[0:00:16] So we're ready to conclude this amazing day. I think you'd agree with me. And maybe it's a good opportunity to thank Yasmin again. And for the very, very last session of the day, we have Sam Altman from YC, Y Combinator, which, as we all know, redefined what it means to be an accelerator and is now going beyond that. I'm sure we'll hear about that.
[0:00:49] And he's going to be interviewed by Michael Eisenberg, who's one of the most prominent VCs in Israel. I've known him for 20 years, plus, minus. And he's just closed his second fund at his new firm, Aleph. Anyway, I'm sure it's going to be a very interesting session. So with that, thank you all for hanging around for this session.
[0:01:24] Sam and I concluded before the session started that since we're the last thing between you and happy hour, you'd all be much happier if we concluded the panel a little early. So that's just to set expectations if we don't fill the full hour. So Sam, thanks for coming out. Thank you for having me. Just to start, tell us about what you think Y Combinator was when it first got started and how it changed and what it is today and going forward. Y Combinator started with what at the time was a radical idea that you could fund
[0:01:54] A lot of startups with a very small amount of capital. You could fund teams that didn't have business experience, but people that knew how to build products, and that the demo was more important than the business plan. That was almost the entire set of founding ideas for Y Combinator. And at the time, no one thought it would work. So YC was started in 2005, and they funded eight companies, one of which was my company, Loot.
[0:02:23] There were all these other things that became obvious really quickly, the most important of which was this idea of a community. And that if you funded startups together and the startups were loosely affiliated but not legally bound to each other, that you would have the shared sense of community and peers and a support network, which is one of the things that you miss most as a founder. And so pretty quickly after that, Y Combinator changed a little bit. It grew, and it became
[0:02:53] At its heart, a community of people trying to build new things in the world, which I think is what Y Combinator is today. Today we fund about 250 startups a year from all over the world. We fund them at different stages. We have a late-stage growth fund. We fund fundamental research, which won't make its way into startup for years or decades to come. But we try and take a very long view on being an organization and a community of innovators. So I'll come back to the research.
[0:03:20] Some would argue that the growth in quantity has resulted in a reduction in quality and that the signal-to-noise ratio has gone down. How would you counter that, or would you? Only VCs ever argue that, as far as I can tell. I think when I got into YC, one out of five startups that applied got in. I think 40 people applied for eight spots.
[0:03:46] So 20% acceptance rate. I think now it's unfortunately about 1.5%. We'd like to figure out a way to fund a lot more startups. None of the companies in the batch that I was in went on to be worth more than a billion dollars. Reddit still will be, I think. Since then we've had 11 billion plus dollar companies. I think that is generally a myth spread by VCs that
[0:04:16] Are unhappy about YC's power in the ecosystem or whatever. I don't think any of the founders that go through the program or would like to go through the program feel the quality has gone down. I think we've been able to, I think the quality has gone up too much. I think we have failed operationally to scale as quickly as would have been good. And so the bar is higher than it should be. It might mean we're, because the bar is so high, it might mean we're missing companies that could be great someday.
[0:04:47] It takes a very long time is the hard part. Quality and valuation eventually roughly converge, but it takes 5 or 10 or 20 years. And so it's very hard to measure quality on the short term. The lag is really bad. But I think the rough metric that we use, because we like something numeric, is just the total valuation of each batch as time goes on.
[0:05:15] Going back to that point about venture capitalists, is there now, in your view, a contentious relationship between YC and venture capitalists? Well, we don't think there is. I think some VCs definitely felt like or feel like they'd like to go back to the good old days when VCs had a lot of power. However, I think this is a be careful what you wish for thing. I understand that YC has been a force in the ecosystem that has shifted more leverage to founders.
[0:05:41] But I think this has been great for everyone, including investors. And if valuations in the seed stage have gone up 2x, but exit valuations go up 5x because we're able to make these companies better, disseminate more knowledge, build a community, that'll be really great for everybody. Are there any billion-dollar companies in the ones you mentioned that don't have venture capital investment? I mean, this is... No, I would say no. I think there will be one or two really oddball examples that come up over time.
[0:06:11] That never officially raised a venture. But even then, in those cases, the ones I can think of, the founders eventually sold some secondary to a venture firm. So there's like a VC on the cap table somehow. But no, one thing that I believe is that if you're trying to build a really big company, it is in your interest to raise capital. Use that to grow more quickly than you could otherwise. And have a great advisor on your board. So we definitely advise our companies to raise capital.
[0:06:40] But we don't require it. Is it only valuations and founder friendliness that you think creates tension with the venture community? In your word. I don't live in this ecosystem, so it's easy for me to ask the question. Or is there something more than that, maybe? I really don't think there's that much tension. But I guess I could be oblivious to it if there is. If you think there is, I'd be very curious to hear what. Like I said, I'm in Israel. It's far away. No, I think we have a generally quite symbiotic relationship.
[0:07:11] What's your role now at Y Combinator? So I'm one of the partners. I advise startups. I spent today advising YC startups. And that is what all of us do most of the day. Everyone thinks that it's this very glamorous job. And actually, day to day, it's a total grind, but one that I really enjoy. There's probably not that many people who would want your job because it's such a grind. You know, there are two ways to do the job. There's one which is like,
[0:07:41] Go to a bunch of parties and network. I don't even know what these people do. And there is the way that I think we do it at YC. Paul Graham has this analogy that he always used, which is, I don't know where it comes from, but you just knit and knit and knit, and every day it looks like you're doing the same thing and not making much progress, and you look back after a year and you've knitted this mile-long scarf. And I think one of the things that has worked well about YC
[0:08:09] It is kind of a treadmill. We just do the same thing again and again. We try to meet founders. We read applications. We advise companies. And it is sort of a grind, but it is a rewarding grind, if that makes sense. And although any given day can be repetitive, the leverage and organizational output in the end is quite high. Is that the full scope of your role? No. So I'm also the president of YC Group, which is we have a number of things that we do. We have the core program that people know us for, the research lab.
[0:08:39] The Hacker News, the Continuity Fund, which is a $700 million growth fund we raised last year. Startup School, which is our outreach and education thing. But, you know, a surprising amount of my time is still advising startups. Talk about the Research Institute. So our mission is to increase the amount of innovation in the world. And in general, I think startups are the best way to do that. I think startups...
[0:09:09] Are phenomenally effective at getting things done and getting new things done. I think that there is a powerful way in which the joint stock corporation was one of the most important inventions of the last several hundred years. No one ever talks about it because it's not a piece of technology you can look at. But it has been a model to align a lot of people and capital to do incredible things. And so I'm super pro startups. I think startups are a great way to get things done in the world.
[0:09:36] There are some cases where it doesn't work. And for those, we still want to be able to have that be part of our community. And so we have the research lab. Those could be things like superhuman AI. I don't believe should be owned by a single for-profit company because I think that will be misaligned with the good of everyone in the world. So OpenAI, which was the first project we ever funded, is trying to build that as a nonprofit. Our basic income project.
[0:10:03] If OpenAI is successful, if a lot of YC companies are successful, we will see a reshaping of the economic landscape in the world, and we'll see massive job destruction and job creation and a very, what I think will be rocky transition period. So I think it'd be really good if we could experiment now with new economic models, like a universal basic income where everybody in the world gets enough money to live on. That obviously does not make sense for a startup, but that I think is important research, and I think we need to do that now.
[0:10:31] Because 10 years from now, when we need to make a policy decision, it would be good to know if that works, if people are happy, if they're fulfilled, if societal fabric holds up, changes. We'd like to find that out. Why should this happen in academia? Why should it happen at Y Combinator? You're a for-profit corporation, right? Why academia is not doing more of this, I can't answer. I will just say that they're not, and I think it's important work. Aren't you a for-profit corporation, though, also?
[0:10:58] The Research Institute is a part of YC, which is a for-profit corporation. You even said kind of, if our startups are successful and the AI lab is successful, then good things will happen. No, Y Combinator, the two funds are for-profit funds. We're trying to make money on those. But YC Research is a separate organization and a non-profit. But you're involved in both. I am involved in both. But so are lots of other people that do for-profit and non-profit work. So talk about, for a second,
[0:11:27] Your attempt to shape social policy, particularly around universal basic income. So if you go back to the onset of the industrial age, people also thought there'd be nobody working anymore because machines and tractors and everything else would do all human work. And here we are again, but yet more jobs were created in the last 100 years than ever before in the history of mankind. And one would imagine the same thing will happen in this technological revolution. Oh, I think there will be new jobs. I think human demand and creativity and desire is limitless.
[0:11:57] I think they'll be new. I think they won't look much like they look today. I don't pretend to know what they will look like. All I can say with any confidence is that I think the jobs of today, many of them will go away, and then people will figure out new things to do. And I don't feel like I could predict what they'll be any better than somebody on the other side of the Industrial Revolution could have predicted programming would be the important job. I think it would be that different. So I think there will be new jobs, but I think...
[0:12:26] There's going to be a rocky transition. And also, I believe that as the cost of a really great quality of life goes to zero or near zero, it's just the right thing socially to do to have a simple support net that's equal and fair and where we don't have to have all these other crazy debates that we have. So why has government not thought about this and decided this? And why has Sam Altman and Y Combinator decided that this is the best path forward for society?
[0:12:54] And why do you expect the politicians to adopt that? Because you still need them, right? Yeah, I mean, governments are. One of the things, we put out something we didn't expect to get much attention. We put out basically a job ad saying, this is research we'd like to fund. If you're interested, please apply. And it kind of just turned into this, it was like throwing a match into a field of gasoline. We didn't expect it, but it just became this explosion.
[0:13:17] Now, at least two governments, Finland and Canada, have announced that they're going to do studies. We've been contacted by many more that have said they're going to as well. So I don't think it's just us anymore that are thinking about this. You feel it's a little bit, what's the right term, paternalistic to kind of parts of society saying, hey, you're not going to make the transition, so let's give you some money. No, what I think is paternalistic is governments telling people how to live their lives and how to spend money. I think that a basic income where the government says,
[0:13:48] Here's what everybody gets. Totally up to you how to spend it. If you want to work really hard and make a bunch of money, there's no ceiling, that's great. If you want to sit around and smoke pot and play video games all day, that's fine too. Do we really think that's fine? I don't have a problem with it. I think it'd be paternalistic to think that that's really bad. I don't think we should make judgments over what people choose to do with their time on this earth is fine or not. But the reason I like the basic income is because I think it's not paternalistic.
[0:14:18] And I think it's very different than all other social services that are. You've been outspoken about the dangers of AI. You mentioned it a bit earlier. Expand on that for a second, and then we'll drop into a conversation about it. And I saw Microsoft today joined up also. I'm not... I'm not... I think there are dangers. I think AI is very high vol. And when we develop...
[0:14:49] Something that is a billion times smarter than the smartest human. I think it's just really difficult to have a confident prediction at all about what that's going to be like. And I know a lot of people do. A lot of people take very strong positions on how that's going to play out. I just don't feel confident enough in my own abilities to pretend like I know what's going to happen there. But I think it's going to be very different and very important. And I think we should...
[0:15:18] Think about the kind of world we want to live in. Do we want a single organization or person or government to have that power? Do we think that should be democratized? To the degree that we are going to design our own descendants here, do we think everyone on earth should have some input and vote into how that works? And yeah, I can see it being really dangerous and bad. I can also see it being really great. In any case, I don't think we're going to stop it. This is not a case where
[0:15:48] You know, like in the case of nuclear weapons, it wasn't actually the secrets of the technology that contained it. It was that enriching uranium took so much energy in the early days that it only took the resources of nations and only very rich nations. In 1943, I think the US, like 22% of electricity in the US was being used to enrich uranium. AI is going to be very different than that. You know, this is something that will be able, possibly developed by a small number of team, a small team.
[0:16:17] Less than 100 people. They'll probably need a lot of compute power, but maybe not. It's like every technology curve, right? Nuclear weapons have proliferated, and now AI will take some time to spin up and then proliferate. My guess is that it will proliferate much faster, and so we need to think about a different strategy for AI than nuclear weapons. Is government part of that strategy? Because that's what's contained nuclear weapons, right? Government has managed to somehow... Well, that was my point. I think the thing that contained nuclear weapons,
[0:16:49] was not a technological secret, which I think is what most people would say it was, but it was really just this luck of physics that it's really hard to enrich uranium and that only people with massive resources could do that. And it was very difficult to hide because you need 22% of the electricity in the US. That would argue, by the way, that as nuclear weapons proliferate, which they are right now, we're going into a very, very destabilized society, right?
[0:17:15] Yeah, I mean, I'm certainly scared of nuclear weapons. You know, when Obama said last year or whatever was the thing that keeps him up at night is a nuclear bomb going off in Manhattan, I was like, yeah, I think people should be more nervous about that than they are. You know, it's a very complicated worldwide geopolitical game. There are other things besides this now, and clearly now we have technologies. It's very different than 60 years ago. We have technologies now to make it much easier to enrich uranium.
[0:17:42] But we have a whole bunch of other systems now. We've had those decades to build super advanced detectors and imaging, and we have a very complicated geopolitical standoff of sorts. So my point was that that happened slowly, and that was good. And I expect AI to happen very quickly. So what's more dangerous, AI or nuclear weapons?
[0:18:05] You know, I'm terrible at those games, which is worse. All I can say is I honestly don't have confident probabilities on that. What's more dangerous, AI or ISIS? AI or Donald Trump? Yeah. Interesting one, huh? So when you look at the speed at which AI is attacking different industries here, what are the first ones that you think...
[0:18:35] Well, the thing that we have working today is perception. Three years ago, humans were better than computers at effectively all perception tasks. And today, it's basically exactly the opposite. You know, computers are better than humans at basically all perception tasks. It wasn't that long ago that Siri couldn't understand anyone.
[0:19:02] And now Siri can understand people I can't understand. Crazy background noise, thick accents, no problem. It wasn't that long ago that computer image recognition was a joke. Now computers can write captions for images. It wasn't that long ago that a human doctor was better than an AI doctor at looking at radiological images 100% of the time. And now the AI radiologist is far better than the human radiologist.
[0:19:30] Perception is only one class of problems, but it's a large class. And I think at Y Combinator, we have been incredibly excited to fund narrow vertical AI companies solving perception problems. What's one you're most excited about? Too early for me to pick one that I'm most excited about, but to just pick a few. One is what I just mentioned, looking at medical images. Another one is like a sort of AI lawyer looking at case law.
[0:19:58] And basically the entire body of legal text to do output. One of ours that got acquired, our biggest acquisition of this year so far was Cruise, self-driving car company using deep learning. But the list would go on for quite a while. I think this is a very fruitful area. And it's not theoretical. You can start great companies here today. Why do you think, and this is a good segue to another topic I want to talk about, why was Cruise acquired at the stage it was acquired at, A, and then B,
[0:20:28] Talk for a second about why Silicon Valley becomes Detroit, so to speak, in automotive information, and what that means for job creation and job destruction in other places. Well, they were acquired because the founders chose to take the offer, which I suspect was a mistake. There's a long history of companies, of founders selling too early. But, you know, I hope Kyle will be the next CEO of GM years or a decade down the line. I think that would be great.
[0:20:57] But you're unhappy about the acquisition. I'm happy for Kyle. I never go against a founder. I think it will turn out to be the suboptimal economic decision, but it may be the right decision for Kyle and Dan and their team personally. And we're always back a founder that wants to sell, or not sell for that matter. One thing I have learned is you never go against a founder who wants to sell or doesn't want to sell, either way, whatever the founder wants to do. I told him that I thought he might regret selling for this price, but that if he wanted to do it, I'd support him 100%.
[0:21:26] In terms of Silicon Valley becoming the next Detroit, well, I've been to Detroit recently. I hope not. I actually think there's an important lesson there, which is the world changes very quickly. And Silicon Valley should not try to become any static thing. But the best thing I could hope for for Silicon Valley is to always be on the front edge of wherever the most interesting...
[0:21:55] New technology is being developed. So you moved from St. Louis, right? And we just said that AI and Silicon Valley innovation is eating Detroit. And then you talked about medical, and medical and pharma is in New Jersey and places like that, and that becomes Silicon Valley. Roll this forward for me. So no Detroit, no New Jersey, no... Well, my guess is that the Silicon Valley ethos will spread around the world, and the whole thing will just become much more diffuse. And you will see...
[0:22:23] Technology companies get started everywhere because technology will seep into everything. So my guess is, I think it is, well, I hope Silicon Valley does not become Detroit, but Detroit becomes more like Silicon Valley. You're bringing everyone here, right? They're not going back to Detroit because you were just in Detroit and said you hope it doesn't become that way, and they're not going back to New Jersey. No, they do. They do. We bring people here during our program, but a huge number of them go back to where they're from around the world afterwards. We view ourselves like a university.
[0:22:52] You travel, you do your studies, and you go back, or you go somewhere else. How many meaningful companies are really created in these geographies? Silicon Valley is sucking all this in. I don't know the exact numbers, but I think there are many really substantial companies not in Silicon Valley, and certainly not in the U.S. Someone mentioned, and I keep meaning to look at the data, that this is the first year where there were more billion-plus dollar companies created in China than in the U.S.
[0:23:23] But even within the US, if you look at the recent very large companies, Oculus, Snapchat, those are both in LA. Seattle, I think, is an incredibly rich startup environment. How many of your billion-dollar companies are outside of Silicon Valley? There's one that I think will cross that threshold soon that I was just talking to that's top of mind. It's in the Midwest. But other than that, I believe they're all in the Bay Area. Aaron, do you know if any are not?
[0:23:53] I think they're all on the bay. I think of our 10 billion plus dollar companies, they're all here. I once had a teacher say that if you can't make generalizations, you can't come to any conclusions, even though we hate doing it. So you meet entrepreneurs from all over the world, China, India, Israel, the United States. Talk to us about the difference from an entrepreneurial perspective and from a network perspective of the entrepreneurs you meet from all those different places. I have been surprised to the degree that
[0:24:23] Over time, they have become more alike than different. I think one of the cool things about this community that we're all in of technology and startups is that it has very quickly become this global community. We now fund more founders born outside the U.S. than inside the U.S. I taught a class called How to Start a Startup that was viewed more outside the U.S. than inside the U.S. There was a time when I would have said, and this would have been many, many, many years ago, where...
[0:24:52] You know, Chinese entrepreneurs are different, Israeli entrepreneurs are different, the US entrepreneurs are different. But at this point, and maybe even a little too much of this, the entrepreneurs from around the world feel pretty similar. There must be something less politically correct than that. Come on, there must be some differences. I mean, literally we were talking about this recently, how there's just less and less over time. I think the US entrepreneurs are less intense than Israeli or Chinese entrepreneurs.
[0:25:22] But that is good and bad. So I can't say that that necessarily cuts one way or the other. But it's much harder now to make generalizations than it used to be because we have kind of converged to this one startup culture. So if we go back to talk about some of the earlier companies started at Y Combinator versus now like Cruise, it feels like there's been a transition from leaner startups or lean startup, the Eric Ries.
[0:25:51] A book to something that is now called somewhere between full stack and thick startups that are really reaching into the depths of hardware commodities. Investor in WeWork, right? Same thing, four walls and buildings. Take us through how you think the funding model changes for that and the model for innovation changes for that and the locus of innovation changes for that. You know, there are all of these different terms like full stack, light stack, lean, thick, and I have never...
[0:26:20] I don't think I've ever really truly internalized what those mean. What I want is to invest early when the valuation is low in a startup that will go on to be worth tens of billions of dollars or hundreds. That would be great, but it hasn't happened yet. And I think whatever way a particular startup gets there is okay. And this idea that there's only one strategy I think is just wrong. I think they're really different, but they're all really valid. And I think it is wrong to dismiss Facebook as not being valuable because they started just with
[0:26:49] There's a PHP app on top of servers, and I think it is wrong to dismiss Rigetti Quantum Computing because they have to fab their own chips and that takes hundreds of millions of dollars. Again... No, no relationship. I don't think. But what I think is really important is, is this the right strategy for this company? I think there are some invariants.
[0:27:15] That makes startups successful. Which is, for whatever constraints the startup has, are they being as frugal and fast as possible? You don't see that change much, but the constraints are wildly different. Is the culture right? Is the startup picking a good initial problem for a real customer pain? And have they found some customers that desperately want what they're building? But beyond this, I think many different strategies can work. And the mistake is to sort of subscribe to an ideology. The mistake is to say,
[0:27:44] Lean startups are right, or thick startups are right, or whatever. And the right way to do is, for this company, building this product, what is a strategy that makes sense? What are some of the attributes you see in the entrepreneurs that are in your batches that you know are going to succeed? That's a great question. The one that has surprised me the most, although now when I say it, it sounds so obvious, is determination is the biggest predictor of success.
[0:28:13] 10 years ago, I would have said intelligence or product or something like that. But these startups take so long and things go so wrong and they're so dynamic and they evolve so much that determination, adaptability, intensity, effectiveness, these are really important. The product strategy is also really important. So, you know, there are a lot of founders that build a product and it's kind of cool. And you ask them, like, how, tell me any plausible story.
[0:28:42] How do you ferret out the termination in the YC application process?
[0:29:10] I knew the answer to that. We'd be much richer. That is really hard to do. I have been studying that for years now, and I feel like I've only gotten a little bit better. You look at what the founders have done in the past in their previous jobs or school or life, you look for some evidence of exceptionalism. But that's really hard. I'd say I'm still a student at that. What did YCC and you that they took you into that first batch? I have no idea. I'm very lucky they did, though.
[0:29:40] No idea? I was... Maybe the move from St. Louis. Anyone who can make it from St. Louis to Stanford? You know what? I think actually what it was is Paul... I'm not sure. It's pure speculation. Paul tried to... I had to move my interview day because I had this thing at school, this business plan competition, and we really wanted to win it because we needed the money. So I had to move my interview to the next day. So I emailed Paul and told him that, and he was like, eh. Paul gets very annoyed with anything logistical.
[0:30:08] And so he was like, eh, just don't even come. You're a freshman. You can apply later. And I just said, well, I'm going to come. I'll see you tomorrow. And I showed up. So I think that's probably a piece of it, but I honestly don't know. Since this is an Israeli and California gathering, if you were going to impart one piece of advice to Israeli entrepreneurs at the YC stage first and then at a later stage, what would it be? We fund lots and lots of Israeli entrepreneurs.
[0:30:38] Some in this room. Nice to see all of you. And they're doing great. I think they've had plenty of advice imparted. I mean, I'd be shocked if Israeli entrepreneurs are not a continually more every year important part of our program. Since it's Israel, you have to ask the opposite question, which is, what's the critique you'd give? Again, I understand the desire to have a specific comment about Israel at an event about Israel.
[0:31:08] Am I trying to get you to vent? No, no, no, no, no. But I really don't have one. I mean, I'd like, if we could get a lot more Israeli companies, we'd be really happy. When was the last time you were in Israel? I was just thinking about that. It's been many years, but I'm going for New Year's this year, so I'm looking forward. Any questions from the audience? Speak loud. We do that today. We have a computer program. We call it Helm 9000, and it reads every YC application in three seconds, and it gives us a predicted score.
[0:31:46] Pretty good and very helpful to our workload. No other questions? Everyone really must want alcohol. People want a drink. You can hear a pin drop. Sam? No, it's all good. No problem. I don't want to stand in the way of drinks. Thank you for having me, though. Thank you. Yeah.