Transcript: James Cham on Venture Capital, Risk Taking, and the Future Impacts of AI – Episode #12

Steve: Hi, this is Steve Hsu, and this is Manifold. Our guest today is James Cham, a venture capitalist at Bloomberg Beta. Corey couldn’t make it today, so it’s just me and James. We got to know each other, I think, starting many years ago through a kind of unstructured Silicon Valley meeting that has become popular. These meetings have often, they’re broken into sub-meetings through a kind of self-organization process, and the sub-meetings are often on all kinds of wacky topics ranging from, you know, how to raise your kids in the digital age, to what are the five best books you’ve read in your life, to how do you fix your failing startup. And in all of those various kinds of discussions, I’ve always found James to be a very insightful guy who asks really good questions and also has really unique perspectives and insights. So I actually think of him – no offense to other VCs — as one of the most thoughtful VC friends that I have, and so I always like to get James’s perspective and input on…

James: That’s great very kind but might be a low bar, might be a low bar.

Steve: [laughs] … not just things related to tech, but things related to life or the future or almost anything, culture… So let me start, James, by asking just a little bit about your life history, which I actually don’t really know that well. Where did you grow up?

 James: So I grew up in the San Gabriel Valley, which is just east of downtown Los Angeles. My folks lived initially in Montebello, and then now they live actually fairly close to Caltech…

Steve: Wow.

James: … where they moved after I graduated from high school — while I was in high school. But I do remember sneaking into Caltech and playing with various VEX machines because, of course, one, Caltech had notoriously bad security problems by design, and all the best computers.

Steve: What high school were you at?

James: I went to a great public high school called Montebello High School.

Steve: Wow. The area of LA you’re describing is where about half of my cousins live, ranging from Pasadena to Hacienda Heights to Pomona, yeah, so it’s like a Chinese, not really ghetto, but Chinese area of Los Angeles.

James: Yeah, the part that I grew up with was just south of that. So Montebello at the time was probably 60-70 % Latino, and so that was post-migration in the ’70s and ’80s.

Steve: When you were a kid, like when you were in high school, what did you want to do, what were your life aspirations?

James: You know, when I was in — my mom found this a few years ago — when I was in the fifth grade, which would have been, I guess whatever, 1984 or ’85, I wrote a report about what I wanted to do when I grew up. And in that report I lived in Silicon Valley, and I worked — and this is not a lie — I worked as a developer evangelist. I wanted to convince other developers about what frameworks they should use. And so this is like sort of… The oddly specific thing, I think that was probably because I must have read something by Guy Kawasaki or something, right. And so yes, that was my initial dream. And then when I was in high school, I was very interested in journalism and was quite involved in the school newspaper. And then when I went to college, I sort of fell back into programming, because honestly, there’s nothing quite as satisfying as software development, right. There’s nothing quite as satisfying and maybe nothing quite as, like, maddening, but there’s that joy in actually getting something to work. And so I was a CS major in college.

Steve: So I’ve heard it said many times by my former students and colleagues who are really active software developers that there’s nothing cooler than code, because you’re creating your own little universe and you get to control what happens in that little universe, and only you created it at the end of the day. So I think I definitely understand where you’re coming from.

James: Yeah, there’s definitely a way in which I was quite pretentious in high school and the early part of my college career, in part because I felt like computer science was the only honest science, because it admitted that everything was entirely human-made and everything was pragmatic, as opposed to pretending that something came from some, you know, sort of grand design as opposed to something that people put together.

Steve: Right. Now, as a physicist, we could have debates about the…

James: That’s why I said that…

Steve: … primacy of computer science versus physics. The physics attitude is like, what are these guys doing, they’re just writing little execution lists, you know, algorithms, just a little set of instructions. They’re not really getting at the deep reality of nature and things like this. But let me leave that there for a moment. Now you went to Harvard, which many people think of, maybe me too, as perhaps the greatest university in the world, but it isn’t really the greatest computer science department in the world. So, given your already existing interest in computer science, why did you choose to go there?

James: Yeah, to be honest I probably would have made a different decision if I thought I was going to major in computer science when I actually attended.  I think this was the case where I thought I was going to do something, you know, either I was going to be pre-med or else I was going to do journalism, and I spent a bunch of time in the school newspaper. And so yeah, it’s one of those cases where I sort of fell into it sophomore year.

Steve: So after Harvard, you worked for a while, and then you went to the Sloan School to get your MBA, right? You’re a Sloan, MIT Sloan MBA?

James: That’s right. And so for those — I actually don’t know your audience, like the demographics of your audience — but there was this time, you know, sort of like in the late ’90s, when there was a little boomlet that’s nothing like the boom that we’re in the middle of right now, and I was a software developer in Boston at that point. And then, you know, everything fell apart, right, everything fell apart in 1990-91 — I mean 2000-2001 — and then of course 9/11 happened. And I was part of a startup that was failing and got saved by being acquired by a large funding corporation. And that was kind of a terrible experience, but I will admit that the most maddening thing about it was that I felt like there was this secret language that the people who weren’t building stuff would speak in, and a secret language around economics and accounting. The best way, at least for me at the time to escape, was to escape to business school. And I really enjoyed Sloan, and I probably took it a little bit too seriously, because it certainly does away with… I actually took no economics courses or accounting — Harvard doesn’t offer accounting, right — but, you know, I took none of those sort of courses as an undergrad, and I kind of fell in love.

Steve: It’s funny, because at the meetings where I often hear you talk and ask questions, I can tell you have an economics worldview, because you’re often analyzing things in terms of markets and market response. And so I would have guessed, if someone had just asked me, oh yeah, James took some economics courses at some point in his life, because he knows the lingo and all the specialized terminology.

James: Yeah, that’s all from business school and with economists afterwards, but I do probably come with come at it with the enthusiasm of a convert, rather than someone who was forced to take Acct. 10.

Steve: So I think you went to business school with a friend of mine named Michael Ibrahim, am I correct?

James: I did indeed.

Steve: So just a shout-out to Mike Ibrahim. Mike did a PhD in theoretical physics at Yale, that’s how I got to know him — I was on the faculty when he was a graduate student — and he joined my first startup, called SafeWeb, where he was the CTO for a while. And Mike is one of these guys who’s just got an incredible amount of ability and he can work like a horse, and so he played a big role in that startup before he went to Sloan. And so just by coincidence, I guess, we both know Mike.

James: That’s right.

Steve: So okay, so after business school, how did you get into venture?

James: So during the summers, my intention was to do something software related, and then at the time I thought I’d want to be a product manager. But I actually… and then, after returning to LA for family reasons, I worked at the Boston Consulting Group, which I enjoyed. And then, as it is in life, a friend of mine actually poked me out of the blue while I was just on paternity leave, actually, saying that I should talk to this VC firm in San Francisco called Bessemer, because they were looking for someone to join and he thought I’d be a good fit. And he was right. I really enjoyed it, you know, sort of there’s a way in which investing matches a lot of the things that I like and that I’m good at, right. I mean, there’s the sense in which you get to constantly look for the novelty of the new and see what’s different, and so that’s a lot of fun. There’s a way in which the problems are varied, that’s a lot of fun. And there’s also a way in which you get to talk to lots of people. You know, I didn’t really have an extrovert-focused job for most of the first half of my career, but, you know, I tend to like people, I’m curious.

Steve: I want to delve into venture and venture capital in some detail, because here we have a very special beast, a Silicon Valley venture capitalist, to talk to. You know, I’m recording this from Michigan, which is, you could say, an underserved area when it comes to venture capital  — the whole Midwest is in some sense, even Chicago, I would say — and so what I often tell people around here — and even sometimes when I’m talking to people internationally, like if I’m in Hong Kong or in Berlin or, you know, places like that — I’ll tell people that this is a new aspect of the economy which is extremely central to having economic dynamism. If you want a pathway from the laboratory or from the individual inventor to the marketplace, you really need a mechanism which pools capital, and that capital is willing to take risk, risk on new technology, new business models, new attempts at organizing society. And so that’s sort of what venture capitalists do. But I think it’s a new enough phenomenon that almost nobody — like if you go to Modul University campus, even Modul Business School and talk to a business school professor — they don’t really have a deep understanding of what all these people are doing, and I think these people are playing a super central role in our economy. 

James: Yeah, how old is modern venture capital, it’s about 60 years old?

 Steve: You know, it really depends because, for example, when I got to Silicon Valley in the late ’90s we had that little bubble, the one that blew up around 2000-2001 as you said. But just the old school — where, you know, the VC would invest like a quarter of a million dollars and take a huge chunk of your company, and it was a hardware startup — that kind of old-school venture capital, there was still a lot of evidence for that around. And so that had persisted I think for twenty-odd years, but there were only a few funds, and they all knew each other. And it was from there that this huge thing exploded, you know, and kept exploding to become what we see today. So when you ask how old it is, it existed as a very unique narrow thing for a while, but I think it only really got big in the late 90s, I would say. That’s kind of my informal history of it.

James: Yeah, you know, others have, you know, sort of deeper histories of venture capital. I think, you know, how much of this is actually venture capital, and how much of this is the reflection of the way that network software is able to grow and create new business models, and expand extraordinarily quickly? It’s not actually clear to me, right. I mean, you know, sort of as a VC — I don’t know, most of your listeners will have noticed that VCs spent a lot more time talking about themselves these days, and part of that is just because of the nature of marketing and information. I think it’s all really good and the things we talk about, these issues are really important, etc. etc. The only critical thing is that it’s great as long as the VCs don’t actually believe what they’re saying, right. So the venture I think plays an important role in the ecosystem, but the core of what happens, or why the economy grows a certain way, is because of sort of some series of unique characteristics about software, which I think just can’t be underestimated, and then also the fact that in Silicon Valley and a few other places, you’ve got customers who are willing to try new things much more quickly than they would have before. And I think that results in a bunch of benefits to the entire ecosystem.

Steve: You know, I think you’re right in taking software as maybe the sort of best model for the type of technology that’s involved here. But let’s not forget there’s also hardware. I think there were hardware startups really before they were even big software startups. And there’s also biotech, for example, which is a whole different beast, right, and consumes or absorbs quite a lot of venture investment right now. You know, if we were sort of more careful historians, I maybe would have prepared a graph that, you know, showed the percentage of free investment capital that was actually devoted to venture, you know, as a function of time, and then maybe we would see some big uptick in the late ’90s that just continued to grow. One thing I wanted to mention is that I think both Bessemer and Bloomberg, the two funds that you’ve worked for, have — am I correct in saying they have only one LP, only one — no, they have multiple LPs?

James: You know, Bessemer’s roots go back to the Phipps family, of course, sort of steel money, but by the time I joined, I believe they started taking outside investors. And by now, you know, they take a bunch of outside investors who are all clamoring to try to get inside Bessemer, because it’s just an extraordinary fund. And then in the case of Bloomberg, that’s correct, our fund has only one investor, Bloomberg LP.

Steve: Okay, I want to come back to the issue of LP. So for our listeners who are not really experts in this, LP means limited partner, and those are the people who put the capital, put the millions of dollars into the venture fund, which the venture capitalist then allocates to startups. So the venture capitalist is in a sense a money manager, and instead of buying stocks with the money that the investors give him, the VC is investing in shares of small startups. Getting back to what you were saying, I think you were making some modest noises about the role of VC. You know, one of the reasons why I can take the other side — of course, as an entrepreneur I agree with you completely, it’s the entrepreneurs who do all the work and, you know, get screwed sometimes, so as an entrepreneur I tend to be a little bit more anti-VC — but as somebody in the Midwest who works at a big university and tries to get technology from the university to the marketplace or tries to foster economic growth here in Michigan, I can see how important VC is, because when you don’t have it, it becomes incredibly hard for the entrepreneur to make progress. And so, you know, of course Silicon Valley’s become a magnet for really talented entrepreneurs, and so there’s just a plethora of people that are well suited and have the skill set and the risk tolerance to start companies. But there still are people like that, you know, in the Midwest, and it’s just much, much harder for them — even if they have a great idea and they’ve developed some great technology, maybe with their PhD adviser — it’s much, much tougher for them to get going because we lack a kind of critical infrastructure, which VC actually is. And so that aspect of it is kind of large in my mind.

James: You know, my partner, Roy Bahat, has led a bunch of trips of, you know, a mix of VCs and politicians through places that are under-invested in venture and technology — and he just did one a couple of weeks ago and I joined him on a couple — and I think that that is true, but there’s a little bit of a causation, sort of ordering-of-things question, where, I don’t know… I think that the VCs come where there are great businesses, and the great businesses come where there are great entrepreneurs and people with lots of good ideas; but to be honest, the thing that no one thinks about enough — at least I don’t think people think about this enough —people don’t talk about the demand side enough. I think that part of what made Silicon Valley great was sort of this massive over-investment by the Defense Department in Silicon Valley that was driven partly by some very clever things that people at Stanford did. And that meant that you could have this phenomenon, right, where you could have companies grow really quickly, because companies grow really quickly because people buy things from them right. And that turns out to be, I think, underappreciated because, you know, when we visit a bunch of great towns, all of them would have the signals of entrepreneurship, and they’d have the talent, and they’d have like the cool coffee shop and the accelerator; but the ones that looked like they were doing really well had local customers who were saying, you know what, I’m willing to take a risk and be aggressive, you know. So I think we’re seeing that in some cities. Where have you seen that — you look at, I don’t know, like we talk about China occasionally, but you look at China, and you look at the adoption of new IT software frameworks in China, and the willingness to like basically move your entire workload to some new Berkeley project that’s been around for like six months because you just need it to work, like that actually is what drives innovation, right. So I feel like that goes under discussed, and that’s part of the reason why, I don’t know — you and I talk about politics occasionally — I would vote for a president who made project management and procurement the center of their policy, because I think like that’s the thing that slows down the economy and that slows down innovation. If that actually changed, we’d all be better off.

Steve: Did you mean specifically the way federal government procurement and project management work, or meaning just emphasizing that skill set in the economy?

James: Emphasizing that skill set and making that the primary thing that people should be fighting about, you know. The New York Times should be covering the details or intricacies of procurement rather than the intricacies of, I don’t know, some movie star that I probably don’t know about. And I think like if we did that, we’d all be better off.

Steve: Well, I definitely want to agree with you that, as somebody who’s been involved in startups, I realize that there’s this whole skill set and time spent reflecting on that skill set — which is, you know, how to go from zero to one, how to actually effectively build the team, how to decide on a strategy, and what technologies to use to build the product, how to take the product to market — all of those things are very specialized skills, which honestly I don’t even think you can really learn them in an MBA program. You kind of have to learn them in the environment, and so the places that have that, and have lots of people floating around that have done it before, and not just done it but actually spent a lot of time thinking deeply about it. So, you know, I can remember this guy Mike Ibrahim that I mentioned — before he went off to Sloan, where you met him — you know, we had spent years working together, trying to figure out like what is the right development process to get this done, what’s the right way to handle the hardware procurement for our appliance and, you know, all those problems we had spent a lot of — and he’s a smart guy — a lot of smart-guy hours just trying to figure out what’s the right thing to do. And in Silicon Valley you can go to any cafe and you’ll just bump into somebody who’s been through that. In other parts of the world, not so much — maybe Beijing, but certainly not here in Michigan.

James: That’s right, it’s a huge missed opportunity. And there’s also a way in which… One of the things that we haven’t figured out how to do well over… Social media is so good at propagating information in so many ways, but it’s not really good at propagating that class of things that are almost semi-secrets. There are things that are known but not widely known and, you know, sort of figuring out how to do that, that feels like a missed opportunity. Because it’s certainly true that… You know, why is something like — I don’t know, right down your street, Duo — why did Duo do so well, is in part because Doug and team had a bunch of roots all over the country that meant that they knew how to scale a software company differently than, you know, let’s say the brand new undergraduate.

Steve: Yeah, you know, I have a feeling that, you know, if you and I were sitting at one of these Silicon Valley meetings, like we’re sitting around the Googleplex and some young guy comes up to talk to us and inevitably just starts pitching the startup that he’s working on, that you and I would have a lot of like, you know, we wouldn’t necessarily agree on all evaluations of the things he said, but there’d be a lot of commonality; whereas if it were a person from my team here in the research office at Michigan State, or even a business school professor who hadn’t done a venture, that person might have like completely orthogonal or very, very different takes on what the guy said than I do. And so there’s that common knowledge and experience in, you know, what works, what doesn’t work, you know, what’s likely to happen in the future. And I think that that is the intangible thing that is very hard to instantiate — like, do they have enough people in Pittsburgh who get that? You know, maybe they do now, maybe they don’t, but, you know, it’s still a long way from Silicon Valley.

James: Yeah you’re right, that it’s like a case where there’s a lot of unsaid assumed knowledge that is very hard to articulate clearly.

Steve: Yep. Now let me go through, you know, for the benefit of my audience, a little bit of like, kind of just a brief intro into venture. And I’ll give my own cartoon version of it, and you just correct any stupid stuff that I say [James laughs] or just make it more nuanced, okay. So my friend Joe wants to start his own venture fund, okay. Joe has been a successful entrepreneur, he’s got some of his own money. First thing he has to do, though, is attract some limited partners, some LPs to invest in his fund. So he goes out and meets with these guys, and some of these guys, you know, are, you know, super-high net-worth individuals, or maybe they run institutional money, like they manage the endowment of a university or something, and he says to them, you know, this is my fund, it’s going to focus on some hot thing — let’s say it’s going to focus on AI startups, maybe this guy says oh I’m going to focus on AI startups which really, you know, use narrow AI to automate factories or something, right, something like this — he’s got some theme for what his portfolio is going to invest in, and he goes out and tries to raise the money. Now, when he gets some investors that are willing to invest, he has to basically create a legal entity, which is a fund. The fund has a certain life cycle, right, and some schedule of dispersing the profits, which presumably they’ll make by selling the shares in the companies they invest in, or get from an IPO from those companies back to the investors. Say a little bit about how that looks to the guy who’s starting the fund — like I’ve heard things like, you know, you’re not really ready to go unless you’ve raised, you know, fifty million, a hundred million, the minimum amount it takes to operate a decent venture fund. Can you just say a few things about that?

James: So, you know, the person who… It’s interesting, because venture capital is so opaque and hard to understand in many ways, and the thing that is even more opaque and harder to understand is sort of the folks who invest in venture capital, the limited partners, the mix of family offices and wealthy individuals and institutions. And each one of them have slightly different incentives. The person who actually has… You should do a call-out to Beezer Clarkson. She’s done a very good job of trying to increase transparency and allowing VCs to understand how LPs think. I think that the first thing to realize is that for most of these types of investors, venture is just a small part of their overall asset allocation, right — they’ve got real estate, they’ve got a bunch of public-market investing — and there’s a way in which venture is basically a small chunk of their sort of portfolio in order to add a little bit of randomness, and hopefully the randomness is on the positive side rather than the negative side. For those investors, you know, the venture exposure is important but not the critical thing. So that’s the first part. And then I think, as far as strategies for VCs starting new funds, you know, it entirely depends on the strategy, right. I mean, there are a bunch of folks who will want to deploy hundreds of millions of dollars, or there’ll be a bunch of folks who will, you know, start these sort of “smaller funds” that will be, you know, about 50 million dollars, and they’ll start off by investing, you know, in the single-digit million dollars. So I don’t know that there’s a strong rule, so much of it depends on your strategy. And the other thing that’s true about venture these days is the range of investment now goes to everything from, you know, investing hundreds of thousands of dollars to investing hundreds of millions of dollars, right. So we live in this very confusing and exciting world in which innovation around the private markets and what to do with companies that have yet to go public, that information and all the new ways of being exposed to that, that’s changed a lot, right, so that’s constantly shifting.

Steve: Yeah, I’d love to get into SoftBank and some of these other things a little bit with you, maybe in a few minutes, after we just finish this like brief intro to venture, because I think yes, it’s exactly what you said, like what’s happening to pre-IPO unicorns, and what is SoftBank strategy, these things are all extremely interesting. But the floor for this sort of minimum fund size that I had in mind is that, well, you’re going to have to pay your team, and you’re going to have to get your team from the… Now, in the old days — I’m not even sure if this is still true — the old structure was 2 and 20, so you would get a management fee of 2% of assets under management, roughly speaking — could be lower, right, could be higher —

James: Or higher for smaller funds, depending on… If you’re just getting started, you might make money doing other things as well. A number of some of the interesting funds that we’ve seen that are thematically based will sometimes make money by doing recruiting, or by doing other services, or by, you know, taking a larger ownership stake in a company by helping to start it or something like that. And I think that dynamic, you’re right, is shifting. So I’m a little reluctant to do a hard and fast rule, but you’re right that the core economics of a VC firm — so how does the VC firm sort of actually make money and earn its keep — there’s some portion that comes out of management fees — and that’s a percentage of the assets under management that then get put into the pocket of the partners over the course of either five years or ten years — and then there’s what’s called carry, which is, you know, some percentage of the money that you make, right. So, you know, for the guys who made hundreds of millions of dollars, you know, sort of investing in Zoom or something like that, you know, some portion of that then goes back to the partnership, and the rest of it gets returned to the investors.

Steve: Right. So the number that I most commonly hear — which is also the number for hedge funds — the numbers are typically 2 and 20. So 2% management — and this is all just rough justice —but 2% management fee, and maybe 20% of, you know, maybe profits, or profits relative to some hurdle, which is some benchmark — it may vary — but something like 2 and 20. But the 20 is going to come years later, maybe. The 2 is what you use to pay your team right now, and if you raise 50 million, then you have a buck — you guys don’t say buck, but the hedge-fund guys say a buck for…

James: Yeah, the hedge-fund guys live in such a different world, of course.

Steve: Yeah. I’m unusual, because I’m sort of halfway… I know tons of hedge-fund guys from my physics background, because most of the physics guys who left basically work at hedge funds, but I know a lot of VC guys from the startups that I’ve done. So I kind of know both, I can speak to both populations of people, and there’s actually some crossover nowadays. So…

James: Oh yeah, there’s a lot of crossover. So some of the most innovative venture practices come out of guys who come out of the public-market hedge-fund world, right, where they’re saying wow, you VC guys are like such small fries, and you’re like sort of unimaginative about the structures that you’re using to invest in things.

Steve: That’s exactly right. You know, for the current startups that I’m involved in, I think there’s a kind of even mix of traditional VC-type guys and hedge funds that are actually investing as well, so it’s interesting because, as you as you point out, their worldviews are quite different.

James: Right. But you’re right that the right way to think about it is, there’s some base amount of money that venture capitalists make every year in order to pay their bills; but the real goal, the real hope is that at some point the companies that they invest in, or some percentage of the companies they invest in, will return so much money that it will make more money than the initial fund size, and they’ll be able to return a portion of that to their investors and keep a portion of that for themselves.

Steve: Right. Now to finish this intro, I’ll describe what it looks like from the entrepreneur side, since I’ve never worked as a VC. From the entrepreneur side, it looks like this. The hedge fund is a team, there’s maybe a few partners and then maybe a slightly larger number of people called associates, and then there are some kind of lower-level nondescript people who are, you know, just I think getting coffee and doing other stuff. You know, the thing the entrepreneur learns right away is that, even if the associate likes your idea and has spent a lot of time with you, it doesn’t mean anything unless one of the partners is really pushing, is a champion for your idea to actually get the fund to invest in it. Is that a distorted perception that we entrepreneurs have, or is that a realistic impression?

James: Yeah. You just think about the incentives, right. So the way that the incentives work ends up being that the partner, or the sort of people who own some of the fund or a big chunk of the fund, you know, a lot of their compensation reputation depends on the companies that they invest in, and thus, you know, funds will have different ways of allowing the investments to get approved. And typically, most funds will have some sort of mix of consensus or some vote-taking method in order to get investment, and typically someone puts their neck on the line to say, you know, I’m sponsoring this investment, and my career will be made or broken based on whether this works or doesn’t work, and that will typically be someone who’s called a partner. It’s all gets confusing because, of course, many VC firms realize that startup entrepreneurs realize this, and so they call everyone a partner [Steve laughs]. But my only observation out of that is that, you know, that is meant to make sure that good decisions are made, and that we are good stewards of our LP money. Sometimes that means that VCs overthink things or they spend too much time doing irrelevant diligence, and that’s always a balance. And the other observation is that, you know, VC partnerships are highly delicate sort of structures that, you know, are kind of designed to fall apart, and when you go look at the history of venture, there are very few institutions, you know, very few. So outside of, let’s say Bessemer, Sagil, Sequoia, you know, most of these funds fall apart, right, because they’re made up of individual high-ego partners who think that they know better, and when they’re wrong they suddenly lose a lot of political power, and when they’re right their heads get big unless they fight, you know. And so I think that’s sort of the nature of investing, and I won’t claim that the decision-making is, you know, sort of the soundest. I’ve seen a couple of funds hire outside consultants or business school professors to try to improve that decision-making process, but it’s still not ideal.

Steve: Yeah, it’s still got to be kind of an art. And so, you know, situations that I’m familiar with, you know, you could have one partner that has maybe the best relationships with the LPs, and so he can claim hey, I brought all this money in, so I’m entitled to the biggest chunk of the 2%, you know. Another guy could say yeah, but I’m the guy who is responsible for the most successful investments that we’ve made, and so I deserve, you know, this. And so, you know, obviously everybody’s got their own distorted sense of their contribution.

James: And it’s worse than that, because, you know, in public markets you can always know your score, because for better or worse, there’s some objective thing out there. In private markets, the only way to keep score is to try to keep track of some business metrics, but those business metrics may or may not represent future behavior. And then you get to keep score by saying that, you know, someone else invested at a slightly higher price, but that may or may not reflect like the actual value of the company either, you know. And so the politics of representing how well you’re doing, you know, is confusing and opaque.

Steve: Yeah, for sure you’d better be able to talk a good game, otherwise… I don’t know of any VCs who can’t at least talk a good game, right, because then how is anybody going to assign credit to you — unless you actually finally get liquid, right, like your company actually goes IPO or gets acquired, and you literally turn it into cash.

James: That’s right, sure.

Steve: Now in the hedge-fund world, a very important and common term is alpha. And so what alpha typically means is that, for a given level of risk that you’re taking in the firm with your investments, are you beating a portfolio, a kind of benchmark, with similar risk characteristics? So in other words, suppose you invest in just standard, just plain stocks that are in the S&P 500. One could ask hey, are you getting a return which is greater than the S&P 500? And if you do, then you have positive alpha. If you underperform, then you have negative alpha. And so a big issue in the hedge-fund world is like, does my PM, my portfolio manager that we just hired, does this guy actually have alpha, or is he just lucky? And it’s like the most fundamental question that you can ask in that area of finance is whether someone really has alpha, can we actually measure alpha, or are we just rewarding luck or people who talk a good game. Does that kind of discussion materialize as much in venture? I mean, I feel like venture is even more volatile and, you know, subject to very small events. I think Peter Thiel once admitted that like he made more from his single Facebook investment than everything else he’s done in his life. And so, how do you guys talk about the equivalent of alpha in the VC world?

James: Yeah, I think you’re right. I think it’s very, very hard to figure out. I think that part of it is also… Remember, Peter Thiel’s point is two… It’s worth unpacking, what he just said there, right, because on the one hand it’s a small investment, on the other hand it also represents why venture is different. There’s a way in which you can not just make, you know, multiples on your money, but you can make orders of magnitude multiples on your initial investments. And there’s a way in which the outcomes from venture on the upside are sort of unbounded, right, depending on the market that you get in, and I think that informs a lot of the decision-making and a lot of ways that LPs treat venture and that realization.

Steve: To give you a funny example, I had a good friend, someone I’ve known since undergrad, who went the whole physics career but then ended up leaving and becoming a hedge-fund manager, and he invested in tech stocks for quite a long time. And at one point he left the, you know, multi-billion-dollar hedge fund that he was actually at to start his own fund under his own name. And to raise money, to get LPs to invest in his own fund, he once said to me, you know, Steve, all I have to do is show them my tax returns, because if I show them my tax returns for the last ten years, they’ll see that I was making a — excuse me —shit ton of money for my previous firm, because otherwise I wouldn’t have gotten that compensation over the last ten years. So in their world it’s very easy to demonstrate alpha, in a sense — it’s like, I got compensated this much for running, you know, this fund for this bigger hedge-fund entity. Now if, for instance, you were leaving your firm and trying to get a job at another firm or raising money from LPs, what would you point to? Yeah, go ahead.

James: You have to look at your existing investments and the outcomes from those investments, right. But you’re right that like that’s… So the nice thing about venture is that these are all thinly traded stocks. So, you know, in the case of a PM, they might say, you know, they have a fairly complicated transaction log, right. In my case, my transaction log is pretty straightforward, like I invested at some point, I invested a little bit more at some other point, and at some point something became valued at some point, you know…

Steve: Yeah.

James: … so that ends up a much less complicated process. And then there’s also a way in which, for better or worse, you know, a star hedge-fund performer may do very well inside the system but may not do well on other systems, while in the case of venture oftentimes, you know, a lot of the best guys are, for better or worse, solo practitioners, right. And so, I don’t know, Bill Gurley’s ability to find a new investment is less dependent on benchmark than the other way around.

 Steve: Yeah, I sort of feel like in your world… So this friend of mine wouldn’t have said anything like, you know, everybody knows me, they love me, you know, because he’s investing in public markets, right, so he would just say look, this is how much money I made for the firm in the last ten years. Whereas, I think in Silicon Valley you might easily get hired because people say like, well the entrepreneurs really love this guy, they think he’s really smart, they value his opinion, they all want to run their ideas by him, he could get good deal flow. And even if the guy, that particular VC didn’t have huge exits — maybe they were still in the pipeline — he could still be a desirable quantity, because he presents well and he just has a great reputation in the community. Does that sound realistic to you?

 James: I mean, and that’s partly just a reflection of the fact that in the public markets your ability to get access to investment is based on your ability to use a Bloomberg terminal and talk to your, you know, sort of like various intermediaries.

Steve: Yeah, there’s no deal flow. That’s not the issue, right, so…

James: That’s right, and so the game is slightly different.

Steve: I think maybe I’ve given a brief intro. So, you know, these venture guys, they accumulate the capital that they’re going to invest, they need to talk to lots of entrepreneurs, they need to decide which ones they want to invest in. There’s typically a lengthy due-diligence process — sometimes not so lengthy if there’s a kind of bidding war, the startup is really hot, and people just want to get their money in — but typically it’s a fairly rigorous due-diligence process where they meet the investors, maybe they — sorry, they meet the entrepreneurs — and they actually look at the technology a bit, and then maybe they even talk to customers that are using the product. There’s a whole process there. I find in my experience it’s usually all the difficult, lengthy work is done by the associates, and the partner kind of swoops in at the end and makes a decision about what to do. What does it look like from your side of things, like what’s the allocation of time you spend on, you know, diligence, trying to get access to deal flow, talking to LPs, all those different activities?

James: Yeah, I mean I think that so much of this also depends on what stage you’re investing in, right. So that when you’re investing in a company with, you know, whatever, seven million dollars of revenue, that’s been around for four years, it’s a very different thing than when I invest right now, which is, I will invest when it’s three people, there’s a promise of a market, and it may or may not work out, right. And so then so much of my time, in my world, is spent assessing the team. Some of it’s spent assessing the technology and the market, but a lot of it is spent just trying to figure out the team and their dynamics and whether they’ll work well together, and then figuring out whether their belief about some new market or some new opportunity is differentiated or special enough that I think they’ll be able to break through the noise.

Steve: And do you ever feel — so I think you and I maybe talk a lot about, like, decision-making under uncertainty and rationality, etc. — do you ever feel like you could be fooling yourself on your ability to evaluate people and teams, and that in fact your “alpha” in that respect is a bit illusory?

James: Oh my god, all the time, right. I mean, I think there’s also a little bit of a sort of George Soros-like reflexivity thing here, where, you know, to what extent is my belief going to merit a belief of either potential buyers, or employees, or future investors, right. And so there’s a way in which I’m constantly trying to measure up whether my — I guess this is a weird and subtle point — I’m trying to measure up whether my biases match up with the biases of others, rather than some objective truth, right.

Steve: Yeah.

James: You know what I mean? Like there’s a way in which the best entrepreneurs, basically, are a little bit, you know, they’re creating something from nothing, and there’s a way in which how they do that is a mixture of creating belief and creating the sub-system, or something to believe in, that both customers and employees and investors will follow.

 Steve: I totally agree with what you’re saying, and it’s related a little bit to Keynes’s old beauty contest remark. And so, when you look at the team you’re deciding a) Can they impress future investors? Can they impress future employees who will want to join their mission and their vision? And then a totally different thing is, can they actually execute objectively to build the product, because that’s actually a little bit different from can they impress investors. It is more related to whether they can impress the employees.

James: And then correctly, some of my worst investments have been cases where the individuals or the founders are very, very good at executing themselves, but they are unable to get other people to join in and buy the mission.

Steve: So you need…

James: In other words, Steve Jobs was a terrible chip designer, and Steve Jobs actually had like relatively bad taste initially, right. But because he surrounded himself with a set of people who had good taste and were great chip designers and UI people, and he was a good editor and constantly made them feel bad enough about themselves that they wanted to work for him, you know, he was able to create something amazing. And I think like that assessment, it’s true, that’s a weird assessment to be making of people.

Steve: That’s right. So we agree that there are these different aspects of the entrepreneurial team that you want to evaluate. I’m curious if you detect cultural differences. So for example, if you go to East Asia, the standard level of aggressiveness and extroversion that Americans want to see from like, say, a CEO, is it any different in these other cultures? Because it seems to me like in some of these other cultures, the sort of standard Silicon Valley CEO would be regarded as a weird, pathological person in these other cultures, and that person would be evaluated negatively by people in that culture. Do you ever see anything like that?

James: I think that’s really hard for me to figure out, right. So, you know, I come from a very specific context and time, and you know, I’ve had historically a difficult time then assessing different enough cultures, right. And I don’t know, as a nerdy guy it took me long enough to figure out like sort of whatever, early 21st century American culture and business practices. And so I do you hear you that expectations are very different, and also practices are different, but I don’t know that it’s aggressiveness – not aggressiveness. I mean, don’t you find that sort of some of these Chinese CEOs are incredibly aggressive and ambitious and, you know, sort of quite extroverted in ways that I find uncomfortable.

Steve: They are, they can be way out on the tail just like a lot of the standard Silicon Valley guys that you meet. Let me give an example of somebody that I think you and I both know. My former postdoc who now is in AI, not just a researcher but he leads like a whole team for a Chinese company — you know, I think here his English is actually pretty good, his Chinese Mandarin is actually not that great either because he grew up in Hong Kong, so actually his main language is Cantonese — you know, in the US people would not listen to this guy at all, basically. He would come in the room and he’s this kind of quiet guy, and although he was very smart and well spoken, people just like ignored him. But it just happened that the companies that he went to after he left physics to do AI-related stuff were Chinese companies, and there it seemed like people listened to him and then quickly realized this guy knows his shit, and basically he got, you know, quite a lot of responsibility right away, and now he’s running a big team. And so that did seem to be somewhat culturally determined, but I’m curious what you think about it.

James: Well I don’t know, I think cultures are not static, right. I mean, I think there are different practices. Like one of the good things about the US is sort of the willingness to try new things, and one of the best things about Silicon Valley is that when something’s working, then everyone talks, and everyone says oh, you know what, actually this thing is working or have any of you not tried this, and that extends to both technology practices but also business practices and cultural expectations. You saw the way that — I don’t know if this has hit your radar yet — but Amazon requires people to write these six-page memos. You know this practice?

Steve: Yes.  You mean before a meeting.

James: Before a meeting, yeah. It’s an incredibly elaborate process by which you get a six-page memo with the appendix pages sort of written down and presented. And what’s funny about that is, you know, Amazon’s been doing that for a long time. At some point, enough people talked about it, the word got out, and a bunch of my companies are doing it now. And it totally changes the balance of power, because now it is that quiet, really good writer who suddenly has a lot of influence. And you know, this is something that’s changed in the last six to twelve months, right, and it’ll be interesting to see whether that is an idea or a business practice that will continue, but I mean, I guess I’m reluctant to think of culture as super fixed or over determined.

Steve: Yeah, I’m not saying that cultures don’t change. It’s just that at a given moment a particular culture, like the one built by Bezos, might be focused on rigorous analysis and writing a plan, whereas some other one might be to totally cowboy it and get in the room and dominate the conversation, and that’s how you get your idea accepted by just, you know, force of personality and the ability to, within a 90-minute period, with nothing written down, just dominate the discussion. And I’ve been in cultures like the second type as well.

James: Well, I mean the second type is the dominant culture of like, most reality for the last, you know, since human civilization’s been around.

Steve: Yeah? If you go in a physics department or a team that’s building, you know, a big accelerator, there’s a lot of detailed calculation and planning and stuff that will carry the day over someone…

James: Oh yeah, I’m sorry, I’m just joking on the fact that like, human beings, you know,

Steve: Yeah. Yes, it’s as old as people.

 James: … historically, right. And then we’re fortunate that we live in a civilization or at a time and place where like careful, rational considerations sometimes occasionally might overrule the [unintelligible] mostly guy…

Steve: No, I totally agree. I mean, the big-man culture is the oldest one probably, right? And the geeks analyzing things using abstract symbols to come to a non-obvious conclusion and then the team agrees that’s what we should do, is very alien, right, [laughs] to the human species. But…

James: And incredibly useful.

Steve: Yeah. I would actually argue that most of the really non-trivial things being done now require the second thing, the second kind of activity, even though maybe the vast pools of money that power that second kind of activity are decided in the old big-man way.

James: Right. Do you believe… In some way I think we talk about systems in which, you know, one bad thing destroys the whole system…

Steve: Yes.

James: … and systems where one single good thing ends up like, sort of allowing the system to sort of survive. And, you know, civilization is a mix of both, right, and to the extent that we try to encourage the system in which the small good things accumulate rather than the small bad things kill all of us, then we’re probably better off.

Steve: Yep. So I wanted to come back to the SoftBank discussion. So this is probably something that people who aren’t really kind of deeply into the industry, you know, maybe don’t follow but… So my perception is this guy Masayoshi Son, who runs SoftBank, has developed like a completely different kind of high-stakes capitalism, you know, he’s really a gambler and taking risks. He’s raised, you know, like sort of on the order of a hundred billion dollar-type funds…

James: Say that again, people might not have heard you when you said that — How much money again?

Steve: I think it’s at least a hundred billion, right?

James: That’s right. Which is, you know, I think that’s more money than will be deployed in venture in…

Steve: Yeah, I mean I’ve met people who are like, I run the sovereign wealth fund for X, and I control, you know, kind of on the order of a hundred-billion or maybe trillion-dollar type thing, but that’s very unusual, and those guys are usually extremely careful and they parcel out the money to many, many different PMs. Whereas in this case, SoftBank and this guy Masayoshi Son — a hundred billion dollar, huge bets, where he’ll go to some company — like he might go to Lyft and say, we’re going to give you five billion, you have to go big, and you have to just destroy your competition with this five billion, and if you say no we’re going to go over to Uber and give it to them — is that a caricature of the kind of thing that happens with those guys? Is that at all realistic? That’s what I’ve been told.

James: I mean, I wouldn’t be surprised if they have sharp elbows, and certainly a bunch of journalists reported on ways that they’ve created systems in which they, you know… Yeah, so things like that. Although I honestly think like that sort of bet is only possible today. And it’s only possible to deploy that kind of capital today because everything is so connected and, you know, sort of your ability to expand into different markets is really, really fast. And you know, the hope is that that continues to be reality, you know, that we continue to live in a world where things can grow very quickly, and people can become important players in a short amount of time. It’s not clear that will happen. It’s also not clear in what markets that’s true. If you wanted to deploy, you know, fifty billion dollars into real estate, it’s not clear you can do that, right. There’s something unique about this, like these sort of information tech-centric companies, where you can actually deploy that money and you can actually create, you know, sort of like interesting, dominant businesses in ways that you couldn’t have done before.

Steve: Right. So to unpack that a little, so you know, you could imagine if somebody gets ride-sharing down perfectly, or Airbnb platform perfectly, they can unlock some huge market, and it’s replicable from city to city and country to country to some extent, and so it’s worth it making just a ginormous gamble that you’re going to dominate or be one of the main players in a space like that. On the other hand, then, I go to some city and see like millions of scooters clogging the sidewalks because somebody had the great idea that they’re going to dominate the scooter transportation problem, and you know, so not all of the bets in this vein I think are going to pay off. Maybe a few of them will pay off. I kind of view this guy Son as almost like a performance artist with, you know, a hundred billion in capital — like he’s got these very strong beliefs, and he’s like a gunslinger out here doing this stuff, and nobody knows how it’s really going to turn out.

James: Yeah, I mean the same way that, you know, 120 years ago you’d see JP Morgan making these claims and everyone relying on him, and that mix of like really hoping he figures it out, and also half the time thinking he’s full of it, right. And I suspect that there’s a similar dynamic, and you know how we’ll know, we’ll know like fifty years from now.

Steve: Yep. I think Elon might be reaching a special moment, right, because he’s rolled SolarCity into Tesla, and now the Tesla investors seem to be souring on just giving him free money. So I can imagine a kind of critical event happening with those guys at some point.

James: I mean, I think the thing, so… So I don’t know Elon Musk but, you know, I think the thing about the business figures… I think we are better off when the business figures, and the titans that everyone talks about, are more engineers than financiers. I also think that there’s a way in which… I can’t predict whether someone will fail or succeed, but I’d much rather people be ambitious about, you know, sort of sending someone to Mars, or figuring out how to build, you know, certain types of machine-learning tools than, you know, their ambition was to sell more junk food.

Steve: I agree with you. I think we systematically, as a society, take too little risk. You can see, like if you just look at the incentives for most CEOs or university presidents, they’re not toward risk-taking, they’re actually toward actually taking less risk than is optimal. And so having people that are larger than life, you know, a few of them out there — and some of them are going to crash, maybe most of them are going to crash and burn — I think, net, it’s good for society.

James: Yeah, and I think honoring, or you know, sort of taking that seriously is something I don’t know how to do in both popular culture and business practice, right, because certainly there are lots of ways in which there are lots of high-profile failures. But I think the lesson — at least I want my kids to come out of it with — is, you know, and therefore you should try great things, right, rather than therefore you should settle for some life of, you know, middling mediocrity.

Steve: Yeah, I mean I’m trying to do that with my kids, too, just get them used to the idea of taking more risk. Now they’re lucky because, you know, their dad is in a position to cushion them a little bit. If the risk they take totally doesn’t pan out, they’re not going to starve. Other people are not so lucky, they’re not in that position, right. They have to get a safe job and they have to hold that job, otherwise they’re not going to survive.

James: And it doesn’t have to be that way, right. We can create systems to sort of find talent all over the place and encourage that. And for whatever reasons, we make it harder than it should be.

Steve: Yeah, you know, although I think, you know, America is certainly better than, for example, any European country in getting its younger people to take some risk. In China, maybe because of their special history right now, they have a generation of people that are really willing to take a lot of risk as well. There are plenty of countries where, you know, you go to France and talk to the average person there, they don’t want to start a company, that’s the last thing they want to do. They want to get a nice job with the state, you know, in some elite capacity in the state, and that’s their dream, right.

James: Yeah. I mean, I don’t want to over-valorize start-ups — although of course that is my business, and I also don’t want to  specifically make just the one thing that I do the best thing that everyone should do — but I think that that idea of starting new things in general, right, is worth valorizing.

Steve: Yeah, and it doesn’t have to be fancy technology. I mean, if you go to a good like micro-brewery or restaurant, I just see a lot of innovation in the way that people are handling the food, or dealing with the… You know, I see innovation all around me, much more than I thought I saw, you know, twenty years ago in America, so maybe I’m overly optimistic.

James: I also don’t want to… Like it’s funny, someone asked me what I believed, you know. So I’m like, I go to church and I’m a Christian, so I believe in Jesus, and I believe in software — like those are probably just like one of the things I really believe in. There’s a way in which software is magic, and it’s incredible that we’ve been able to do so much with it. These are long numbers that you would need to break up in systematic ways to do things for us, right, and it’s just so shocking that we all agreed to do it.

Steve: So James, is it software that’s magic, or is it some applied physicist who got you a million, a factor of a million in compute during…

James: I am very grateful, I am very, very grateful.

Steve: Where is the actual — like, has software gotten a million times better in your life, or is it hardware [James laughs], is it the semiconductor physics that got a million times better in your life?

James: [laughs] I’m very grateful.

Steve: I often joke with my software friends that, you know, you got a 10x, right — your coding tools and such, they’re sort of 10x better than when I was a kid — but they’re not a million — or maybe you could even say they’re a hundred times better — but are they a million times better, because you literally got a factor of a million in compute during our lifetimes.

James: That’s right, and that is sort of one of the most shocking things… You know, we do live… Are we at the tail end of the S-curve?

Steve: Maybe, yeah.

James: Right. I think a lot of people who are smarter than I am and who actually think deeply, worry a lot about the end of Moore’s law, and I think that’s probably fair, I don’t know… But at the same time, I think that there’s so much catch-up for the rest of society to figure out ways to benefit from that million — actually more than a million X, you know. So like, increase in capability and, you know, like human institutions and people — we’re still slow, right. And that ability to catch up to what the hardware guys had done, and to create institutions and practices that actually reflect that as the overwhelming sort of way of thinking — like that’s going to take time, and maybe we’ve got some time to catch up now.

Steve: I totally agree with you. I think even though — just at the pure number of transistor features per dollar kind of thing — Moore’s law is actually effectively dead right now, nevertheless there’s tons of low-hanging fruit that is still available with the compute that we have, or the sort of slowly improving compute that we’re going to have. That low-hanging fruit is still going to transform society — I mean, just look at Airbnb and Uber for example.

James: Yeah, I used to think a lot about what’s new, and now I probably think more about tech diffusion, diffusion of like either new technologies, but also new techniques and ideas. And I think that those things just move at a pace that’s different, you know, and so it’s worth looking all over the place to see where are the new ideas that should be practiced by others and connections that need to be made.

Steve: Now you mentioned software is magic, and I think your fund is machine-learning AI focused, is that correct?

James: You know, the fund… When we pitched Mike Bloomberg and team, the pitch was that we were going to invest in the future of work. And it just so happened there were a bunch of people who talked about how long it took for economic gains from the steam engine and electricity, etc. etc., to show up in the economy. And it typically is two generations, two generations of managers, because it takes time to figure out what to do with it, right. And so the pitch was that we’re now twenty-some years into the ubiquitous network computer, and now we’ll figure out finally how to take advantage of it. So that was the original pitch. And then because we started, you know, about six years ago, that was just as all the big-data projects were failing, and people were trying to figure out what to do with it, and machine learning and especially deep learning was making a resurgence. And so in that way we got lucky, thanks to, you know, some happenstance and the vision of a couple of my partners, Shivon Zillis and a few others, so we ended up focusing a lot of our attention there.

Steve: I want to get you to make some projections for me about where… so let me see, a couple of categories. One, what’s the most exciting impact of kind of narrow AI — maybe toward work if you want, automation or whatever it is — in the next say decade or two? And then secondly, what’s your projection for AGI?

James: Okay, so I have no idea on AGI.  Seriously, I go to this blog,…

Steve: I hear that’s pretty good.

James: I heard it’s pretty good, particularly if you’re looking for some projection on AGI that you can, you know, sort of repeat at some later time. But on the narrow AI piece, I think we are so early, and it’s one of those cases where we barely have figured out the outlines of how to build and deploy machine-learning models, and there’s so many opportunities right now around that. And to me, the biggest sign of how early we are is the fact that there’s no straightforward framework for building machine and learning models and deploying them, and so there’s no equivalent of agile or extreme programming. And so we’re still in mostly talky-talk world, right, where people will pontificate rather than actually build and suffer. So you only see the building and suffering in a few places out there, where people are actually depending on machine learning. So that’s one part. I think the other part that you see right now about machine-learning models and the way that, you know, we’re so early, is that people don’t have good economic models. They’re talking about how to think about it, you know, sort of mostly it is in fantasy world or macro-economic world, rather than micro-economic world. I think that’s one of the biggest opportunities out there, where we just don’t know. And so a way to think about it is, you know, sort of about forty years ago, when we figured out we should split up hardware and software insofar as how we sell it, you know, out of that came Microsoft, right, and it was a mixture of like some technical innovation but not really, but a lot of business and marketing innovation around both how to sell and talk about what they were doing. And then, you know, like ’95, ’96 it was clear that network computers were going to be ubiquitous, and everyone sort of figured someone else will host my applications, I guess. But it took another seven years before Salesforce started, and it took another seventeen years before Salesforce was taken seriously enough. And so then we’re in this moment now, where everyone talks about SAS as a subscription software service, you know, as if that was a sort of fait accompli, but it took a bunch of people to make up a bunch of things, and to be honest, not all of it was technical, most of it was like marketing and business model. And you know, as a result of that… You know, I’m in San Francisco: Marc Benioff has this really big tower that overlooks all of San Francisco… And there’s going to be some man or woman who figures out like sort of how to think about machine-learning models and AGI from a business point of view, and a marketing point of view, and will actually have a deep understanding of the economics of it. And that person is going to take advantage of some variation of a network effect that you and I haven’t really thought through. And then they’ll have a whole campus [unintelligible] [Steve laughs], they’ll like rule the world. And that search I think is still out there, we’re still figuring it out.

Steve: Yeah, I agree with you that the equivalent of what you described there — which is that it wasn’t the tech that changed that much, it was sort of figuring out what the best application was, or maybe convincing the customer to buy it, etc. — that still hasn’t happened with narrow AI and things like that. But we know of enough point applications where, if they can just get over some threshold — whether it’s autonomous vehicles, or robotics in, you know, manufacturing, call centers — you know, we can already point to a lot of places where a lot of people are employed right now, where if the machines just reach another milestone, they’re going to put a lot of people out of work. And do you see that happening soon, or is that sort of more past the horizon for you?

James: I think that that concern is real, and I don’t want to poopoo it, right — I mean, I think that is an entirely fair concern. And the rate in which that could happen, I don’t want to poopoo either — I think that that is entirely legitimate. At the same time, I think the race is not to slow that down, but instead the race is to figure out what are new valuable things for people to do. And I think that only comes out of having actually implemented machine-learning models out there working in the world. There’s this great paper I read the other day by two Stanford guys called “Street-Level Algorithms,” and their point is when you actually deploy machine-learning models out in the real world, there are lots of cases where it just doesn’t know what to do, right, and in which it will make bad decisions. And you need an entire infrastructure of people to manage those models, find out the gaps, fix the models, paper over the mistakes of the models, and all that sort of work is not being done right now by the biggest companies, because in their fantasy land you’ll replace people with models, and then everything will be great. But then if you think that way, you’ll come up with bad models and you’ll come with bad business practices. But instead…

Steve: Yeah, I want to agree with you that the world of narrow AI requires a lot of very smart humans to train the AI, figure out how to deploy it, know when to turn it off, all those things… But it is a fairly small sliver of the population that have the cognitive abilities, in my view, to really do that kind of work. So let me give you the famous horse versus car historical analogy. So, you know, we went from a world where horses were doing a lot of the work, and then automobiles came along. And two different fates we can discuss. The fate of the horse… So the horses were not able to adapt to the new world, horse population went way down. I don’t see a lot of horses employed in my town anymore, they used to be all over the place, now they’re not employed at all. The blacksmith was able to retool. The guy who was working as a blacksmith could go over to the Ford factory and start working on tires, or this or that. So, do we have a huge chunk of the population whose fate is more like the horse, and perhaps a smaller chunk of the population whose fate is more like the blacksmith, or is that a bad analogy?

James: I mean, I think that that is a fair question. I think, I guess my emphasis… You know what it is like, okay, so this is going to sound too harsh… So there are a bunch of very, very smart AI academics and philosophers and ethicists all thinking through these issues, and a lot of them are very well-funded right now by very well-meaning, very smart people. And I think that it is mostly wrongheaded. It is mostly wrongheaded in the way that, if I were in France in 1880 and I was building a new engineering school, and I wanted to figure out how to deal with bridges, you know, and I wanted to figure out the ethics of bridges [Steve laughs], if I was really smart I’d say, you know what, we shouldn’t build new bridges because, fifty years from now, a country next to us will have these things that look like horses but they’ll move faster than horses and will have armor, and they’ll move very, very quickly, and they will thus be able to invade Paris, right. And thus we should focus on the issues of security of bridges, right, rather than just figuring out how the heck do we even get bridges to work, right. And I think we’re still in a world where we’re just trying to figure out how the heck do we get models to work, how do we get labeling to work, how do we get the economics of replacing models to work — all those sorts of questions we kind of need to figure out before we’re able to really speculate in smart ways about like the actual profound implications of that. And I think it’s dangerous. I also think it’s very dangerous to anthropomorphize these models, and to treat them as if they have like deep human agency. Right now, one of the ways I talk about AI is, you know, sort of it’s bureaucratized collective intelligence. [Steve laughs] And then you talk about that and it’s like, that’s much more boring, but it I think it actually captures a lot of what’s actually happening with these models.

Steve: You know, one of the most interesting things to me about AI is that oftentimes the actual kind of “reasoning” or inference that it’s engaging in is very simple — it’s just adding up some numbers or something like that — but it’s able to track a thousand numbers at a time, whereas no human can track a thousand numbers at a time. So in the case of…

James: That’s right. There is this ability to like… The economics, the fact that it’s now cheap, and consistent, and able to deal with much higher dimensional space than we are, I think that that is very interesting.

Steve: To me, that’s the biggest factor right now. Now eventually you may get much more subtle models being learned and things like this — and I think already, for example, in the case of like AlphaGo that’s true — but a lot of what AI is doing — like figuring out what ad to show you, or optimizing something for Amazon, or even in case of genomics, predicting some phenotype —the actual operation that it’s performing is very simple, but it is tracking like a thousand different genetic variants at the same time.

James: That’s right, and thus able to do unimaginable things, things that we wouldn’t have been able to do before.

Steve: Exactly, and no human can really follow it very well. But it’s not that intricate, whereas probably what AlphaGo is doing inside is pretty intricate.

James: Yeah — I mean yes, that is right. But at the same time I feel like I would much rather lots of people spend their time trying to figure out how to make things work right now, and figure out the implications of what we’re dealing with right now, rather than… and solve the problems right in front of us, of which there are so many, right. You know, you can tell when you go to some business and they’re really worried about like how often to fix their models, versus they’ve got some very theoretical thing about like the CEO watching too many episodes of The Terminator or something like that, right. And I think like there’s a way which people are afraid to test and try and come up with new solutions. And there’s also a way in which, by anthropomorphizing and treating these AIs as like agents or like sort of other beings, we concede power to them in ways that we shouldn’t yet, right. You know, there’s a widespread sort of misapplied or badly described version of AI right now, where people say well, you know, the machines are, you know, sort of just making predictions, and we get to make the decisions, or something like that, right. And I think that although the economic insight there is right, it makes it too comfortable for people, and it elides the actual problems that we’re trying to figure out right now, of which there are many. That’s where all the excitement is.

Steve: So when I first came to Michigan State seven years ago, we were still in the era where you wouldn’t use the word AI, you would just say machine learning, because if you said AI, people would still remember the sort of previous AI winter, when those guys made a bunch of aggressive predictions.

James: A friend of mine reminds me all the time that when he was raising money six years ago, I tried to get him not to use the word AI.

Steve: Yeah, I once said this to a Chinese guy who was starting his company that’s now worth, you know, like a billion dollars, and I said don’t use AI, people think you’re a bullshitter if you use AI — this was a while ago — and he just looked at me like I was crazy. And of course he used AI, and now he’s a billionaire and I’m not. But in the time that I’ve been in this job, we’ve gone from don’t use AI because you sound like a BS artist, always be careful and say machine learning, to now people who have really no idea of what is going on inside any of these systems will just routinely talk about AI — AI is going to help us do this, well we’ve got to watch out for AI — and so I feel like there are quite a lot of business leaders and educational leaders who don’t really have a good feel for even what we mean by AI now, but they’re using the term constantly.

James: Yeah, okay so I’ve got many, many thoughts on that. I think there’s like the economic ones, but my one pitch is for anyone who’s a senior executive and interested in using this term and secretly feels uncomfortable, send me an email. There’s a guy who teaches this course only for senior executives and for like congressmen, in which he basically spends three hours, and he has them walk through and build their own machine-learning model. And then in that process, they see how fiddly and temperamental it is, but they also get that magic of, but it actually works, how amazing it is.

Steve: Right.

James: And at the same time, they experience how badly it screws up when you have data leakage and you sort of like accidentally misattribute a bunch of factors, right. And so I think that sort of nuanced view of what actually happens, that is definitely the frontier for most business executives and like political figures, but the ones who figure that out and understand it, like they’re the ones who are going to like be the next, you know, Marc Benioff or whatever.

Steve: Yeah, I think that’s extremely valuable that you said it’s like a three-hour course or video or something?

James: That’s right. It’s a three-hour course where basically they modify — you know, it’s not that complicated, right — but they make some modifications, they sort of classify a bunch of images, and they see how well it can work. And they understand that part of it at its core is really complicated, and part of it at its core basically is fairly straightforward economic change, for suddenly this thing which we thought people had to do, or we thought was really expensive, becomes really cheap, and really, really fast.

Steve: That’s great. It reminds me a little bit of, there’s a famous textbook and course at Berkeley called Physics for [Future] Presidents, in which they teach…

James: Ha ha, I’ve never heard of that. That’s a great idea.

Steve: Oh you should check it out, there’s a great textbook. And it’s basically just simple physics concepts but that actually have impact in the real world. And we hope — we wished — the president would understand all these things.

James: That’s great.

Steve: So, I think we’re out of time. So going to have to end our conversation, but I hope we’ll have another one before too long.