies get too big. It reaches a point where their market cap is too big — we’ve observed this — for them to even consider affording the ability to then take on that company. So, it’s a really big deal that some of these companies can’t then innovate themselves. So, it’s a big problem you’re talking about is how to get at that. So, anyway — so, given this condition like what have you seen about how people are becoming — I mean, isn’t there a word called intrapreneurs? That’s been around forever? Like, what does that mean?

Eric: Yeah, you know, I don’t actually like that word that much…

Sonal: I don’t either.

Eric: …because I feel like, you know, an entrepreneur is an entrepreneur. It doesn’t matter if they live in a garage, or they wear a suit, or they have health benefits, or if they eat ramen noodles. Like, the surface details don’t matter. What matters is the fundamental, you know, reality of their job, which is — they’re trying to create something fundamentally new. And, you know, a lot of big companies actually have outstanding research labs, where they’re doing breakthrough science, and they manage the scientific uncertainty really well. And yet, as soon as they take those discoveries out of the lab, it’s like, “Okay, we’re done with the science, now the astrology.” And it’s like, they take these world class scientists and, like, “Forget everything you know about science, now we’re going to build a business plan. Tell me what’s going to happen in the future, and then make it happen through the power of your mind.”

Sonal: And that’s a great analogy from science to astrology.

Eric: And the scientists are like, “What are you talking about? That doesn’t make sense to me.” I’ve now worked with a lot of big companies, and a lot of companies with high science research labs. I meet these scientists, and I say, “So, tell me about some of the great breakthroughs you’ve had in the lab.” And they’re very excited to tell me about it. I say, “Great. Tell me which of those have been commercialized and are in products today, and which ones are sitting on a shelf.” And now you may as well start the violin music, because it is really depressing. Life-saving treatments, unbelievable breakthroughs, sitting there. It’s like, these companies can spend, through their technology readiness level, you know, analysis and — like they can do the smart research to spend $5, $10, $50 million to have a breakthrough. And then they often are not able to spend the, like, $2 million extra dollars that would be necessary to commercialize it, because they’re organized around functional silos. And there’s nobody whose job it is to actually take it out of the lab and make sure that the businesses that operate — that are mostly tied to quarterly short-term incentives, you know — have the ability and the incentives and the time and the space to figure out how to commercialize.

Michael: How do you mean that it’s nobody’s job to do that? I mean, $2 million —  that should get done, and then we’re off and running, right?

Sonal: It’s, like, everybody’s job.

Eric: Yeah, it’s everybody’s job, which means that it’s nobody’s job. My observation is that in most companies, there’s a missing function for entrepreneurship. So, there’s just nobody in charge of making sure that new ideas are taken from concept to execution. There’s not a disciplined, systematic way of testing new ideas. So, I used to think — I’m a Silicon Valley person, quite arrogant about the world. Our way is the best. You know, I used to think if I sat a big company person down, I said, “Hey, do you have any ideas for how your company could be better?” That they wouldn’t have any good ideas, because people are dumb if they work in big companies. That’s what I used to believe. And, you know, what I’ve learned the hard way is that that’s actually a dangerous question to ask, because you gotta have four or five hours to spare to get the answer, because you can’t shut people up. They got tons of amazing ideas of things the company could do better.

Sonal: Totally. The biggest thing is that disruptors always know what’s coming. It’s not like they don’t know what’s coming.

Eric: Yes, definitely. The information and ideas are in the company already, and the talent is in there, too. If you want to shut them up real quick, just say, “Okay, tell me the process to test out those ideas to see if they’re any good.” And they’re like, “I guess I got to ask my manager to ask their manager to go across the silo to the other manager.” So, like, visualize — you put the idea in a pneumatic tube, it gets sucked up the org chart somewhere, sent somewhere else, sent down and they’re just like, “Forget it.”

Michael: Because nobody pays attention.

Eric: No one pays attention. It’s, like, just discussing the process is, like, so painful. They’re like, “Forget it. I’m just gonna go back to doing my job.”

Sonal: But isn’t that…

Michael: So, here’s the thing, it sounds to me like these large companies are coming to you, in some sense, for youth. Like, you see in the movies where, you know, the witch is sucking the youth out of a child. But isn’t it the natural course that these companies — they get big, they get old, and they get plowed under?

Eric: I mean, that’s a very common belief in Silicon Valley.

Michael: But are you gonna change that?

Eric: I basically used to think that, too, but I don’t believe it anymore. I think that — you know, I come in as a consultant, or — so I come in as an outsider. And one of my strengths is, I don’t run a consulting company, that I don’t have 50 associates I’m trying to feed. I come in and I can tell companies the truth. What I tell them is, “Listen, as a consumer of products in your category, I don’t care if you live or die. I know that 5 or 10 years from now, the person who provides me this service — it’s going to be technology-enabled, it’s going to be developed according to these principles, it’s going to be rapidly evolved to suit my needs. So, as a consumer, I’m fine. Either because you will have adapted to that new reality, or some startup my friends down on Sand Hill Road are funding right this second will disrupt you and displace you, and I don’t personally care. So, whether you live or die.”

Now as a consultant, that’s not generally considered a nice thing to say, but it helps, because people who don’t want to hear that kick me out of their office and we save ourselves a lot of time and heartache. The ones that have been willing to say, “Okay, what would it take to do the transformation,” and it’s hard — I think I’ve seen really dramatic results. So, I have become a believer that even — that the bureaucracy, and slowness, and, kind of, ossification that we take for granted as a result of scale is not an inevitable development, but is a choice about the systems of management that we use.

Sonal: So, can you tell us a little bit more then about what you’ve seen on a big company side? Because frankly, I think, yes, you’re right — there isn’t, like, a chief entrepreneurial officer that owns a function, or the process for that matter. But there are groups within a company that try to — like, they have weird titles often, which is probably also a sign of not a good thing — but that do own this in some way, shape, or form. I mean, how do you prevent the risk of that just being yet another idea that doesn’t go anywhere? Like, how does lean startup, kind of, help with that? Like, what have you seen on the front lines of that?

Eric: Someone once came up to me after a talk and they said, “I have a question for you. There’s this guy in my company who has a C-level title.” I think his title was, like, chief innovation officer or something. He said, “That guy always comes to work in red pants. He has no responsibilities. He doesn’t do anything, as far as I can tell. He has no operational — he’s not responsible for any quarterly targets.” He’s like, “If I came into work dressed like that and talking like, I’d be fired in a heartbeat.” He’s like coming up to me like, “Can you explain to me what this person does?”

Sonal: That’s so funny. It’s like therapy.

Eric: Yeah, the guy in the red pants. “You mean that guy in the red pants?” You know, I was like, “Like, don’t blame me.” Like, paying lip service to innovation is easy. Doing it is really hard. And the question I always have is, like, if I want to find an entrepreneur inside a large organization, I can usually go to the middle manager. I say, “Listen, I got this kind of wacky, crazy project. Do you know a lunatic who would be dumb enough to sign up for this suicide mission?” They’re like, “Well, let me show you my secret black book.” There are these certain people that are known in the organization. If I pull their personnel file from HR, they are full of black marks. “Does not play nice with others.” My favorite is, like, “refuses to obey the standards.” They defy the standardization of work and that drives people crazy.

In a lot of companies, they get fired and bought back more than once, sometimes, for their startup. And it’s just, it’s crazy. Like, these people exist. Then it’s like, “What’s their job title?” In some companies, they’re a product manager, they’re an engineer, they’re a marketer, whatever they are, they — like, how do they get promoted? If they’re good at what they do, if they were really good entrepreneurs, how do they get promoted? Where do they live in the org chart? Who do they look up to? And what I’ve been working on lately — I’ve been thinking about — is, like, a grand unified theory of entrepreneurship, which is this. In most companies, including, by the way, startups that have gone through hyper growth…

Sonal: Like Google. I mean, how many people leave Google to start startups now?

Eric: I mean, it’s unbelievable. You have basically four completely different jobs that, to me, are the same. <Okay.> You have somebody who’s, like, a product manager, tasked with leading on brand new product development. So, you say, “We’re going to enter a brand new market with something that’s radically different. We’re going to try to be the disrupter for once.” You’re that person. You have someone in charge of a new internal system. And think about how many new IT systems you spend years and millions of dollars on, and they’re dead on arrival. It’s like, it’s the same old waterfall development. It was the old school Silicon Valley way — too much money, too little customer feedback, too long development. That’s true for new HR policies, new finance policies. I mean, you name it. That’s actually an entrepreneurial challenge too.

Then you have somebody in the business development, you know, office who’s supposed to be evaluating outside startups for purchase, and they make these catastrophic errors. You know, they’ll buy a startup for $900 million, and 3 years later, it’s worth $15 million. You know, like that — in most parts of the corporation making catastrophic errors like that would get you fired, but in Biz Dev, it’s like, we don’t know how to ask the right questions to figure out who’s doing a good job and who’s not. We wind up flooding the entrepreneur ecosystem with dumb money. And then you have people who are responsible for partnering with startups. Most big companies are terrible partners. They don’t understand how to pilot things. They don’t understand how to work with startups in a way that don’t kill them. They spent way too much time on contract negotiations, and they just — they’re unreliable partners.

Sonal: Oh, and then you have the “not invented here” syndrome, which pretty much kills anything…

Eric: Right. Totally terrible.

Sonal: …if you have an internal group of any sort.

Eric: So, what all those jobs have in common, to me, is this entrepreneurial reality — that they deal with situations of high uncertainty. And therefore, we need discipline as a company to be able to look at what are the right metrics to hold those people accountable? How do you identify who’s actually good at that job and who’s not? How do you share best practices across these similar things? So, you start to add up these tasks — a career path, a sense of professional pride and accomplishment, standardization, you know, having the right metrics — you’re like, “Gosh, that sounds a lot like a corporate function, right? That’s what — we do that in marketing. We do that engineering. We do it in R&D.” And people are like, “Well, entrepreneurship is too creative to be managed.” But it’s like, if we can manage R&D, like, we can manage Muppet Labs. You know, people working on [a] Nobel Prize, they can be managed.

Sonal: That’s right.

Eric: So, I just — I don’t buy it. I think that we have just made a mistake about how the companies were organized, so that we can pay lip service to innovation and claim we want to have continuous innovation. But I ask these CEOs that I meet with all the time, “Who’s in charge of making sure that that happens?” And they don’t know. There’s nobody in the organization they can point to for accountability on that score.

Lean principles for large organizations

Michael: So, what are these organizations that you talked to who are, you know, stuck in the present and perhaps in the past? You know, when you think about the organization of the future, what does that start to look like? You know, when you look at the org chart, or when you look at it, sort of, structurally otherwise?

Sonal: And just to pause there for a second, I think we’re not just asking about, you know, “What is lean startup applied to a big company?” It’s really about reinventing the nature of the firm.

Eric: Yeah. So, people talk about lean startup for startups and then lean startup for the enterprise, which I think is really silly. I understand why people do it, but it doesn’t make sense to me. And the reason is, only the bad startups are small companies, right? The people are unintentionally small companies, but that’s not what they’re trying to do. And I meet all kinds of entrepreneurs who became an entrepreneur because they hate working for big companies, and they find them bureaucratic and sclerotic. And I always ask them the same question, I say, “Listen, if you hate big companies so much, why are you trying to create a new one?”

And what happens is, five years later, they have all the success, they achieve product market fit — you know, in the blog posts, in the books — you know, in everywhere except for Ben’s book. It sounds like when you get product market fit, all your problems are solved. But you know, like, the reality is, everything gets way harder. And the curse of it is, you have these founders who I meet with all the time now, who have 100 to 500, 1,000-, 5,000-person organization — and they’re like — you know, you’ve got to get them privately off the record. I’ll be like, “I’m not sure I would even want to work here.” I mean, I got a good gig because I’m the founder CEO, and that’s pretty fun. But, like, if I wasn’t, would I actually want to, like, be a regular employee here? And if I was trying to do something entrepreneurial here, would people understand how to do it? And they don’t… 

Sonal: Even the founder CEO is, like, desperate to hold on to that feeling of how it was in the first year, first five years.

Eric: Right. Yeah, they can feel the loss of it, and they feel the frustration because people come to them with plans. I mean, I was just talking to a very famous recent mega success story. And they’re telling me this unbelievable story, where the founder was being pitched on a new app, you know — the new big line of business for them. The founder was like, “Okay, that seems like a pretty reasonable experiment. It should take about two…” And he’s like, “I could probably code that in a week or two weeks.” He’s like, “I can do it in a week. So <inaudible>.” And the team was pitching him on, like, a 12-month, multi-million dollar, like, mega plan that’s, like, overengineered to the max. And he’s just like, “What have I done? How can I possibly have a company where people think that’s a good idea?”

And the challenge for me, you know, talking to them is to be, like, “Look, I hate to be the bearer of bad news, but you need to look in the mirror. And now you’re looking at the problem, because you have to make a fundamental choice as any kind of leader. Are you trying to preserve that entrepreneurial feeling for yourself, or are you taking the steps necessary to push that entrepreneurial opportunity down into the ranks of your company?” And, in fact, most of the CEOs who are good at this realize that they are so used to being on one side of the accountability table — they’re the entrepreneur pitching on their board, and their VCs always asking them about progress and having that negotiation — what they don’t realize is that now the roles are reversed.

For the people inside their company, they’re the VC. They’re the source of funding and political capital that everybody needs to sustain what they do. So, when they’re being pitched crap, it’s the same as — I know a lot of board members in a lot of companies that are like, “Why are these companies always give me these stupid reports and these dumb updates?” Like, because that’s what they think you want to hear. So, if you want them to do something different, you have to be the one to say, “Here’s how I intend to hold you accountable.” And then, when we have that real conversation, a lot of these entrepreneurs, they themselves have amazing intuition and really good natural instincts for, like, what are the right metrics to look at? You know, they all, kind of, naturally gravitate to the minimum viable product. And the idea that, you know, a small number of extremely passionate customers is way better than a large number of people who kind of are indifferent about your product. But they don’t understand why the people that work for them don’t have those instincts.

Michael: So, what’s the hard — I mean, it all sounds pretty hard, I’ll be honest. But, like, what stands in the way? Like, again, if I know what I need to do, if I’ve done it before as a startup, now I’ve — my startup has grown and successful — like, what stands in the way in the companies that you talk to from them actually realizing their entrepreneurial, kind of, style and flavor and goals?

Eric: Yeah, the problem is strictly scale. So, the founder — they can, kind of, go on a side project and be like, “Forget it, I’m going to do it myself. And I’m gonna step out of the CEO chair and go show this project team what to do.” But they can only really do that one, like, at most one project at a time. But these companies are too big for that. They need to be doing — you know, if you want to have a new successful disruption every couple of years, you need to have hundreds of experiments going at any one time. So, then — it’s like, you need to have a way to train and reward the entrepreneurial people in your organization. 

And they have natural instincts for that. But that’s really different from saying, “How do you teach that approach to other people.” And that really, I think, is why lean startup is taken off inside these larger organizations — which, by the way, is both legacy organizations that are, like, 100 years old and are now adopting it recently, as well as these companies that started as lean startups but then blossomed into this traditional company.

Sonal: And, by the way, by organizations, that includes governments.

Eric: Oh, yeah.

Sonal: Because I’ve heard like the former CTOs of the United States talking about how they’re trying to adopt a lean startup-like methodology inside of government.

Eric: And it’s amazing. I mean, I was actually just in D.C. the other day and meeting with teams. They showed me this almost unbelievable story about this team inside of the Immigration Service that processes applications still on paper. And the paper applications can’t be stored in a normal office building, because the backlog is so large that the physical weight of the paper requires a structurally reinforced room. And, therefore, at one of the processing centers — I think they said in Kentucky, the processing center is literally in a cave. Like, that’s not a metaphor. Now, they have historically had these big outside contractors come in and do these multibillion dollar IT initiatives. They were telling me about one they spent, I think, a billion dollars in seven years, and it couldn’t process even one form faster than paper.

Michael: But the great thing is the solution was to move to a cave.

Eric: Right, right. I always think about the human cost of these bad management systems. Think about the poor people actually trying to get this work done the best way they know how, and that’s the best they could come up with. They sent a lean startup team in — I think from the United States Digital Service at the White House. And, you know, they partnered — it wasn’t just IT people coming and telling everybody what to do. But they had real partnership, real user-centered design, real lean startup experimentation techniques. And they built a small team of technologists and people who are experts in the processing center. And they’re now processing something like 40% of the applications digitally. I think it took, like, six months.

So, instead of spending a billion dollars in, kind of — I call it the healthcare.gov plan. Instead of executing the healthcare.gov plan, we did something a little bit better. And I love those stories, because when I meet with private sector folks — I meet a lot of CIOs now, and they’ll tell me about some new major initiatives they have going, and I’ll say, “Oh, that’s great. Sounds like the healthcare.gov plan. I’m sure you’d be fine.” And they’re like, “How dare you suggest such a thing? That’s government.” I’m like, “Listen, let me draw a little chart. Here are the things they did and healthcare.gov. Right, big upfront design, no customer, no iteration…”

Sonal: RFP, multi-stakeholder consensus, blah, blah, blah.

Eric: <crosstalk> And let me show you that chart for what they did and what you’re doing. So, what’s the difference?” And they get mad at me. But I said, “Look, the truth is, that system of managing work is not a good one for our time. Maybe it made sense in a different era, but it really doesn’t make sense now, and there is a better way.”

Impact of software

Sonal: What you’re really getting at — and I don’t think this is as evident to people who aren’t necessarily inside the software industry — is that, in a lot of ways, lean startup is almost synonymous with the world being eaten by software. Because it’s really about a mindset for how people move fast, have a certain methodology, the ability to democratize, as you talked about, the ability to be agile — whatever all those adjectives are, they actually have meaning. They’re buzzwords, but they have meaning. Tell me more then about how software reinvents the firm as a consequence of this.

Eric: It’s really interesting, because — I’ll tell you two stories I think are actually — one, I was talking to a company that was really struggling with its agile transformation. And forget lean startup, they’re still trying to get their software that they write. They employed lots and lots of software developers. They’re trying to get everyone to go agile. And I was meeting with the folks at their — in the software part of their business. And they said to me, you know, “We’re really having a hard time getting the non-software functions in the company to do agile. We’re doing pretty good in the engineers, but like getting people in the hardware, manufacturing, supply chain, but also HR, finance,” they’re like, “Straight up leadership managers to understand agile,” they’re like, “Like, I don’t want to do some, like, software thing.” Lean startup has created a neutral terrain where different functions can come together to do this. And I think there’s just a natural resistance to doing a methodology from someone else’s function.

So, like, if you ever tried to get software engineers to do design thinking, or try to get non non-manufacturing people to do lean manufacturing, you know, you try to get non-operations people to do DevOps, it’s like people are gonna be like, “That’s not my thing. That’s their thing.” And what are they trying to do? The lean startup is a neutral terrain. It is, you know, not associated with any one specific function and therefore — and it’s denominated in terms of business results only. So, people talk about, “Oh, we show people in finance our burndown chart, and they can see how fast we’re making this in.” People in finance don’t care, like, “Well, how much money am I gonna make?” And, you know, we have this thing in a startup called innovation accounting, which is a formal methodology for translating what we are learning about customers and our business plan into financial performance results that give us leading indicators and confidence about the future. It’s a very important part of the method.

So, that’s one thing is — is that although this is software enabled, it is not a software-specific thing. But the point you made that is exactly right — and this is true for every kind of software or semiconductor-related change — it really is about mindset more than tools and materials. And I’ll tell you a funny story. I was working with a consumer electronics company, and they were building this new device. For their confidentiality, I won’t tell you what it was. But whenever I work with hardware — I’m a software guy by nature. I grew up in my parent’s basement programming computers. Like, that’s me. So, whenever I deal with hardware things — because I’ve worked with, you know, the GE’s of the world, the Toyota’s of the world on big physical — if it can explode, I tend to be very humble. You know, I was like, “The nice thing about software is it doesn’t tend to explode.” I always appreciated that. So, I was working with this team, and I said, “Gosh, it’s probably going to be very hard for us to build a product.” We had to build a minimum viable product. Instead of building 10,000 units or 100,000 units, what would it take to create 5 or 50 units, or even 1 unit?

And I was like, “Gosh, that’s probably gonna be really challenging.” And of course, whenever you have engineers in the room, once you frame the problem correctly, they’re like, “Oh, that’s no problem. We’ve actually already done that. We were just playing around with some 3D printers and soft tooling. We have a prototype of it in our office right now.” Well, then the problem is, how do we find an MVP sales channel? You’re not going to get Walmart to carry some unknown prototype device. We’re probably gonna have to find a local store. So, I start to, like, “How can we find a place we can get…” They’re like, “Well, actually, we run a model store in our company, you know, where we can showcase new technologies for customers. So, customers are in there 24/7, you know, looking at new things.”

And I said, “Oh, well, the problem must be, then, that that store is really far away.” I’m like, “How do we get access to the store?” And they’re like, “No, it’s in the same building where we work.” I was like, “Okay, is it like on a different security system? You don’t have the right badge to get down there? Like, is it a different team that operates?” They said, like, “No, we operate ourselves.” And I was like, “Okay. Do we need, like, a dolly or something to move the thing the hundred yards from your office to the model store?” And they’re like, “No, we could just pick — it’s heavy, but the four of us could easily carry it there.” And I was like, “Okay, timeout. You have everything you need. You have the <inaudible>.”

Now, if you look at that story, why do they have a model store? Why were they able to produce this prototype? Like, there’s software lurking in the edges — “here there be dragons” in that story, in many places, but it’s not really about software. It’s about the fact that the company has the capability to work in this new way, but it had never occurred to them to do it. When I said, “You should take this thing out of your office, walk 100 meters to the store and offer it for pre-sale to customers,” They were like — they thought I was crazy. They looked at me, like, with wide eyes, like, “What are you suggesting? That seems nuts.” But then we really walked through the method and walked through the, you know, reasoning from first principles about why that would be a good idea, and it revolutionized their business. And they eventually got to a place where they can show — they can iterate on this so quickly now. They can build a new version of that device every week. Used to take them three to six years to build a new model, now they can do a new version every week. So, they’re constantly getting testing and iteration with customers. And, like, that’s not gonna scale up to a million units, but that could get us — you know, it’s like, well, if customers don’t ever want to buy the thing, we’ll never have to scale it to a million units. We just saved ourselves a lot of costs and time and energy.

Sonal: So, like, what’s the high-level moral of that story?

Eric: So, there’s a couple things. One is just people use it as an example, I think, to see how a minimum viable product thinking can work, right? So, reduce scope, reduce the number of customers affected, try to figure out — what is that experiment that can help us learn whether our strategy is actually right? And in this case, what they learned was that what they thought customers wanted was wildly different than what they actually wanted. But the other moral of the story is that this, before it is anything else, is a management issue. It is not a technology issue, it is not a process or tools issue. It is really about “how do we manage people?” Most companies have all the raw material they need to work in this more entrepreneurial way — including, by the way, the actual people that you would need to act in this creative way. I mean, I’m amazed at the caliber of people who work at these companies who are being told what to do.

Sonal: You’re right. And in fact, I think people are born and really do have — I always think of this analogy of children always coloring when they’re little.

Eric: That’s incredible.

Sonal: Like, who beats the coloring out of them? I mean, now there’s actually this trend where there’s all these adult coloring books, which is something in and of itself. But the point I think is that that creativity never dies. It never dies.

Eric: Yeah, that natural creativity. I’ll just tell you one more story. It’s such a cliché, and I hate even saying it — that we’re going to unlock the creativity of our people. But I’ll tell you this story. You tell me how to describe this, that [it] doesn’t sound cliché. I was once sent to one of my workshops. A 25-person team was sent by their company. It was the true multi-headed hydra of the most despised functions in corporate America. It was a joint finance and IT committee, tasked with creating a new finance IT system that would be a new global standard for how this giant corporation would do…

Sonal: Oh, my God, it’s not even just the entities you’re describing, it’s the fact that there was a task, and a committee, and a standard. I mean, there’s just a lot of crazy stuff in there already.

Eric: When I talk to startup-y and product people about the story, they start moaning and groaning before I even get to the — I haven’t even got to the setup, let alone the punchline, and they’re like “Oh, God, right, you know exactly…” And I was like, these people were not happy to be in this workshop. And they’ve been sent to this thing. And when I went there, I was like, “Okay, if you wanna do this, you gotta think like a startup. You’re gonna adopt a customer service mentality, and really, like — you got to understand that the people who use this product are your customers, even though they’re employees of the company.” And I thought I was gonna be burned alive. I mean, the looks I was getting from people were just like, “Who is this kid telling us what to do?” But we did the work, and we went through the method. Like I said, I think being able to derive lean startup from first principles is very helpful.

You know, accepting people’s skepticism is natural, and being able to walk them through — look, be skeptical, but what is the experiment that could demonstrate. After three days working with this team, they were totally transformed, and they changed their plan from this — their original plan was, like, this huge committee, they basically were gonna spend 18 months gathering requirements. Hand those requirements off to all these implementation teams around the world, who would spend another 18 months doing the implementations.

Sonal: Healthcare.gov approach basically.

Eric: I mean, it’s just healthcare.gov all over again, right? It’s like, “Listen, first of all, one of my rules is, the laws of physics are required — everything else is optional.” So, the word “requirement” just doesn’t apply to, like, a giant worldwide focus group of random things that customers ask for. Those are not requirements, those are hypotheses. Those are guesses about what customers might want. So, instead of doing this big global thing that’s gonna take three years, there’s gonna be no accountability, right? Because the committee will disband, the things won’t be done correctly, there’ll be no productivity savings. We’ll be in the same IT finance mess we’re in now.

They took a different approach. And they decided, on their own accord, to condense down to a five-person dedicated cross-functional team. So, no 25-person committee. They went to their customers, the P&L leaders of the different businesses in the company, and they made them an offer. They said, “Whoever says yes to this offer, the next day, our whole team is on a plane to your headquarters, wherever you are in the world. We are gonna sit with your people and build the software live, before your eyes, and we will show it to you every month, or every couple of weeks,” I don’t remember the sprint iteration cadence right now, “But we’ll show it to you periodically. And when you voluntarily decide to adopt it, you adopt it — so, no corporate mandates. We will show that it’s better than what you have now, and we will not leave. We will keep iterating the software for as long as it takes to prove to you, P&L leader, that you have an actual productivity improvement.” So, not “we met the requirements and now good luck,” but we will keep measuring how much work is actually done in this function — and it’s kind of complicated what the thing was, but we had the metrics to prove that it’s a good idea. And then and only then will we take it to a second P&L, a third P&L, and then eventually scale it up to the whole company.

Sonal: Right. This goes back to your point about the methodology for scaling, and that being one of the biggest challenges.

Eric: Yeah, exactly right. One of my — someone told me that the easy way to remember it is — think big, start small, scale fast. It’s, like, that’s really that is — so, like, prove that it works at scale X, then prove that it works at scale 2X, and just repeat until you have the whole thing. These guys transformed into honest to God entrepreneurs, every bit as enthusiastic and creative as the people I meet, you know, in Soma every day.

Sonal: Right. So, even though it is really cheesy, I agree, to say like, “Unlocking human potential and creativity,” I do think it’s a really important argument for a future of a world — because when we talk about everything becoming software — like, companies changing, everything changing. For a world where more things are getting automated. And being able to really have something to contribute as a human being with judgment and creativity, something you can’t codify into a program.

Eric: The debate over robots stealing all the jobs and everything, like, has baked into both sides of the disagreement. This fundamental premise that work is boring.

Sonal: Yeah.

Eric: And so, like, should we let robots do the boring work, or should we let humans do the boring work? And it’s like, no, no, no, that’s just evil. I cannot buy into that idea at all. First of all, work that is monotonous and routine should be automated, because every human being has a right to use their creativity in their job. This is a lesson going all the way back to the Toyota Production System of years ago. Even on the factory — like, the canonical job that was supposed to be — all creativity sapped out by Fred Taylor back in the days, right? Someone just doing the repetitive, you know, stressful work on the line.

Even there, human creativity can work to our advantage. How much more so in knowledge work and in management, and all these kinds of systems we’re talking about. So, to me, this is saying these companies have locked up a massive amount of human potential. I think the scale of this, we are only — it’s hard even to fathom, because these companies are so large, and so many of the people are trapped in systems that prevent them from exercising their independent judgment and creativity. I mean, that’s just — like, people debate whether that could be changed or it’s a law of nature, but it’s a fact that is happening today. And what I have seen is that you can change it. It’s hard not to sound utopian about it, but I really think it’s going to have a profound impact.

Michael: I just wanna ask — the distinction, though, between — does that mean everyone needs to be an entrepreneur or entrepreneurial?

Eric: Yeah, that’s a great question. It’s a source of great confusion, because the companies that have adopted this system formally. They’re usually, like — even lean manufacturing. The Toyota Production System, it’s not called lean manufacturing, it’s called the Toyota Production System. Every company makes it their own. So, like, at GE, they have this program called FastWorks, which is their version of lean startup. At Intuit, they have a program called Design for Delight. Everybody has their own version of it. And one of the questions they get all the time is, “Well, does this system apply only to special projects, or does it apply to everyone?” And every company has had to deal with this duality, where they say, “Actually, there’s two versions of this. There is the version for, like — when you wanna make a big bet — disruptive new product. And I think we’re talking about as a startup, as an atomic unit of work. It’s like one of the reasons why Amazon is so effective, they have the two-pizza team rule, and they can say, “That’s a good idea. Let’s throw a startup at it.”

Sonal: Right. Just to clarify that two-pizza team rule being that the team should be big enough in size to only be fed with two pizzas.

Eric: No bigger than you can be feed with two pizzas

Sonal: Exactly.

Eric: It’s a basic way of saying small teams, small teams can try things. So, they have like a — it’s a tool in the management toolbox to throw a team at something and not let it turn into a big sprawling committee, but keep it focused, keep the people cross-functional and dedicated. No multitasking, no passing work between silos. That is the death of innovation. But then you also have, like — at GE, they call it FastWorks Everyday. And they say, “Look, no matter what work you’re doing, even down to…” They always give this example of, you’re preparing a PowerPoint presentation for a meeting. Even that very simple task, you can ask yourself, “Who is the customer for this task? And is there a minimum viable product version? Is there a way to test and experiment? How do I really know that this work is valuable?” And the number of people in corporate America who do work where I asked them this question, I say, “Listen, what is the evidence? How do you know that the work you do every day matters to anybody except your boss?” I used to think people — everyone would just be like, “Oh, of course, I know.” And a number of people who are like, “I don’t know, I just assume.”

Sonal: We’re just kind of moving along like zombies in the workplace.

Eric: And you’re like, “What kind of job is that? Like, how are you going to feel at night when you go home, saying, ‘Gosh, I hope I accomplished something today?’” Versus now, like, imagine a world in which everybody knows it in their bones, because they had the scientific rigor to always be testing. And when you see that, I mean, I’ve seen the before and after photos of the people who’ve gone through that transformation, and it’s truly powerful. Even in places, you know, I was thinking about like a factory where we did this transformation. And we’re talking to, like, the union reps for the people in the factory, and to see them go through the transformation. Even places where I think we have a prejudice, they’re like — like, in certain cultures, certain places, certain functions, those people would never get it. I’ve seen it everywhere that you can see it, and it’s a wonderful thing to see.

Implementing lean startup

Sonal: Everybody gets it. So, I wanna wrap up and then, kind of, revisiting the question of, how do people adopt this amazing movement and mindset — as we’ve said, it’s important — and the tools that come with it — without veering into cult territory. Where they start holding onto, like, the letter of the rule, versus the principle behind the rule. Because that seems to be a phenomenon that happens with anything. But also we’ve observed [it] happening with — even with things like lean startup. So, how do you sort of…

Eric: Oh, certainly. I mean, I hear — many of my VC friends complain, like, about lean washing, they call it. And it’s like the same old crappy…

Sonal: Oh, lean washing. That’s so funny, it’s like green washing.

Eric: …the same old crappy venture pitch, but now it’s lean. It’s, like, lean crappy venture pitch. And I’m saying, “Look, first of all, do not, please — if you’re listening to this, if you’re trying this at home — do not use the terminology to, like, dress up your dumb idea,” okay? In fact, I don’t care if you use the terminology at all. In fact, I have a secret product managers, like, meetup group that I do for one of the big tech companies, where we get together — it’s secret so I can’t name the group…

Michael: What’s the name of the group?

Eric: Yeah, I can’t tell you. I can’t even name the company. Yeah, because they work in a company that has a real strong “not invented here,” culture, and lean startup is forbidden. So, if they talk about minimum viable product, or pivots — like, inside the company is just like, “Do not want to hear about it.” So, what they have figured out is, like, they have to create a company-specific vocabulary version of it. So, we meet periodically to talk about, like, “How do we get people to adopt these concepts without…”

Sonal: It’s like putting code names around it.

Eric: Yeah, because, like, to be honest, what matters is that you have a precise and clear language that you can communicate with your co-workers about. I don’t care if you use my language or somebody else’s, as long as it has a rigorous foundation. And the people who argue about — like, I periodically get a phone call from someone and they’re like, “So and so is blogging a bad thing about lean startup, like, they’re using it wrong. You need to make them stop.” And I’m like, “I’m not the Pope. I can’t excommunicate anybody. Like, what power do you think I have to make somebody stop?” Like, that kind of, like, inside baseball…

Michael: “Hold on, let me get the stone tablets.”

Eric: Yeah, it’s like, chisel an extra — you know, and it’s funny, because I said in a recent talk, I was like, “Listen, lean startup stands for what works. So, if I said something wrong — if we discover ourselves as a movement — we use our own scientific process to discover new things. They obsolete the old things. And we should always be getting better, and what we wrote five years ago should always seem a little bit creaky, or we’re not learning.”

Sonal: To close it out then, let’s just take the two most popular terms, which are also often specifically overused — and I think, therefore, have a lot of misunderstandings around them. Pivot and MVP. And I’d love to start with MVP, or minimum viable product. Because one thing that I’ve seen on the other side is that people sometimes mistake doing an MVPm when sometimes the big visionary ideas require you to over-investm and not under-resource a minimum viable product. So you, actually — sometimes you don’t wanna— so, I’d like you to talk about that.

Eric: Yeah, Oh, I’m happy to talk about that. It’s really interesting, because people really love the bumper sticker version of lean startup. Okay, there’s certain concepts that just people really gravitate to. Pivot, MVP, continuous deployment, you know, and some of the, like, lean buzzwords are people all excited about. But if you actually read the book — which I think a lot of people who are doing this buzzword Bingo haven’t actually done — if you read the book, the vast majority of it — the bulk of it by weight is not about, you know, this jargon, but it’s about the management system. It’s about the innovation accounting. It’s about the math that underlies this way of working. And you know, I get that accounting doesn’t make a good bumper sticker, and we can’t cram the math into — so, I understand why the slogans get the attention. That makes sense to me.

But one of the problems with that is people who have a kind of a flip understanding don’t really have the context to do it correctly. And minimum viable product, or MVP, that is a pretty common problem. The core mistake I think people make is they think that it’s a minimum product, right? Like, if you’re trying to do something small in this world, you don’t need to do an MVP, you just do it. Like, if the thing itself is cheap and easy and has an obvious application, and…

Sonal: Especially if it’s software.

Eric: Yeah, you don’t need to test it, you don’t need to experiment, just ship it. It’s only people who have a consequential vision that this is a good idea for or is necessary for it all. In fact, the first part of my book is called “Vision.” Because people think like, oh, taking a scientific approach to innovation means taking the creativity out of it. Which I think is very insulting to our friends who are scientists. I’m sorry, I think science is one of humanity’s most creative pursuits. And someone once said to me, “If you could turn entrepreneurship into a science, then everybody could do it, and that wouldn’t be good.” And I was like, “First of all, science is a science, and very few people are good at it, okay?” It’s actually, like, it’s really hard. With all due respect, like, some of the smartest people who’ve ever lived have been our scientists — and not just smartest — our most creative people. So, science is not just turning a spreadsheet, right? It really is having those leaps of insight to form good hypotheses in the first place and that is what — and all MVP is saying is, take those hypotheses and put them to the test in a cost-effective way, so that we can sustain our investment in the vision over time. 

Some visionaries in this world are independently wealthy, and can sustain investment in their projects for as long as they want. Some have the knack for raising money, just on the strength of their personality, and some can even take a company public and resist the temptations of Wall Street and just do what they — I mean, there are some people who have that ability to sustain their vision indefinitely, but most people do not. Most people, even the ones that have very, very powerful visions, struggle with, “How do I command resources? How do I get resources put to me?” And they feel, frankly, a lot of resentment about the just huge amount of bullshit that is required to do what we call success theater — putting on a show about how good your thing is doing, so that you keep attracting investment. And I think every ounce of energy we force visionaries to put into success theater is an ounce of energy they didn’t put into making their vision a reality, they didn’t put into serving customers.

Sonal: Just making it happen. Right.

Eric: So, I think of lean startup, and MVPs in particular, as a way to demonstrate progress towards the vision to sustain that interest overtime during the long flat part of the hockey stick, when the vanity metrics are really low and there isn’t instantaneous overnight success.

Sonal: So then, the last word — pivot — completely overused. I mean, it’s actually just, God, we waited so long to even say it on this podcast.

Eric: Yeah, I mean, I have to, like — I’m a little bit embarrassed now almost how much it’s become an overused buzzword. And I saw this — there’s this cartoon, you can Google it. There’s a cartoon in the New Yorker magazine a little while ago, where there’s a man and a woman sitting in a café, and she says, “I’m not leaving you, I’m pivoting to another man.” I just thought, “What have I done? I’m sorry.”

Michael: No. Well, it is — I mean, you called it a movement when we were talking about it, and, I mean, that’s kind of part of it, right? I’m sure people have “pivot” tattooed. That’d be a weird tattoo to get.

Eric: I’ve seen some pretty odd things.

Michael: Yeah, but it is part of this, kind of, I don’t know — language and culture and fascination with what you’ve done with lean startup.

Eric: Yeah. You know, first of all, I don’t apologize for the fact that pivot is a very useful concept for startups. In fact, you can go back and read stuff that was written about startups before that word came into the common vocabulary. And people struggle to explain this weird phenomenon, which is — it seems like the great entrepreneurs persevered through everything, and they, you know, stuck to the vision no matter what. And yet, they also were super flexible about certain details. So, like, that’s odd. Like, “What do you do?” And so, like — for the people who are new to the concept, the right definition of a pivot is a change in strategy without a change in vision, right? So, the vision is our true north, our destination. But the specific strategy we’re gonna be like, “What is the business model? What kind of product is it? Is it software? Is it hardware? Is it a device? Is it service?”

I think about, you know, the Google search appliance and the pivot to AdWords, right? That didn’t give up on the vision of organizing the world’s information, but they said, “That’s a dumb business model, and this is a good business model.” So, like, that’s okay. But if someone had said, like, “Let’s get out of the search and information business and just sell cars,” you know, they would have been like, “Oh, that’s not a pivot. That’s abandoning our vision.” Although, of course, now they are gonna sell cars, so the vision expands as you have success.

Sonal: Actually, I think it is still connected to all the world’s information, right? Especially as cars become moving computers.

Eric: Oh, yeah, well, exactly. So, you know, vision is personal, and it’s deep in the minds and the souls of the founders and eventually into the whole company in its DNA. So, from the outside, it can be a little bit hard to understand, like, what really is the vision and what are the incidental bits? Like, I think people, you know, certainly would have thought, at a certain point, that Netflix is all about sending you DVDs by mail. And when they first — I remember when people — it was very controversial when they started to become an online streaming service that would actually have fewer movies and less options. It seemed like it was getting worse. Like, it’s not an abdication of the vision and then…

Sonal: And now to programming their own content.

Eric: I mean, yeah, so you never know where — you know, you have to be flexible about the specific strategy, but you have to be willing to invest in and stay true to the vision. So, that really is why the concept of pivot is so important.

Sonal: That was the origin of the word.

Eric: And the reason it’s such a critical part of lean startup is, if we can get to the moment of pivoting sooner, cheaper, faster, it’s like magically extending the runway of the startup without raising more money, and that’s why it’s such a powerful idea.

Michael: It’s a movement. It’s not a cult. It’s gathering momentum, you know, outside of Silicon Valley and beyond. And we’ll try and extract the information on your secret product management group later.

Sonal: I’m gonna start, like, tailing you and driving following you everywhere to see where that goes.

Michael: Thanks so much for coming. We’ll look forward to the next book.

Sonal: Thank you, Eric.

Eric: Thank you very much.

• 

  Eric Ries
• 

  Sonal Chokshi is the editor in chief as well as podcast network showrunner. Prior to joining a16z 2014 to build the editorial operation, Sonal was a senior editor at WIRED, and before that in content at Xerox PARC.

  Follow Twitter
• 

  Michael Copeland

Disruption is such an overused buzzword. But the word itself does have meaning: As defined by the Oxford and Merriam-Webster dictionaries, it is a “disturbance…that interrupts an event, activity, or process” and that causes something “to be unable to continue in the normal way.” It’s also the name for an influential theory about innovation first coined by Clayton Christensen in a 1995 article and later publicized through his 1997 book, The Innovator’s Dilemma.

But that was nearly two decades ago! Not only has the concept been much misunderstood and mangled since then, surely it’s changed given the advent of new tech and business models today? Is it still relevant, given cases that seemingly defy the theory and its application? Are we at risk of overfitting this “verbally inflated” term to everything, and in doing so, are we missing what disruption theory really says — and doesn’t?

Michael Raynor, co-author of the followup book on disruptive innovation with Christensen — and author of another book that later tested the predictive power of the theory — joins this episode of the a16z Podcast, in conversation with Sonal Chokshi, to answer these questions and more. He also hints at some nuggets from an upcoming article in Harvard Business Review with Christensen and others that addresses the latest formulations of this theory of innovation.

Show Notes

• What disruption theory is, and how it’s been misunderstood [0:34]
• Popular examples of “disruption” that do not conform to the formal definition [10:23]
• The difference between disruption and innovation [24:42]

Transcript

Sonal: Hi, everyone. Welcome to the “a16z Podcast.” I’m Sonal Chokshi, and today we have as a guest on the pod Michael Raynor, who is a director at Deloitte Services and a co-author on one of the seminal books on disruption, with Clayton Christensen, “The Innovator’s Solution.” Raynor also later wrote a book called, “The Innovator’s Manifesto,” which is one of the very few works out there that actually tests the predictive power of disruption theory. We invited Raynor on the podcast today, since he, Clayton Christensen, and Rory McDonald have a new paper coming out this December in Harvard Business Review, defining what disruption theory is and what it isn’t. Welcome, Michael.

Michael: Thanks. It’s great to be here.

Defining “disruption”

Sonal: Thank you for coming. So, actually, why don’t we just jump right in, and let’s just start with talking about what disruption is. And I think it’s come top of mind, because there’s been a lot of articles written in the last year that, sort of, slam it. And, to be clear, I think people are slamming both the word, because it is very overused, and I think part of it is also slamming the theory, which is what we’re talking about here — disruption the theory.

Michael: Yeah, I mean, as I read it, I’d say it’s even more than part of it. I mean, there have been a couple of critiques that have come out. There was a piece just over a year ago in the New Yorker that was very pointedly about Clayton and disruption theory. There’s a piece [that’s] come out recently in the Sloan Management Review that’s similarly, sort of, an attempt to look at the theory and say, “So, here’s where it goes too far, where it makes mistakes, where…” And I think that’s important, actually. I mean, that’s how we progress, right? Constantly saying we got it right the first time doesn’t take you anywhere new. Conceptually, at least, it’s easy to embrace that kind of give and take, that kind of discussion.

Sonal: Yeah, no, I think that’s great. What was the gist of the critiques? I mean, I agree that conceptually, it’s important to have that kind of — but what were people, sort of, slamming about the theory?

Michael: Yeah it’s — I guess, in the first instance, they were going after a phenomenon that I think is definitely something worth trying to, I’ll say, combat at the risk of overstating the case. Which is that the term “disruption” I think has come to be used far too frequently with far too little precision, and I think that’s unfortunate. I mean, in the first instance, disruption is a well-formed English word. You can look it up in the dictionary, right? But when people use it, especially when it comes to — in a business context, they forget what the word actually means. So, disruption means to hold up something, to slow it down, to interrupt an otherwise smooth and even flow. So, a disruptive student is not something that makes class better. That’s someone who makes class worse. A disruption in subway service doesn’t help you get where you want to go better or faster or cheaper.

Sonal: Yeah, you’re right. It doesn’t have a positive connotation at all.

Michael: It’s just a bad thing, but people that’s not what people say. People say, “Oh, you know, this company is disrupting things.” And that’s supposed to be good. So, when they say that, they’re invoking — whether they know it or not, they’re invoking the connotation that Clayton gave to the word, with disruptive innovation and disruption theory. But unfortunately, they’re using it, then, in a way that is, to my way of thinking at least, more often than not entirely disconnected from the specifics of what disruption theory actually describes. So, we’ve run into a circumstance where people use the word very frequently to describe all manner of phenomena. And the downside of that and this — and here’s why this matters — the downside of that is that as a consequence of that verbal inflation, we actually lose our grip on the power and the insights that disruption theory brings, and that would be a shame.

Sonal: I totally agree. And I think that’s actually a great turn of phrase — verbal inflation — and one could argue it’s a disruption bubble. When Clay originally coined the phrase — used the word “disruption theory” — like, what was the intent behind the original meaning?

Michael: So, his first book came out in ’97. It was called “The Innovator’s Dilemma,” and it started slow. It came out in ’97, sold, you know — I’ll get these numbers precisely wrong, but close enough, close enough. Sold a few thousand copies over the next couple of years, and then it exploded in 1999. I think it was the January issue of — I think it was the cover of Forbes, and it had Clay with Andy Grove. Caption read, “Andy Grove’s big thinker.” And Clay is 6′8″ for those people who haven’t met him, and Andy Grove is probably, I don’t know, 5’6” so it was a clever turn of phrase.

In any event, then the book took off. And what that book described is a particular class of phenomena whereby companies are able — small, under-resourced startups, very often — are able successfully to enter markets that are dominated by well-managed incumbents. And so, that was a puzzle. How is that possible? How does this small, you know, scrappy, little upstart — how is it able to successfully overturn a successful incumbent? And so, Clay chose the word “disruption” to describe that phenomenon. And it captures half of what’s going on, right? The disruption is to the incumbent, and it describes a very particular pathway by which a startup and a new entrant, more generally, is able to enter an established market.

Sonal: Right, and if I remember correctly and, actually, I remember learning this from you a few years ago. It’s, sort of, the startup comes, in traditional disruption theory, from the lower end of the market, usually with either lesser features, or just reaching a niche customer set that no one is otherwise reaching. And then the other part of it, the second part of it, I remember, was really key — is that there’s some kind of accelerator through technology that then drives them to be able to go upmarket.

Michael: So, this notion of what you call an accelerator, what I refer to as an enabling technology — what Clayton has referred to as an extensible core. So, the fact that the language hasn’t quite settled down shows that this is a relatively new addition to the theory, but I think a critically important one is one that a lot of people have walked past and fundamentally ignored.

Sonal: Right. Are they able to ignore it because they’re just, sort of, fitting the theory to everything? Or is it because you don’t actually require that accelerant in order to reach that sort of quote disruption?

Michael: No, I think it’s a necessary condition. You don’t have that, you don’t have a disruption. So, you pointed to a couple of things, which is that a disruptor starts at the low end, a niche market. That’s a defining feature. There are really three necessary and sufficient conditions, right? The first one that we pointed to was where you start. So, first and foremost, disruption theory is a theory of customer dependence. You tell me who you’re selling to, and I’ll tell you whether you’re embarking on a potentially disruptive trajectory of innovation. So, disruptors start in segments of the market that incumbents aren’t motivated to fight for, or fundamentally don’t see. So, we refer to that as either the low end, or an entirely new market competing with non-consumption. So, that’s step one.

Second is, you have to have a fundamentally different business model that allows you to serve profitably the niche that the incumbents don’t want. There’s a reason the incumbents don’t want to serve those niche markets — because they can’t do it profitably enough, right? And so, you have to come up with a way to serve those segments profitably. If they lose money there, and you say, “Well, I’ll go lose money there too,” that’s not gonna be a disruption. That’s just losing money. So you have to have a different way of serving those segments. Then the last piece is this enabling technology. It’s something that allows you to now take that same business model and begin to serve the mainstream markets that the incumbents do care about. But now it’s too late, because the incumbents can’t respond, because you have broken the trade-offs that they were depending on. The trade-offs that made it impossible for them to serve the low end, you have now broken.

Sonal: So then what has changed today? Because one of the observations that I have, and some of this is definitely anecdotal, is that there seems to be — it seems to be happening a lot faster, for one thing, because of “software eating the world.” There are arguably new patterns. I mean, I’d love to hear your thoughts on this. 

Michael: So, I guess what I’d say is that when it comes to things happening faster, that speaks to the rate of change in the underlying enabling technology. So, if you look at disruption in the steel industry, right, how long did that take? So Nucor is the archetypal disruptor in the steel business. Where did it start? It started with rebar, which is a low-volume, low-margin segment of the steel business that incumbent steel makers were not motivated to defend. Nucor built a fundamentally different business around the mini-mill, and then it took 43 years for Nucor to become the same size as some of the largest integrated mills in the U.S.

So, why did it take 43 years? Well, it took 43 years. What was Nucor’s enabling technology? Well, it was electric arc furnaces and continuous casting. And those big iron — both, literally and metaphorically — those big iron technologies improve relatively slowly. They improve on a mechanical clock speed, and so it took 43 years. And then you look at disruptions in the tech space, and you say, “Well, what about the personal computer?” So, the personal computer is clearly a disruption to mini computers and mainframes for all the same reasons. Started as toys, sold to hobbyists that couldn’t do anything. What was the enabling technology there? Well, it was the microprocessor, and the microprocessor — that gets better pretty quick. The difference is not the underlying phenomenon, nor the theory — it’s an empirical observation. Which is, how fast does the enabling technology get better? That will tell you how quickly it will break the trade-offs that preclude it from serving mainstream markets.

Sonal: Actually, I’ve actually heard from Alvy Ray Smith, the co-founder of Pixar, that they used that exact formula in their head to actually then map out how they would intentionally disrupt the making of animated films.

Michael: Sure.

Sonal: Because they were able to actually use it to, like, almost predictive power in a sense.

Michael: Part of the reason that I find disruption theory so powerful is that now when people say, “Well, this is completely different because it’s so much faster,” I’m like, “Actually, no, it’s not completely different. It is a quantitatively different outcome, but it is qualitatively the same phenomenon.” We can use the same theoretical toolkit to understand what’s happening. The specifics are different — that’s why we play the game — but we can use the same theory to understand and, to your point, predict and maybe even control.

What disruption theory is not

Sonal: So, speaking of prediction, you know, again — you’re one of the few people who actually applied and studied the predictive power of disruption theory. What are some of your high-level findings from that work?

Michael: “The Innovator’s Manifesto” came out in 2011, and it was an attempt to do, as you say, to actually use the theory to predict outcomes. And it’s tricky. It was a lab experiment, and it was done using MBA students largely. I’ve had a chance to replicate it using executives now, and have achieved essentially the same results.

Sonal: Oh, that’s good to hear so that reproducibility… <crosstalk>

Michael: Right, exactly, yeah. Although the executives weren’t too thrilled to hear that they weren’t doing any better than the MBAs but, you know, sometimes the truth hurts. And we did that using a portfolio of businesses that had been launched by Intel over the years. And it was a randomized, double-blind, you know, study to say, “Right, we went to the MBAs and gave them a bunch of business cases and said ‘Pick winners and losers’ and then we taught them disruption theory, and we said, ‘Now, try it again.'” And I’m glossing over all the details that make the findings, I hope, believable. But what we found is that the users of disruption theory improved their accuracy by up to 50%. That said, in absolute terms, we have to be modest. Their success rate was around 10% at picking winners, and it was about 15% picking winners with disruption theory, because it’s a big, noisy world.

Sonal: And you also have a sample set that’s an internally captive VC arm, essentially, a venturing arm that’s inside a company.

Michael: Right. All kinds of delimitations. You know, my experiment, you know, like, every other has its share of imperfections. But to your point, it was an attempt to actually try and take seriously the notion that the theory can be used to predict, and I found the findings encouraging.

Sonal: We do have a tendency. You know, I think the predictive power matters, because there are entire businesses built upon this theory, and some of them which have become incredibly successful. One question we have is — there’s a tendency to, kind of, equate technology means disruption. Just sort of, you know, to over-apply and overfit the phrase to everything. So, what is disruption theory not? Like, what’s not disruption theory then, to help people kind of understand what it is?

Michael: That’s a great question. In fact, both Clayton and I and another professor at HBS named Rory McDonald have a piece coming out in the December issue of the Harvard Business Review that tackles that.

Sonal: Oh, give us the early preview.

Michael: Yeah, exactly.

Sonal: That’s why I want you to tell us all your secrets.

Michael: Sure. Well, I’ll give you an example. It’s — and I’ve asked this question at various conferences and workshops I’ve been part of — I ask for a show of hands. How many people think Uber is disruptive? Every hand in the room goes up.

Sonal: Ours included. One of ours went up.

Michael: Yeah, exactly. And it’s not. In fact, it’s the…

Sonal: So why?

Michael: Well, let’s review the theory, right? I mean, disruption theory is, first and foremost, a theory of customer dependence. Whom are you selling to? So, whom did Uber sell to? Was it selling to a niche of the market, the low end of the taxi market that established taxi simply couldn’t be bothered to serve? Was it selling to people who found hailing a cab and paying for it so inconvenient and so expensive that they just had never used cabs before? No.

Sonal: So, it’s not capturing that different consumer market, right.

Michael: They were going after, and continue, for a large part of the business, to go after folks who want a cheaper, more convenient, cleaner, nicer cab ride, right? There was an article in Businessweek — again, I think I’m remembering this largely correctly — and it was stating that Uber had gone from — again, close enough, close enough — 350,000 rides a month in Manhattan to 3 million rides a month in Manhattan. And over that same period of time, what do you think the drop off in yellow cab rides was? Son of a gun, about 3 million rides a month.

Sonal: So, ride-sharing is clearly a huge market, but you’re saying that because they’re competing with the same exact customers as the taxi industry, it doesn’t count on that one criterion, so far, as disruption.

Michael: Well, exactly. And so, remember, what disruption describes is a pathway — a particular way in which a small under-resourced entrant can succeed against well-managed, dominant incumbents. So, it’s a pathway. It’s not a description of your impact on the established market, which is how people have tended to use it. Say, “Oh, Uber is disruptive because it’s turned the industry upside.” Well, it has revolutionized the industry. It has had a huge impact on the industry. It is not…

Sonal: But it’s not technically disruption.

Michael: Well, but you say that as if somehow it were a minor distinction.

Sonal: No, right.

Michael: Yes, it’s not technically — it’s not disruption and that matters, because if we think it’s disruptive, then other folks who want to pursue a disruptive strategy will think, “Well, I need to do what Uber did…” What did Uber do? Uber did something that, to my mind, at least, is a fairly long-odds proposition. Which is, they just built a better mousetrap.

Sonal: So, that’s interesting, because one of the theses that one of our partners put forth a couple of years ago is something called the full-stack startup, which — you know, he sometimes jokes about how he regrets even calling it that, because it’s sort of like an analogy…

Michael: You should talk to Clay about regretting having called disruptive technology.

Sonal: Oh, I know, right. Clay is probably the one who has a lot of regrets around those things. But the way Chris Dixon articulates the thesis is that, in the past, companies like Lyft and Uber would have tried to build software and then sell it to the taxi industry. But there weren’t even people in the industry who could even have the skill set, let alone to appreciate the software — to evaluate that software and actually say, “Okay, this is what’s going to help us with the problem we have.” Nor were they incented to solve for that problem. 

And so, he argued that instead of trying to go down that path, there’s been a new wave of startups that’s actually been able to “disrupt” — and, yes, I agree, this is not in the technical form of disrupt, but now I’m using it as more of a descriptive adjective — that they’re able to overturn and shake up, so to speak, the taxi industry, because they built something full-stack. Like, from end to end, so they can control the entire experience. And by doing so, they essentially stopped trying to sell their software to the taxi industry and just built an alternative. Like, to your point, a better mousetrap.

Sonal: Yes. They built a better mousetrap, and as Emerson said, the world beat a path to their door. In fact, they probably took an Uber to their door.

Michael: Right. But I do wanna just protest for a second, Michael, because I’m having a really hard time letting go of this belief, and you’re gonna have to convince me a little harder.

Michael: You and everybody else, I’m sure.

Sonal: Right, I am. I’m fighting it. But the reason is because it does feel that — what if it means that disruption theory could be adapted for the software world?

Michael: Well, no, it’s a different phenomenon. And, now, when I say adapted for that — we kind of talked and touched on that earlier. If we are describing a phenomenon in which software is the enabling technology for an entrant on a disruptive path, we’re describing something that starts over on the fringes and works its way into the mainstream, with a fundamentally different business model that is powered by — and this is key — in fundamental improvements in the software over time. All of those things have to be there. And it’s important to understand that, because it feeds into the choices that you make as a manager along the way. How do you deploy resources? What R&D strategy do you follow? What customer segment do you target?

Sonal: Right. Do you build self-driving cars or not?

Michael: These are all things — so it’s important, I think, to underline that — I’m not, sort of, being picky. At least, I hope I’m not being picky here.

Sonal: No, it’s great. That’s why we wanna have this discussion.

Michael: The phenomenon you describe — that one is describing and the meanings that one attributes to these words are critically important, because they determine the choices we make. The way we use the words is critically important because, ultimately, what happens is that everything is disruptive. And when everything belongs to a category, then the category is useless.

Sonal: So then, how has the theory changed? You are already talking about what disruption theory isn’t. Are there any other examples along those lines? And I’d love to hear your thoughts about what has, sort of, updated around here.

Michael: So, I guess there’s a couple of things that I’ve observed that people have — that threatened to lead folks astray. One is, sort of, it happened fast, you know, it’s a big bang disruption. Well, no, we actually don’t need that because we have this concept of the enabling technology, and it’s the rate of improvement in the enabling technology that determines whether or not it’s a disruption. Now, there may be complete transformations of an industry that happened very quickly, for reasons other than the disruptive entry of the startups — and that’s fine. But then we need to be clear, that’s a different phenomenon.

Sonal: Even if the outcome may be the same, actually, in some cases.

Michael: Absolutely. These are, you know, to use a medical analogy, these are different conditions, and you need to get the diagnosis right. So, that’s fine. The other thing that I think leads people astray is the notion of — and you hear this, you know, kind of, top-down disruption — and people will point at Tesla, and they’ll say, “Tesla is a — they’re disrupting the car industry, but they’re doing it from the top.”

Sonal: So, they’re not disruptive.

Michael: They’re not disruptive at all.

Sonal: Okay, so let’s talk about why.

Michael: Well, same reasons, right? Was Tesla targeting a small, unprofitable, unattractive segment of the car market that was of no interest to incumbent car companies? No, they’re targeting people willing to spend 100 grand on a car, which is very interesting and important to companies like Mercedes and BMW and Lexus and, and, and…

Sonal: Right. But it was an underserved market, in the sense of — those folks were not having a car that has software at its center.

Michael: No, no. They weren’t — you think they didn’t have a car?

Sonal: Of course, they had a car.

Michael: Well, then they were consumers.

Sonal: No. Okay, I’m gonna fight this one too. Again, this is different because, yes, you’re right. They would have bought another car. They would have been — they are the typical segment for other car companies, so that’s not a new market in that sense, but they weren’t having — their needs were not being met.

Michael: They were underserved?

Sonal: They were underserved.

Michael: Absolutely. And that’s the sustaining innovation. Disruptive innovations target overserved customers.

Sonal: There we go.

Michael: Customers for which established solutions are too good, too expensive, inaccessible.

Sonal: Okay. So that’s another precision thing that helps us define what disruption is and isn’t.

Michael: So, Tesla goes after a critically important segment of the market, and shows up with, you know, its own version of a better mousetrap and appeals to those they’re willing to buy, and away we go. And so, in fact, if I were to point to somebody that explains the path that Tesla appears to be following, I point to Jeff Moore in “Crossing the Chasm.” As I read it, the way you cross the chasm is that you find very demanding customers, and you create a highly effective solution that solves their problems really well. And then, basically, you ride a cost-reduction curve into the mainstream, right? So, you find the really demanding, early adopters. In a sense, it’s an adaptation of Everett Rogers’s diffusion theory.

Sonal: That’s actually where the whole “Crossing the Chasm” thing actually was hinged on?

Michael: Yeah, so you find those really demanding early adopters, you solve their problem because they’re demanding, they’re willing to pay, you use the profits that you generate from serving those high-demand, very profitable, early adopter customers, and then there’s a lot of things you have to do in order to cross the chasm into the mainstream. That’s completely different from what disruption describes. Disruption describes a very different path from the fringe to the mainstream.

Disruption vs. innovation

Sonal: So, so far, we have Uber and Tesla, which a lot of people — including me, apparently — thought were disruptive and really aren’t. So, are there any examples where the company is actually disruptive but no one really knows it is?

Michael: Well, that I can’t speak to, but you’ve probably heard of Theranos.

Sonal: Of course.

Michael: So, I would look at that one, and here — I’ll probably run out of facts sooner than I should. My understanding there is that they’ve created a whole series of blood tests that are able to give a high level of accuracy at very low expense and very low inconvenience. The way I think about that is that it’s an innovation, because it has broken trade-offs. And something that I think gets in the way, is that when we think about disruptive innovation, we can’t separate disruptive innovation from any other type of innovation, because we don’t have the larger class defined. So, an innovation for me is anything that breaks a constraint.

Sonal: I love that definition, by the way. I’ve actually stolen and used that definition for years since reading your book. I just want to tell people publicly — that that wasn’t my idea. I want to just confess.

Michael: Coming clean after all these years.

Sonal: I actually did credit you, in fairness, but I am gonna say that that is, I think, by far the best definition of innovation I’ve ever heard. I really mean that.

Michael: Well, thank you. We can stop here. So an innovation is anything that breaks a constraint and disruptive innovation is a particular path, right, from not being able to break those constraints to having broken them in the mainstream markets. So, when I look at Theranos, my understanding of it is that their solution right now is kind of — is more for less. They’re having difficulty, I think, finding adoption in, you know, mainstream hospital labs, and so they’re actually finding their foothold, their first commercial applications in clinics and drugstores and relative — essentially, if you will, on the fringes of the core mainstream blood testing market. So, you have something that is — that has broken certain constraints and is following a path from the fringe to the mainstream. What I don’t know enough about is whether there’s, at that core, the enabling technology that is going to allow the “Theranos solution” to, in scare quotes here, improve to the point that it can penetrate mainstream markets.

Sonal: Got it.

Michael: And this is important, because if it’s there already, right, if it’s already more than good enough, right, for those applications, then what we have is not bona fide disruption, what we have is a marketing strategy.

Sonal: Right?

Michael: The need to start at the fringe and move to the middle. And so, the kinds of things they get caught up is people say, “Well, it started small and got big.” Well, that doesn’t make you disruptive. That just means you started small and got big. Almost nothing big starts big.

Sonal: That’s actually a good point.

Michael: Right? And so all of these other characteristics — when people say, “Well, it started small and got big and it revolutionized the industry, therefore, it’s disruptive.” Holy non sequiturs, Batman, none of those things have anything to do with whether or not you’re following a disruptive path. Disruption can be used very precisely, and it describes an important class of phenomena, but it’s not a theory of everything.

Sonal: Got it. So, let’s actually then take on the elephant in the room and talk about — and, again, I don’t wanna make this about Clay. I know we both have immense respect for him. But people often argue that he was wrong about the iPhone. And, I mean, I’ve made the argument. I know others have made this argument, that it was a category error. That he just got the category wrong for what he actually thought it should be when it was something else. What’s your, sort of, take on, sort of, why that did actually apply or didn’t apply in that case?

Michael: Sure. So, I think — and this is a subtle but critically important distinction to make between what I’ll call the cross-sectional problem and the longitudinal problem, right? So, Apple showed up with the iPhone in the mobile phone market with a better mousetrap. Apple did not enter the smartphone market disruptively. And, again, why do we say that? Let’s see. Whom are they trying to sell the iPhone to? People who had phones, right? People who wanted a better phone, right? People who wanted a phone that could do other stuff, right? It’s not as though they were appealing to a niche market…

Sonal: For underserved customers, right, exactly.

Michael: …that established phone makers said, all of them, “Apple can have them. We don’t really want those customers anyway. Who needs 18 million more customers?” Of course, right? So, they were selling to customers. And Clay, I think, was absolutely correct in the way in which he applied the theory. He said, “Look, the data say pretty clearly that if you kind of walk into a bar and punch the biggest guy there, you’re in for a fight, and chances are you’re gonna lose.” That’s not what happened, right? So Apple, in my view, beat the odds in the way that if Tesla is ultimately successful, Tesla will have beaten the odds, and that’s fine, right? That is a class of phenomenon that needs a theory to explain it. How is it that some companies enter well-established markets and prevail when that’s such a long-odds proposition?

Sonal: As you know, like — folks like Ben Thompson, John Gruber, myself — others have made the argument that it was disruption, but because it was a disruption to the PC industry.

Michael: But that’s the longitudinal problem. The cross-sectional problem is how did they enter the smartphone market? They entered it with a sustaining innovation, and it worked. Good for them. Now, how did they realize growth out of that? Well, they were busy racing up the disruptive trajectory displacing the personal computer. Terrific. Every company is playing both games at the same time. They have to be winning the cross-sectional battle they’re in…

Sonal: As well as gaining points for the long game, yeah.

Michael: We’ll go back — you know, sometimes with the benefit of, you know, the perspective that history provides. <Hindsight.> I’m not reinterpreting — if you look at say, Xerox and personal copiers.

Sonal: Right, we talked about this a few years ago.

Michael: Absolutely. This is near and dear to your heart, I know. So, the early personal copiers, they had a cross-sectional battle to win themselves. They were competing with carbon paper and Gestetner machines. So, they were more expensive than those — so they had to be better, right? They had to win the cross-sectional strategic battle for the niche market that they wanted. Now, it was niche to Xerox, but it wasn’t a niche market to the folks who made carbon paper and Gestetner machines, so the personal copiers had to win that fight. Right? And then they followed the disruptive path into commercial applications for photocopying technology. And, by the way, you need a different toolkit to understand how to win that cross-sectional battle. That’s a strategy problem, right? Strategy is about the constraints you embrace. The innovation problem is about the constraints you break, and you need a different toolkit to understand that. And I think a very powerful tool in that toolkit is disruption theory, and there are other tools — diffusion theory, crossing the chasm, there are others.

Sonal: So, another thing that people tend to equate when it comes to disruption — and this actually comes up in the case of the iPhone that we were just talking about — is that disruption equals money. Clearly, not all wildly successful products are disruptive. Is that true the other way around?

Michael: Yes. So, if disruption were defined as being successful, it would be useless as a theory.

Sonal: That’s a good point.

Michael: Right? And so there are any number of efforts that have tried to follow disruptive paths that have ultimately failed. We’ll go back to the core research that led Clay to create or discover the theory, depending on how you think about these things, in disk drives. So, each subsequent generation of disk drives — you start out with, you know, the Winchester drives, and then the eight-inch drives, and then the five and a quarter and then the three and a half. And with each generation of disk drives, there was a ravenous horde of companies that were seeking to deliver that new generation of technology, all eager — and, in fact, quite ably — following the disruptive path. And guess what? Not all of them succeeded. Some did, some didn’t. Back to my earlier observation — they have to win the cross-sectional battle as well as the longitudinal one.

Sonal: The longitudinal one, right.

Michael: And disruption theory doesn’t say anything about that. That’s not a shortcoming of the theory. That’s not, as they say around here — that’s not a bug, that’s a feature. Right? Because it’s not a theory. Theories are powerful when they have boundaries, when you know what phenomenon they are used to describe. It’s like an antibiotic. If you take antibiotics when you got a cold, you’re actually doing yourself harm. And the same goes for any good theory, right? If you start applying it when it doesn’t apply, you’re highly — in fact, you are more likely to make the wrong decisions than if you just didn’t use it at all.

Sonal: We have a lot of entrepreneurs in our audience, and I want to make sure that they — you know, that we’re not just talking theory, that there’s something concrete that we can do with this information. How do you resolve the tension between this — you know, if you’re focusing on the long game, the longitudinal battle, how do you, then, address, sort of, the cross-sectional reality that’s right in front of you?

Michael: Yeah, so that I would put in a category of a strategy problem, right? How do you actually create a strategy? How do you embrace different trade-offs in a different way from your competition, so that you’re differentiated in a way that customers find valuable? The good news is that there, once again, there’s a long stream of both scholarship, theoretical and applied, that seeks to tackle that problem. I’ve tried to make my own contribution to that body of work as well. In 2013, my book, “The Three Rules” came out with my co-author Mumtaz Ahmed, and that was an attempt to try and unpack — what does it take to win in the here and now? When you face trade-offs, which trade-offs should you embrace, and how do you go about remaining committed to those choices over time?

Sonal: How do people decide to make those trade-offs? Like, what should they know?

Michael: Better before cheaper, revenue before cost — and there are no other rules, if you’ll forgive me.

Sonal: That’s great.

Michael: And it’s intended to look at, kind of, the three core questions that I think define any business. In the first instance, how do you create value for your customers? And there’s basically two ways you can do that, right? You can provide superior value or you can provide lower price. And we’ve concluded that companies that deliver exceptional profitability over time focus systematically on better before cheaper. The second question is, how do you capture value for yourself in the form of profits? And here, the arithmetic of profitability is pretty straightforward, right? It’s just revenue minus cost. Guess what? Companies that deliver superior profitability focus on revenue before cost. And then, finally, what do you change when everything around you changes? And the answer is anything, except those first two rules.

Sonal: Oh, that’s great.

Michael: Which is why the third rule is, there are no others. So, those are rules that I think — they pass the test of being falsifiable, right? If I’d said the rules were cheaper before better, I wouldn’t be talking nonsense. There are people who actually think price-based competition is extraordinarily powerful. Look at the big discounters in any industry. What we found is the data point in the other direction. If I told you that, you know, being a cost leader is key to superior profitability, you probably think, “Yeah, that makes sense.” And it does make sense. It just happens not to be true, which is that systematically, over the long term, companies that focus on superior revenue, either through higher unit price or higher total unit volume, are more likely to deliver superior profitability than companies that focus on cost leadership.

Sonal: Yeah, I mean, just one last point on this. I think this is where the studies do get a little tricky, because we’re looking at larger data sets, but every success story — and I admit that there’s definitely a survivor bias when I make this claim I’m about to make — there’s an outlier of success that always just proves every theory. I’m thinking of Amazon, for example.

Michael: No, of course, which is why — no, no, no question, which is why it’s called “The Three Rules,” not the three laws. And here’s what we think the rules are good for — which is that some folks may be of a mind that look, “You can collect the data, analyze the data, and come up with the answer.” The data are always ambiguous, right? What data mean is as much a function of what we impose on them as what they say to us.

Sonal: That’s right. Exactly.

Michael: And so if you can’t be bias-free, because you can’t, the best you can hope for, perhaps, is to have the right bias, right? Play house odds, if you will. So, when we look at it, we say, “Look, the bias should be better before cheaper, revenue before cost.” If the data convince you otherwise, then you should go in the other direction. We’re not gonna say, “Well, I’m just going to ignore reality and follow the rules.” That would be silly. It’s better before — and note, better before cheaper. It’s not better, not cheaper.

Sonal: Right. You’re just saying how to prioritize and make those trade-offs, in that case.

Michael: Well-played. I would agree. Exactly.

Sonal: So, to wrap up a bit then let’s talk about — it’s a phrase that people here tease me about all the time — some “nuggety nuggets” that came out of your…

Michael: I can see why they tease you about that.

Sonal: I know, but I use it when I describe when we’re working on decks, like, “Where are the nuggety nuggets?” But, anyway, what are some of the nuggety nuggets coming out of your paper that you can share with us? Like, other things that, you know, are, kind of, some cool insights?

Michael: “Innovator’s Dilemma” was the first popular expression of disruption theory. It wasn’t the first. In fact, Clay’s theory of disruption was really born in his doctoral thesis. But ’97 is a long time ago, and the first article that introduced it to a popular management article was in the Harvard Business Review in 1995. So, it’s actually 20 years since disruption theory was kind of introduced. It’s not like it has been frozen in amber for that 20-year period, and so it’s important to remember that. Some of what’s happened is that people have picked up “The Innovator’s Dilemma” and read that very carefully, and said, “Okay, I’m gonna go after this.” Which is — again, that’s how we learn. That’s how science progresses. Absolutely. But it makes a lot more sense to grab ahold of the latest formulation of the theory that takes advantage of everything that’s been learned over the last 20 years.

Sonal: I know we have to read the December issue. We’ll wait. We’ll read it, but what are some of the other things you can share with us as an early preview?

Michael: One is that I think disruption has come to be used in a way that people say they are not using in a technical sense, and they do not mean to invoke Clay Christiansen but, indeed, if you use disruption to mean something has revolutionized and improved outcomes, then that’s what you’re doing, right? Because the English word means to introduce chaos, not to introduce a new and better order. Right? So when we use disruption with an innovation connotation attached to it, then disruption theory comes along for the ride. And the bad news is that when that happens, we’re back to the verbal inflation problem. We actually lose the power that disruption theory has to offer, and that’s what concerns me, right? So, my hope is to kind of — that the December piece, in part, will begin to save disruption from its own popularity.

Sonal: I love it. Saving disruption from itself. Well, Michael, thank you for joining the “a16z Podcast.” This has been a great conversation. I’m glad you disillusioned me. I’m gonna probably lose some sleep over some of those. No, I’m just joking. Not really. But, thank you.

Michael: My pleasure.

• 

  Sonal Chokshi is the editor in chief as well as podcast network showrunner. Prior to joining a16z 2014 to build the editorial operation, Sonal was a senior editor at WIRED, and before that in content at Xerox PARC.

  Follow Twitter
• 

  Michael Raynor

When most people think of big data they think of numbers, but it turns out that a lot of big data — a lot of the output of our work and activity as humans in fact — is in the form of words. So what can we learn when we apply machine learning and natural language processing techniques to text?

The findings may surprise you. For example, did you know that you can predict whether a Kickstarter project will be funded or not based on textual elements alone … before it’s even published? Other findings are not so surprising; e.g., hopefully we all know by now that a word like “synergy” can sink a job description! But what words DO appeal in tech job descriptions when you’re trying to draw the most qualified, diverse candidates? And speaking of diversity: What’s up with those findings about differences in how men and women describe themselves on their resumes — or are described by others in their performance reviews?

On this episode of the a16z Podcast, Textio co-founder and CEO Kieran Snyder (who has a PhD in linguistics and formerly led product and design in roles at Microsoft and Amazon) shares her findings, answers to some of these questions, and other insights based on several studies they’ve conducted on language, technology, and document bias.

Show Notes

• How analysis of language can predict success on Kickstarter, affect job listings, and more [0:00]
• Specific words and phrases to use and avoid [9:59]
• Discussion of how the analysis works [16:11], and how language can affect gender bias [23:59]

Transcript

Sonal: Hi, everyone. Welcome to the “a16z podcast.” I’m Sonal, and I’m here today with Michael, and we are talking to Kieran Snyder, who is the CEO and co-founder of Textio, a company that analyzes job listings to predict how well they’re going to perform, and can help optimize them to get more qualified, diverse candidates. And interestingly, they’ve been able to figure out, besides what doesn’t work very well in job descriptions — words like synergize — they’ve been able to figure out what does work well.

Broad effects of language

Kieran: Language, like — in tech, people love to talk about hard problems and tough challenges.

Sonal: But it’s a lot bigger than just about jobs. The ability to understand the words we use and how we use them is pretty important, because even though we’re completely immersed in a world of tech, where a lot of the conversation is around big data as numbers, a lot of the data that we produce — or, the output of our work — is actually taking place in the form of words, and those words matter.

Kieran: Sometimes how you say things is more influential than what you’re actually saying, right, and it’s counterintuitive to any of us who’ve built products before, because you like to think you’re leading with a strong vision.

Sonal: Clearly, words matter. And another place that that plays out is with hidden biases that are often revealed in words. For example, Kieran examined a number of resumes to see the differences between how women and men describe themselves, as well as in performance reviews, to see the ways that women and men were described differently.

Kieran: The word abrasive, which has been talked about since then, ended up, you know, being used in 17 out of a couple hundred women’s reviews, and 0 times in men’s reviews, right. The, sort of, stereotypical, like, “aggressive” was used in a man’s review with an exhortation to be more of it, and in women’s reviews, it’s a term of some judgment.

Sonal: Okay. Let’s get started. Kieran, welcome. So, the reason we actually invited you to the “a16z Podcast” today is because you’ve been writing a lot of interesting work based on the outcomes of your product, where you’ve been analyzing people’s use of language in certain contexts as a way to surface insights. And I think that’s really fascinating, because I think we have a tendency, in our world, to focus on big data as if it’s just numbers — and not other forms of data, because you’re really describing — I mean, what you describe your work as doing is applying machine learning to text and natural language. So, how did you kind of — how does that work, and then we can talk a little bit more about how you got there?

Kieran: Yeah. So, how does it work? Language is just an encoding of concepts, right, and anything that can be encoded can be measured. And so, I was sharing this story the other day — we were actually — originally started out looking at Kickstarter projects, right. So, we started out with this question — could we just look at the text of a Kickstarter project, and some of its, you know, metadata around the text and predict, you know, before it was ever published, whether it was going to raise money. And we didn’t look at the quality of the idea. We didn’t look at whether a celebrity endorsed it. It turns out we got over 90% predictive on minute zero of a project, as to whether it was going to hit its fundraising goal, based solely on things like how long is the text, and what kind of fonts are you using, and how many headings do you have.

Sonal: So, wait a minute, just to unpack that a little bit. So, before the project even went live on Kickstarter, just looking at those features of the text, you’re able to predict whether it [will] be successful or not.

Kieran: Exactly.

Sonal: What were some of the high-level takeaways from that?

Kieran: Yeah. So, longer is better where Kickstarter is concerned.

Sonal: Interesting.

Kieran: Kind of counterintuitive. One thing that broke our hearts, because my cofounder, Jensen Harris, and I both have some design background — you would think these cleanly designed projects, with this beautiful use of single typography would do best. Not so. You want it to look like a ransom note. So, you want to mix and match types. You want lots and lots of headings.

Sonal: Oh, my God. That sounds visually painful.

Kieran: You want images to be frontloaded, kind of makes sense. But a lot of what we found was not intuitive.

Sonal: Interesting.

Kieran: And so, it demonstrated for us the value of actually measuring, because the whole Kickstarter corpus is out there in the world, right. So, you can actually have great training data. You can see how well prior projects have performed. And we saw, “Hey, we’re kind of onto something here,” just looking at the — so very painful as a product person, the quality of your idea doesn’t matter — just looking at the content aspects we could predict.

Michael: And how do you account, then, for all the other sort of outside variables, you know, whether it was at the beginning of the Kickstarter kind of, like, craze, whether it was a certain time of year for that matter?

Sonal: A certain type of product even.

Michael: Yeah. Or geography? How do you know that, in fact, your analysis was correct?

Kieran: I mean, you can look at some of those other factors, right, because you can see when projects are published. It turns out that doesn’t make a big difference. You can see — the only things that really moved the needle in a very short-term way are, do you have a celebrity endorsing you — because that can get you a lot of social media attention. It doesn’t make or break you, but it can help quite a bit. And generally, how good you are at your social media strategy can tip the balance a bit. But none of those other factors turned out to be as significant as we expected.

Michael: The ability to really zero in via just the text — did that surprise you?

Kieran: I mean, we started off with a hypothesis that it would be that way, and that, you know, sometimes how you say things is more influential than what you’re actually saying, right. And it’s counterintuitive to any of us who have built products before, because you like to think you’re leading with a strong vision. We weren’t surprised. We were curious, as we started to apply the technology to some other verticals, whether it would extend. You know, our first big area has really been in the area of job listings, where we’ve looked to see in the first real product application — where we’ve looked at listings now from over 10,000 different companies. 

We’ve measured who’s applied to which listings, and we do see — the content matters. We do see some tailoring by geography. It turns out what works in New York is different than what works in San Francisco. We see a lot of tailoring by industry. So, what works to hire in tech is very different than what it looks like to hire a claims adjuster, or someone in retail, right. So, you see some differentiation. But in all cases, depending on how you’re slicing and dicing the categories, that text leads — you know, we’ve looked at real estate a little bit prior to launching our jobs application, and we’ve seen the same principles apply.

Sonal: So, so far, you’ve been talking about the form of the text — like, the length and the fonts and the design — but, like, were there particular words that popped out as well, in terms of what people said on those Kickstarter descriptions, or anything like that? I’m bringing this up, because there’s just this recent anecdote in the news that I read, about someone saying that you can predict success or default of loan applications based on words people use — like God, or using God a lot will actually mean you’re more likely to default on your loan, for example.

Michael: By God, I’ll pay you every month. I promise. Yeah.

Kieran: In Kickstarter, we didn’t look at that. We started looking at that for real estate listings and then jobs, where we’ve looked at it quite a bit. So, we saw when we were prototyping out the real estate stuff that if you say “off-street parking,” that really moves the needle for low-income homes. But for high-income homes, in terms of the number of people who go to your open house, and then the eventual sale price of your home — for higher-priced homes, it’s actually a negative, because why would you want to highlight that it has off-street parking? It’s just sort of an expectation. So, we saw, you know, vocabulary mattered quite a bit. In jobs, it matters hugely. You know, we’ve identified, at this point, over 25,000 unique phrases that move the needle on how many people will apply for a job, what demographics, how qualified they are.

Sonal: Could you share some of that insight with us, because, you know, the reason I came across your work is because I read an article about how you analyzed performance appraisals and job descriptions for insights about what moves the needle, and the differences in how people communicate. What are some of the things — I mean, just because we have a huge audience that does job descriptions.

Michael: That needs to hire some people.

Kieran: Yes. That needs to hire. Yeah. So, there is, sort of, a set of language that works really well for everybody. These are not surprising on the face of them, but when you look, you see lots of them. So, things like, “We’d love to hear from you.” Be really encouraging and positive in your listing. Using the right balance of talking to the job seeker. So, your background is in science, and you really enjoy roller skating in your free time. And talking about the company. “So, we stand for this,” in terms of the balance between “you” statements and “we” statements, can matter. You know, language like — in tech, people love to talk about hard problems and tough challenges. Curiously, we see patterns change over time. So, my favorite example of this is the phrase big data. So, a year and a half ago, if you used the phrase big data in a tech job listing, it was positive. You know, it was seen as compelling and cutting-edge. In June of 2015, it’s not negative, but it’s totally neutral.

Michael: That’s interesting. I wanted to ask, because if everybody, sort of, gloms onto these best practices, how then does the signal versus the noise shift?

Kieran: Exactly. Marketing content, as with any marketing content, the patterns that work change as they get popular and get adopted. And so, one of the reasons we believe software is so interesting as a solution here, is that it can kind of keep track at broad scale of what’s actually happening right now in the market. So, you may have published a job listing that worked really well a year ago, and probably have a lot of your listeners write their job listings as they go back to that one, and then they try to edit it and tweak it a little bit and fix it.

Sonal: That’s exactly what happens.

Kieran: Right. But it actually doesn’t necessarily work, because the market has changed. And so, there’s a lot there.

Key words and phrases

Sonal: Were you ever — I mean, I’m just curious about this — were you ever able to find or study associations between people’s intent and outcomes in job listings? So, for example, one of the things that we’ve seen happen a lot is that people only become real about what they actually want out of a job description when they actually put words to paper, and words have that power, to sort of help discipline what you’re looking for. You might not even know what you’re looking for until you write it down. Have you ever looked at anything around that, or found — heard interesting anecdotes around that given your work?

Kieran: We have seen that listings tend to perform better when they are originally authored. So, you can see some degradation over time when people patch, you know — I take a little bit from this listing and a little bit from this one, and I sort of stitch them together. And it’s probably because when you’re originally authoring it, you bring that coherent point of view.

Sonal: That’s really interesting.

Kieran: So, a little bit — pretty early for us to have seen that. And we also identify phrases that torpedo your listing.

Sonal: Like?

Kieran: Corporate sort of clichés and jargon.

Sonal: So buzzwords, basically.

Kieran: One of the very common — we call it a gateway term — that kind of torpedoes your listing is the word “synergy.”

Sonal: Oh, my God. That should torpedo any piece of content.

Michael: Yeah. Yeah.

Sonal: I don’t care what it is.

Kieran: But it’s a gateway term, because when people include “synergy,” they’re also significantly more likely to include, you know, “value-add” and “make it pop” — kind of silly, but they’re all over the place. And it turns out, every candidate of every different demographic group hates them. And so, there’s a lot of opportunity to improve in these jobs.

Michael: So, in the, sort of, the editorial world, we would call that jargon. And it sounds like…

Kieran: We also call it jargon, specifically.

Sonal: I think we all call it that. Jargon is jargon. No, totally. Actually, it’s interesting, because, with words like that, they’re obviously in use because they’re useful words, and it’s kind of sad, because — I mean, synergy at some point was probably a useful word. So, it’s kind of interesting, because over time, with your corpus of data, you’ll be able to sort of map how people’s language changes.

Kieran: Exactly.

Sonal: And when you think of dictionaries as, like, these static instruments for capturing text these days, it is kind of fascinating how language is changing in a way that we’re able to track differently now, thanks to online and software.

Kieran: It changes lexicography, like, just as a whole discipline. It changes lexicography for sure. I don’t know that you could do it in a static way anymore.

Sonal: Right. I totally agree.

Kieran: The internet has just exploded that.

Sonal: Right. Exactly.

Michael: So if big data is, kind of, neutral now, is there a kind of job type or job description that’s the celebrity of the job search world right now?

Sonal: Yeah. What word is, sort of, popping out that’s really moving the needle for you guys, or that you’ve observed?

Kieran: There are several. Most of your listeners are probably in tech. It varies a lot by industry. So, “at scale” right now. “At scale” is a very popular phrase.

Michael: A-ha.

Sonal: That’s popular here, too. We talk about that a lot.

Kieran: Yeah. Well, it is. You don’t want to do things and use methods that are perceived to be manual, or perceived to be limited in some way. So, “at scale” is one that shines — and it started in tech, but it spread to other industries, which is common that we see that. One of my favorite examples, given that we spend a lot of time talking to HR people, is — turns out “workforce analytics” is no longer a good phrase to use. You want to use “people analytics.” So, you know, you can get these highly specific, you know, deep in an industry changes — that if you’re in the industry and you’re on the cutting edge, you probably know, but if you’re just a startup trying to hire your first analytics person, you probably have no idea. You don’t have a deep background in the industry.

Sonal: That’s great.

Kieran: Right. Yeah.

Michael: So you’ve described different job listings in real estate. And so, this approach you think can extend in different directions. You started with Kickstarter, but what is it that it’s doing, and how do you — like, it seems a little bit magical, I have to say — that, like — I know that this is a job listing, so therefore, it’s going to have to do this. But a real estate listing has to do something kind of different.

Kieran: Right. That’s a really good question. So, you know, this approach is as powerful as the data set that you have. So, if you want to understand a document type, the very first thing you need to do is collect a lot of examples of the document type. And that means you need the documents, and you also need some information about their outcomes. So, you are publishing a Kickstarter project. We want to know, did you make money or not? That signal for us. You’re publishing a job listing. We want to know, did you attract a lot of good people? Did you attract only men? Did you attract no one? So, you know, for each document type that we take on, the first thing we do is, we make sure we build out a great training data set. 

And then we apply really classical natural language processing techniques. So, we look for patterns, and we say, “Okay. These are the ones that were successful,” where successful is defined as, you know — attracted more applicants than 80% of similar listings, maybe. And then we start looking for the linguistic patterns in the successes, the ones that aren’t as successful, ones that skew in a certain way demographically, and then we play that back. So, sort of a key thing for us, is that you get that feedback in real time, as you’re typing. So, as you’re working on your document, before you ever publish it, pay to publish it somewhere, you can make it good. And so, the training set is the, sort of, core of all of that, because without that outcomes data, then it’s just someone’s opinion.

Michael: And then could you extend that to say, like, “Look, I want to write a screenplay for a blockbuster.” I mean, could you — people have probably tried this, but…

Kieran: In fact, a very prominent Bay Area CEO proposed to us a couple months ago that we start applying this to screenplays.

Sonal: To actually start producing content, or just analyzing them?

Kieran: Sell it to Hollywood.

Sonal: Oh, wow. That’s great.

Kieran: Yeah. So I think any time you’re writing content to sell something, this is really interesting technology. And you could be selling your company. You could be selling yourself — you’re a job seeker with a resume that you want to have optimized. You could be selling your product in an e-commerce setup. You could be marketing yourself. You could be marketing blast emails. Any time you’re writing content to get people to take an action, this is really useful technology.

How the analysis works

Sonal: Well, let’s talk about where this fits, and let’s purposely use some jargon here, and let’s talk about where it fits in the tech trends — like, where it fits in that space. So, it sounds like you’re describing — big data techniques apply to natural language, or machine learning techniques applied to natural language. But natural language has been around for over 3 decades, 30 years. I mean, in the early days, they didn’t have this kind of corpus to train the algorithms on, obviously, so they had to use different kinds of techniques. Like, where does your work fit, and how do you see how it fits in the evolution of natural language — like, how has it been and where are we now, kind of?

Kieran: Yeah. I mean, I think in core natural language processing, empirical strategies have always been really important. So, when I was a grad student years ago, writing a dissertation, collecting data was just a lot more work, right. So, I had to go and record people in the field, and I had to transcribe things. I mean, it feels ancient now, actually, but I actually finished my Ph.D 12 years ago. It wasn’t that ancient. The fact that the internet has codified everything over the last 15 or 20 years, at least in English and most Western languages, means that you have this ready set of corpora available for you. The tricky part is collecting the text and the outcomes.

Sonal: Right. 

Kieran: The outcomes are the part that’s hard. Finding the content is easy.

Sonal: So, you’re describing the difference between just analyzing something and being able to predict something using that text.

Kieran: Exactly. When you analyze something, you can say, “Oh, cool. This word is really popular now. That’s an interesting fact. It might be valuable to someone to know it.” But it’s different than saying, “This word is actually helping your document in some way.”

Sonal: What are some other scenarios where you could use, sort of, this natural language text analysis to predict interesting things?

Kieran: Yeah. So, people are really starting to think broadly about this. We saw a New York City-based company helping people optimize the sale of their New York City apartments recently, using the right phrases. We’ve seen people do things in healthcare that I think are really interesting. It’s not a known vertical to me, but looking at the kind of notes that doctors take about a patient, and predicting the patient’s likelihood of having a major insurance incident over the next, you know, 12 to 15 months. Some really interesting things in actuarial science. Like, I think anytime people are producing text — which, by the way, in businesses, whatever your business is, text is actually the thing you produce the most of…

Sonal: Right. I believe that.

Kieran: …which any industry, and so people produce a lot of text. It’s meant to describe often what they think is going to happen. And so, I mean, the field of opportunity is pretty big.

Sonal: The techniques you’re describing — is it the same underlying technique applied to all different domains, but do you have to also train each corpus on a different domain? Like, there’s a special inside language in each industry. Or are there also universals across all of them?

Kieran: That’s a really good question. You don’t know until you train, is the short answer to the question. So, we have a set of NLP libraries that look for common attributes of text, and we always start out any new vertical by turning them on the documents and seeing what happens. So, things like sentence length — almost always interesting. Things like the density of verbs and adjectives — almost always interesting. Document length — almost always interesting. But the specific phrases that matter, and what it means to write a job listing, is very different than what it means to predict whether a patient is going to become ill, right. 

And so the specifics matter. The goals matter. So, if it’s a document that’s intended for broad consumption, it really probably shouldn’t be longer than 600-700 words. If it’s a stock prospectus, where you’re giving a company some information about how their stocks are likely to perform, it’s going to be pages and pages. And so, you know, the specific benchmarks that you’re looking for often vary vertical by vertical, but the principles of the kinds of things you look for are pretty similar.

Sonal: In the past, it seemed like only really big companies could do this, because they had, like, the type of computing hardware and processing power to pull this off. Like, what’s changed that a small startup could do this?

Kieran: AWS. AWS is what has changed things, right. I mean cloud compute at scale and, you know, Google Cloud and Azure. There’s a lot of competitors now, but AWS did this for startups, I think. And I say that, not because I worked at Amazon before, but it actually is. Like, for our team to set up the server infrastructure that we need is [critical]. You know, so I think that that’s a thing. And just the fact that there’s so much text data encoded on the internet. Google has democratized a lot of access to data. And so, that has helped, too.

Sonal: That’s great.

Kieran: Yeah.

Michael: Did you guys, I have to ask, did you kind of put any Kickstarter projects up there yourselves, just to give it a whirl?

Kieran: No. We were asked this a lot during our fundraising. We did look at pitch decks, by the way. One of the things…

Sonal: Oh, I want to hear about that, by the way.

Kieran: I will come back to your question. One of the things that’s been fascinating about having the beta out there in the world is the ways people are using it. So, of course, they’re using it for job listings, but people are using it for everything. Like, just a couple days ago, I had a material science professor write to me saying, “I put all my course syllabi through.” I was like, “Really? Like how did that work for you? I can’t imagine that that was a good result.” And he’s like, “Oh, I threw out all of the job parts. I just looked at gender bias, because that was a component that I needed for what I was doing.”

Sonal: Wow.

Michael: So, describe, when you say put it through — like, what happens? I understand, like — in my head, I have this idea that I’m typing along and, you know, suggestions come flying at me, but…

Kieran: That’s exactly what happens. So, there’s a website, and you paste or type in your content, and as you’re typing it’s getting annotated and marked up for you with patterns, suggestions, things you might want to change, scores.

Michael: And you can, in the case of the syllabi, right, you can dial it up or down depending on what you want the outcome to be. So, in his case, “Look, I’m sort of tracking for gender bias or…”

Kieran: He was looking for a specific aspect of what we provide. And, of course, the product isn’t tuned for what he wants, but he still found that aspect to be applicable to what he was doing. We’re seeing people put marketing content through, pitch deck content through. So, to your question, about did we initiate any Kickstarter campaigns? We didn’t because we weren’t making…

Michael: But you guys would be genius at it.

Kieran: We might be, yes. We’ve given a lot of advice to people on Kickstarter projects since then. But we didn’t, because we were making an enterprise product, right, and if we had followed through on a Kickstarter product and then it got funded, then we’d have to build it.

Michael: Right.

Kieran: But we helped friends, for sure.

Sonal: That’s great. So, what did you find out about the pitch decks actually? I’m totally intrigued by that, obviously, given who listens to our podcast.

Kieran: I mean, pitch decks are not always highly text oriented, right. So, great pitch decks don’t include just your text attributes, but there are certainly things like length of your deck that matter. Slide titles end up mattering quite a bit, because people are looking to see a certain style of content.

Sonal: And less space. And we’ve all seen any kind of meeting where some one person gets hung up on one word in a headline.

Kieran: Yeah. It can.

Sonal: It always happens too.

Kieran: It can. We didn’t go deep on pitch decks, but we looked at as many as we could find as we were building our own pitch deck in our last round of funding, and found some patterns in the set.

Michael: In the synergy line of questioning, were there words or phrases you should never include in your pitch deck?

Kieran: You know, I don’t know.

Michael: Okay.

Kieran: I don’t know.

Sonal: I guess, there might not even actually be — yeah. I wonder if there’s — there’s never, I guess, a set set of rules.

Kieran: I bet there are. We didn’t identify them.

Michael: Right. Synergy is probably one.

Kieran: Yeah.

Language and gender bias

Sonal: Actually, let’s talk a little bit more about — and maybe we should wrap up on this note — let’s talk a little bit more about some of your findings around gender differences.

Kieran: Sure.

Sonal: So, you said the materials science professor tested his own syllabus — which again, I’m not sure that made sense, like you said, because there wasn’t a reference corpus to, I guess…

Kieran: There wasn’t, but when you have, you know, tens of thousands of phrases that are lighting up, and he’s writing for a science STEM student population, odds are good that there’s going to be some lexical overlap.

Sonal: Oh, that’s great. Right.

Kieran: So, you know, he found some things there.

Sonal: So, describe some of your findings around job descriptions, because — given what your product focuses on right now in terms of gender differences — and how people — what things you picked up on that?

Kieran: Yeah. So, prior to us doing this, there was some really strong qualitative research, right. The National Coalition of Women in Technology, the Clayman Institute here at Stanford — they’ve done some really interesting qualitative work, but the number of phrases that they identified was on the order of a couple hundred. Avoid “rockstar.” Avoid “ninja.” You know, we want to hire more women in technology.

Sonal: Guru.

Kieran: The interesting thing for us — first of all, we’ve talked to a lot of industries outside of tech. And so, while in technology we want to hire more women, when I talked to people who are hiring ICU nurses, or elementary school teachers, bias goes the other way. And so, it’s very important to us that we don’t judge — we just forecast and let you make the right choices for your business.

Sonal: Right. Whatever you’re optimizing for given wherever there is an indifference or imbalance.

Kieran: Right. Right. So, I will say, we have validated much of the qualitative research, which is good, that there’s, you know, some alignment on those points. We have found cases where things are — it’s pretty subtle, right. So, the difference between “fast-paced environment” and “rapidly moving environment” — it’s almost head scratchingly tiny, but statistically, one of them…

[End of Transcript]

• 

  Kieran Snyder