[0:00:01] Not surprising. So, Sam, I'm so thankful that you're here today, and you were very nice to talk to some students we have backstage, which we also really appreciate. There are tons of things we could talk about in not a whole lot of time, so I thought we could kind of talk first a little bit about Y Combinator, get a little bit of a progress report, and then dive into some of the very meaningful interests that you have. Sounds great. Great. So, we were just talking backstage about this Tad Friend article in The New Yorker, which I...
[0:00:28] Found very charming in many ways. That makes exactly one of us. So one of the nuggets that I thought was interesting, and it was the fact that you put together sort of a list of goals every year. And there are two goals on there that you listed last year that I thought we could talk about. One of them was expanding into China. So I just wondered, did you make a decision on whether or not that's going to happen? We don't have anything ready to share yet, but it remains important to us, I think.
[0:00:56] Many of the most talented entrepreneurs that I've met recently have been operating in China. It is logistically super complex, but it's something that we really want to figure out and we're actively working on. Are you courting startups in China to come to Y Combinator actively? Yeah, we have been doing that for some time, quietly. We'd like to do a lot more, though. And if you do move ahead, do you sort of see partnering with an accelerator program there or just recreating YC from scratch?
[0:01:24] I mean, there may be some way to do both. I'm not totally sure yet. Okay. Well, toward that end, another objective, or I don't know if it was a consideration, was expanding YC by 2X. So I wondered if... Yeah, we've gotten significantly more ambitious there in the last year. We were able to work with 3,000 startups at once when this thing we did called Startup School, which is like an online version of YC. And I didn't think that was going to work.
[0:01:53] And then it worked. And now I think it can be a lot bigger. And I think we can figure out ways to make a whole lot more startups happen and really help them. So now I think we are sort of like being not nearly ambitious enough. And I think startups really are a super effective vehicle to make important things happen in the world today. And I think we are nowhere near the limit of how many we can help. Do you have any sort of like specific plans in place to grow YC that you could share?
[0:02:23] Yes, but it really is just, you know, at this point it's just an operational grind. We have this model, it works really well, and it's just, you know, how much can we scale it up? And, you know, every time you scale it up even a little bit, new things go wrong. So we always warn the founders in a batch, you know, YC is always slightly broken, because we're always trying to grow, we're always trying to do new things, and when you do that, you don't find the problems until you experience them.
[0:02:52] A good lesson from computer science that has really applied to YC is that you can always optimize things more than you think, and the bottlenecks are always super hard to predict. So we just try to keep growing and keep delivering a great experience. New things always break, and we find new ways to address them. That's great. Well, so many people love YC, and I've heard many of your talks, and I love the idea that, especially with this fellowship program, the idea was it's better to put $1,000 into...
[0:03:20] I'm sorry, $12,000 into 1,000 companies, then make a sure bet on a super hot ride-hailing company. At the same time, I think people in Silicon Valley are a little bit taken aback by what you've done in your near four-year tenure. I think they wonder, is entrepreneurship that scalable? I just wondered how hard it is to vet these entrepreneurs. And if you sometimes worry that not all of them are
[0:03:49] Out to change the world. And in fact, want to be founders. I mean, part of the thing about our model is, and I understand that it's difficult to, it doesn't fit well with other people's models, but part of our model is make the cost of mistakes really low and then make a lot of mistakes. So we'll find a lot of people doing stuff that sounds really dumb. Most of the time it will be. And some of the time it will sound like a bad idea and be jaw-droppingly brilliant.
[0:04:19] And really what you're looking for, you know, the very best startup ideas are the things that are at the intersection of the Venn diagram of sounds like a bad idea is in fact a good idea. And you're shooting for that little overlap. And so if you're shooting for that, you're going to do a lot of things that sound like a bad idea and are in fact bad ideas. And that's okay. You know, I think we'd rather make mistakes in that direction.
[0:04:45] Then make the errors where we say no to this company that will cure cancer or develop nuclear fusion or whatever. So we will do a lot of foolish stuff. And that's okay because we'll do some jaw-droppingly brilliant stuff. And the payoffs are such, and this was always the thing that was hard for me to wrap my head around, that we can make a lot of mistakes. We can fund a lot of terrible companies as long as we do fund the most important ones.
[0:05:12] So the model is such that you've just got to be willing to make that error. And so, yes, there are people who say that, well, is Y Combinator getting too big? And I always stop and think, if everybody else is saying this thing that YC is way too big, and I think that's what they're wringing their hands about. And my big worry is that we're not nearly big enough. We're off by a factor of 10 or something from where we should be. Maybe everybody else is right, but I don't think so. And at least so far, that would have been...
[0:05:40] Really the wrong thing for us to do. It's already been able to scale much more than I thought. I always like to think about any kind of company, including YC, is this something that gets better or worse as it gets bigger? And given our particular network effect, given how important our alumni community is to what we do, I think we really do get better the bigger we get because the network has just become extremely powerful now.
[0:06:04] Well, I was wondering about that. If there is, it doesn't sound like you think so, but a breaking point where the alums are like, I don't want another startup from YC reaching out to me to be a customer or to ask me for funding advice. Yes, that is a concern. And we have had to kind of change the software we have. We've had to change the way we work. But at this point, the alumni network is basically a self-sustaining ecosystem.
[0:06:33] People really like that. So more people in the alumni network are more customers, more users, more products that you can buy at hopefully some kind of discount. And so we will continually have to evolve as we get bigger and bigger. But university alumni networks are far, far bigger than we are today and still manage to work OK. And we have this huge advantage of we are ourselves a software company. So we are able to write really good software that makes the whole network work.
[0:07:03] Very much a trendsetter, but there are other trends that are sort of sweeping through Silicon Valley right now. One that I'm very interested in your thoughts on are ICOs. Yeah. Is, you know, we're seeing Andreessen Horowitz and Union Square Ventures and some of these other top firms investing in like hedge funds and creating funds to fund other ICO companies. What is your sort of thinking here? You know, like the question everyone wants to talk about are, is this real or is this a bubble?
[0:07:32] And that was the question everybody wanted to talk about about startups two or three years ago or 12 years ago. It's just this endless cycle. And the thing about bubbles is there is something fundamentally real and people get too excited about it at a moment in time. Because otherwise, smart people don't get tripped. I mean, occasionally they do. Like the Tula bubble, there was nothing underlying there. But, you know, like Newton, very, very smart guy, still fell for the South Seas bubble.
[0:08:02] And again, there was something real. People just got ahead of it. The internet bubble. There was something real. People just got ahead of it. And that is what I feel about the current ICOs. I think it's definitely a bubble right now, but there's something real underlying it, which is why smart people are so fascinated by this. Do I think a lot of ICOs that are happening right now are...
[0:08:30] Ridiculously silly bordering on scams. Yes, a lot of individual ones are. Some are just outright scams. But I think there will be a few that are really important. And I think this idea of the blockchain in general is more likely to be really important than not. However, I was not around in the 2000 bubble. I was still a kid. The only memory I have of it
[0:08:58] I was trying to convince my mom to buy Yahoo stock in 1999 because I read about it and thought I was going to keep going up. But I didn't have this visceral sense of what it was like to live through a bubble. And this is the first time where I felt like I've understood it. I've understood this driver where the thing that's happening is people are watching their friends get ridiculously rich. And that's driving them, just making them like, ah, I want to do that too. So they're buying and then more people are buying.
[0:09:27] This bubble is a little bit different than others because there is no underlying asset of Bitcoin. It really is just how big the collective delusion is that this is the thing we all care about. And this is a bubble that can sort of like become, it can, by force of will, it can perpetuate itself into actual existence. But this quote, I don't know, I've heard it attributed to a bunch of people, but something like, there is no...
[0:09:55] I think there's something real. I think people have been overcome a little bit by irrational exuberance, but that's why it's really hard to say. Do you welcome regulation in this area? I mean, it seems to me that if, you know,
[0:10:19] Regulators at least said you have to have lock-up periods of six months to a year or something. I mean, it seems a little bit too much like the Wild West. I definitely think it needs to be regulated. I think one of the interesting things that has shown here, though, is how much white space there is between what people want and current regulation. And so I think you see this especially in the desire of non-accredited investors to be able to invest small amounts of money in startups, which is something I personally think.
[0:10:49] I think that's cool. I think we should try to figure that out. I think other people should try to figure that out. We need regulation for that. We need clarity for that.
[0:11:18] You know, the promise of this is a really good thing. One of the trends that bothers me about Silicon Valley is more and more of the wealth creation that is happening here is not available to most people. And I think that's very bad in a society with already such wealth inequality. And if there is a way that new technology can make it practical and possible to democratize this, I think that'd be great.
[0:11:42] Well, maybe we can segue now into a little bit of politics. Sure. Which is obviously in other interests of yours, including because of this concern of this growing wealth disparity. So I know it has been said that you are not interested in running as governor of California. But I wondered if you were maybe a little bit interested in running as governor of California. I have such a good... I mean, it's sad for me because I think there are a lot of really important...
[0:12:10] Things to do in the world, and we get one life, and we get some number of hours in a day. And I am interested in politics, but I love my current job, and I think there's such important things I can do here that have a positive impact on the world. So sure, if there were no opportunity cost, I'd be interested in trying it, because I do think there's a lot of really good things one could do. But I think I can do really good things where I am right now. Okay, so you're staying put, no political office.
[0:12:39] You've talked about getting behind some candidates. That I want to do. Okay, so tell us who you would back. Well, I will tell about the kind of people rather than specifics. Okay. You can also provide specifics as well. I am really interested. One of the things I've learned about the political process is it is basically all about how much other people think you can raise. And it is controlled by a small number of donors and these very powerful but I think horribly run large political entities.
[0:13:09] And I have met a lot of really promising more junior office holders who I think would be great. They've got great ideas. They work super hard. They have really good values. But they just have no chance because they're not chosen by the existing system. And I think in the same way that Y Combinator found out there was a huge amount of value to unlock by finding unknown people that are really talented but don't have a network and giving them resources.
[0:13:38] I think there's something cool to do there politically. And I think a lot of people are thinking about that. It's something I want to try. It's going to be a small experiment for us next year. But if it works, I think it's something we could really scale up. And I think, unfortunately, right now, the left is pretty disorganized, doesn't really have a great economic plan, and it would be really good to get a new team of people that really believe in some of these things.
[0:14:09] So you've got these candidates that you'd like to support, or not candidates, but people you'd like to possibly turn into candidates. Are they from the tech world? Not necessarily. I think some of them should be, most shouldn't. I was going to ask, I mean, I just wondered, sort of talking on the presidential level, if, for example, having a billionaire businessman right now has...
[0:14:33] Paved the way for another one or completely blown up the past? That is hard to say. I mean, I think very little about this current situation generalizes to anything else for so many reasons. But I do think that the way this government system was supposed to work was that people are not career politicians but have a career and then you do service to your country and you do it for some period of time.
[0:15:01] And you may go through a few different levels. And then you may go back to a career. You may run again later. But I worry that we, instead of this kind of participatory democracy, we have this, what I think is, fairly broken kind of career politician track. And control of that track is in the hands of a few rich, powerful people. And I think trying to say, hey, let's get a lot more people in the door would be a good thing to try. Right.
[0:15:30] Okay, great. So we're almost out of time. I wanted to talk to you about basic income. I don't think we're going to have time, but I would love to get your thoughts on AI. And if there's any update you can give us about OpenAI. Yeah. We had a senior VP here from Google earlier who said, you know, tech leaders talking about the dangers of AI is, you know, I think he said borderline irresponsible. Rodney Brooks, who founded Rethink Robotics and is considered the...
[0:15:58] Father of Modern Robotics, and I interviewed him last month, said the same thing. He said, anybody who works in AI knows it's dumb. So, look, I think you all, two general comments. I think the most important thing when someone is giving you advice or a statement like that is trying to understand their biases and where they come from and why they will say something that is either not true or not really what they believe. And, you know, certainly it's possible to imagine why
[0:16:28] People who run maybe the most advanced AI company in the world would say, oh, you don't have to worry about it. I don't think we need to panic about it. I do think we should be thoughtful about it. I think already at the level of AI technology we have today, there is huge reward and huge risk. And you see examples of this all of the time. There have been a number of examples, just with the most basic machine learning we've had about...
[0:16:57] Bias inherent in these systems. Autonomous cars crashing. I think autonomous weapons are already something to be concerned about. And this is moving really fast. If you study the history of technology, there are many examples of very smart people saying two years before something is invented that it's never going to happen. In fact, there are some examples where the people who invented themselves say that. One of the Wright brothers, I forget which one.
[0:17:26] Two years before Kitty Hawk said that he thought that human flight was at least 50 years away. And then after that happened, he realized he was never going to make negative technology predictions again. And exponential curves are always hard. If you stand on an exponential curve, you look backwards, it looks flat. You look forwards, it looks vertical. No matter where you are on the curve, that's what happens. So that's really hard. And I think we evolved not to be good at understanding exponential trends.
[0:17:56] And there are many exponentials here. We have exponentially increasing hardware. We have exponentially increasing algorithms. We have an exponentially increasing number of brilliant people entering the field. So the thing that I think is weird about the whole conversation about is AGI going to come and do we need to be worried about it is almost no one is arguing that AGI is never going to happen. Is it 5 years or 10 years or 20 years or 30 years away? And all of that...
[0:18:23] Compared to like the long arc of human history is an absolute blink of an eye. And so whether it's five years or 30 years or 100 years, I believe it is going to happen. I believe it will be the most important development yet in the history of technology, probably the history of humanity. And I think a wide conversation about how we want that to work, how that's going to change society, how we make sure that benefits all of us, how we make sure that happens safely.
[0:18:53] I'd rather be a little bit too early than a little bit too late on that conversation. And in my own work at OpenAI, I have generally found that the milestones have come faster than I expected. I would love for you to come back soon and tell us more about OpenAI. That would be fun. Thank you, Sam. Thank you. So appreciative. Thanks for having me.