{"url": "https://www.youtube.com/watch?v=cKpmHoJfO5g", "page_url": "https://www.youtube.com/watch?v=cKpmHoJfO5g", "title": "Mosaic Ventures: Fireside chat with Y Combinator President, Sam Altman", "show_or_channel": "Mosaic Ventures", "date": "2016-11-15", "duration_s": 4059, "type": "interview", "speakers": ["Sam Altman"], "confidence": "medium", "notes": "Mosaic Ventures fireside chat", "people": ["sam-altman"], "id": "2016-11-15_mosaic-ventures-mosaic-ventures-fireside-chat-with-y-co", "audio_duration_s": 4059.2428125, "worker": "jc-w23", "transcribe_s": 170.1, "model": "large-v3", "segments": [{"start": 0.0, "end": 29.68, "text": "So it's a privilege on behalf of Mike and myself from Mosaic to welcome Sam Ullman here. He's a man who doesn't need a lot of introduction as he obviously has a very significant persona in the startup world as the president of YC. But just very quick sort of highlights. Went to Stanford, dropped out, did a startup."}, {"start": 29.68, "end": 59.34, "text": "Raised significant amount of money from some of the best investors in the world. Sold it successfully and moved on to running YC, which is the sort of preeminent accelerator on the West Coast. And as part of that leadership role, is taking the business in a number of new vectors, not just relating to seed financing companies, but many other interesting things that we'll talk about shortly."}, {"start": 59.34, "end": 83.92, "text": "Including AI, which is an area that my partner Mike has been spending a lot of time on and I think we'd like to get into some detail on. But some other very big, important and necessary startup areas including energy and biosynthetic life forms and stuff that startups should be doing."}, {"start": 84.08, "end": 112.27, "text": "The purpose of this was really in a smaller group, a more intimate group, we're doing a larger session upstairs in the bigger theatre with proto-entrepreneurs later this afternoon. But for both YC alums, and there's a number I think that Sam's recognised who've gone through the programme over the last decade, as well as some of our peer investors and founders in the London ecosystem, to, in a more intimate surroundings, have a chance to engage with Sam on some of these important..."}, {"start": 112.27, "end": 141.71, "text": "Developments in the world of the sort of third industrial, third software revolution that the planet is currently going through and trying to look at those developments with a bigger picture. With that, maybe Sam, can you tell us a little bit why you're in London as part of a sort of whistle-stop European tour on this trip and how you see"}, {"start": 141.94, "end": 170.9, "text": "The role of YC not just as a important player in Silicon Valley but over time potentially on the global stage is a good place to start as you introduce what you do. So we don't view ourselves as a Silicon Valley investor. We invest around the world. I'd say half of our companies come from outside of California at this point, maybe more. And our goal, our mission is just to increase the amount of innovation as much as we can."}, {"start": 170.9, "end": 198.74, "text": "And I think if we limited ourselves to Silicon Valley, we would miss out on a lot of important work. And clearly London is at the top of the list at this point of new startup hubs around the world. We have been very fortunate to fund many companies from London. It's nice to see some familiar faces in the room. And I expect that that trend will keep going and that we'll spend more time here and we'll have more companies from London come over and spend a few months with us. But we're here to..."}, {"start": 198.74, "end": 229.06, "text": "Try to meet the next generation of great founders and hopefully a few of them will decide our program is worth considering. Maybe as we begin to understand how YC works, can you just give us a flavor of some of the metrics of your pipeline? How many people apply to YC in each batch? How many of those order of magnitude come from Europe versus Asia?"}, {"start": 229.06, "end": 258.64, "text": "I suspect it's the widest pipeline or funnel of any investor on the planet. Yeah. So we run two different programs, what we call YC Core, which is in person with us in Silicon Valley, and then our fellowship, which is remote. And I think the total for those two together in terms of companies per year is 25,000 to 30,000. And of that, maybe 15,000 to 20,000 coming into the two core sessions per year."}, {"start": 258.64, "end": 291.2, "text": "We don't have perfect data. I mean, we could collect it, but we don't have it on where they come from from around the world. But I would guess that more from Europe than Asia, which would make Europe the biggest market outside of the US, and then a pretty long tail around the rest of the world. And from a founder's perspective, there are lots of choices as to what kind of accelerator they should apply to or how they should"}, {"start": 291.2, "end": 321.04, "text": "Kick off things. What's the most important element of the sort of pitch that you make to them as to why they should choose you over some of the other possible avenues? I think when you're considering any investor, not just an accelerator, there's the good, which is how will this investor help me? And there's avoiding the bad, which is how will this investor not hurt me? And we try to be the best at our stage at those."}, {"start": 321.04, "end": 350.24, "text": "We don't have too much competition from other accelerators, but we have a lot of competition from later stage investors convincing founders they don't need any accelerator at all. So our pitch is usually not why you should do YC instead of accelerator X, but it's why you should do YC even if you can go off and raise a series A today. And that, I hope we have a good answer to."}, {"start": 350.24, "end": 374.29, "text": "Which is that we try to, A, we try to never harm companies. I think this one we're really good at. We try to always do what's in the best interest of the founders. We try to really never hold up anything the founders want. And the way our model works is that when companies work out, we invest at such an early stage, we make so much money and we do so well that we can afford to write off most of our other investments."}, {"start": 374.29, "end": 401.3, "text": "So we never fight with founders over trying to get our money back or anything like that. We never stop founders if they want to sell early. And I think this is actually really important to learn and internalize to be a good investor, which is don't fight in the mediocre outcomes. The long-term brand damage you do and the damage you do to your network and reputation is just not worth it. So we're really good about..."}, {"start": 401.3, "end": 431.47, "text": "Not caring about what happens in anything except the exceptionally good scenarios. You know, when we're fortunate enough to get into business with an Airbnb or Stripe or Reddit or Dropbox, this pays for a lot of mistakes. And so we try to be really friendly on those. And then in terms of helping the companies, we try to build the best network of startups in the world. And I think at this point, we're well on our way to doing that. If you're a YC company, there are more than 1,000 other YC companies that will do anything they can to help you."}, {"start": 431.47, "end": 459.9, "text": "A community norm and cultural value of when you're in YC, you will get any help you possibly need from alumni companies. And then when you yourself become an alumni, you have to pay that forward. And look, I think we give pretty good advice. I think we have good events. But the community that we've built has been by far the most powerful thing that we've done. And that is what founders want to join. And I think that is what explains our success."}, {"start": 459.9, "end": 492.35, "text": "This is fundamentally a network, and so we have a network effect in our business. And that has been difficult for other investors to replicate. I'm sure people will. Now that people have seen that the model works, I'm sure other people will do it. But it's sort of like joining a guild of some sort. And I think you get a significant boost forward as a company in the network. And from a sort of industry standpoint, the numbers that often have been"}, {"start": 492.35, "end": 520.77, "text": "If you're funding 250 plus companies a year across two cohorts, that perhaps one in 100 or so will turn out to be one of these winners that sort of pays for all the mistakes. Is that what you're seeing kind of coming through in the data still as you scale up? Well, I think that's the number for a really big winner. I think it's a little bit better. I think if you look at companies that are sort of..."}, {"start": 520.77, "end": 551.09, "text": "Three years older, older, which is, I think, the fastest you can reasonably expect to get to a billion dollar valuation. It's about one in 70 or one in 80 of the companies that we fund have gone on to do that, which I think, I mean, when the press writes about this, they will often say, wow, you know, like, what a horrible track record. You know, even the best accelerator in the world, only one out of 80 goes on to be a multi-billion dollar company. And, you know, I wish the people that write that would come in and look at the stage that these companies are in when we fund them."}, {"start": 551.09, "end": 579.95, "text": "Because I think it's absolutely incredible that we're able to do that well. And I think it speaks to the enormous untapped potential of talented but unknown, unexperienced, usually young, unconnected people in the world. And if you're willing to put in the effort to sift through that, and it is a huge amount of effort, that there is just an enormous amount of value that has not yet been unlocked of new innovation to unleash on the world."}, {"start": 579.95, "end": 607.02, "text": "However, that number is only for the companies that go on to be worth a billion or more. Far more than that go on to be worth, say, 100 million or more. And so I think we end up probably making some amount of money on somewhere between a third and a half of the companies we fund. That's impressive. But even the one in 80, say, that's still three companies a year that from inception pretty much you have..."}, {"start": 607.25, "end": 635.6, "text": "Good ownership in the cap table. I mean, there isn't a venture firm that is as prolific or if any other venture firm that has three new billion dollar outcomes a year, they're typically buying in at later stages and their cost basis is much higher. Well, there's also no venture fund that takes as many zeros as we do. So that's the trade-off. But it's one we're happy to make. And how do you sort of..."}, {"start": 635.79, "end": 665.68, "text": "Remain on the side of the founder as the company scale up and become more self-sufficient in certain respects. We are just in service of the founders. Brian Chesky, I don't try to bother him too often. I don't try to call him with unsolicited advice too often anymore. I used to do that. But at this point, I figure out he knows what he's doing. But if he calls me and is like, hey, I need this favor, I will drop whatever I'm doing and go do it."}, {"start": 665.68, "end": 694.32, "text": "In fact, that was what happened right before we started this talk. And that's how we try to act with our companies. We never try to control them. We don't take board seats. We don't have any voting rights. We buy common stock. And as the company does better and better, we try to exert even less influence and even less control. You make all your money on the occasional $10 to $100 billion company, and those are almost always run by a strong-willed founder."}, {"start": 694.32, "end": 726.67, "text": "And so I think the best thing you can do is get out of the way. And does the role that you play change at all with the launch of the new continuity fund, which some people in the room may not know everything about, so it may be worth spending 30 seconds describing it. So last year we raised a $700 million later stage fund to try our hand at supporting our companies as they grow. And the genesis for this was there was a class of companies that we became super interested in."}, {"start": 726.67, "end": 756.32, "text": "A few years ago. And we were able to find them and fund them. And they were able to go on and have great success. But they needed huge amounts of capital. And most other investors wouldn't fund them because they weren't software companies. So these are what we call the hard tech companies. One of the first ones that we funded that was just acquired by GM was called Cruise. And they make self-driving cars. They were just acquired for over a billion dollars after two years. That was pretty quick. I wish they had not sold."}, {"start": 756.32, "end": 782.91, "text": "There were a lot of other companies we funded in the same vein. Helion makes fusion reactors. Rigetti Quantum Computing makes quantum computers, obviously. Ginkgo Bioworks makes synthetic organisms, and there are lots, lots more. And these are companies that I think are phenomenal investment opportunities and important to me personally. And they weren't able to raise subsequent funding. Meanwhile, we had a lot of other traditional software companies."}, {"start": 782.91, "end": 812.34, "text": "Where we still thought we were more bullish on them, we understood them better than everybody else. And so we looked at both of those two things together and thought, you know what, it would be really great for our ecosystem if we could be a friendly late-stage investor. And we could support companies no one else will, and we could change the norms for what it means to be a founder-friendly late-stage investor in the same way we did that 10 years ago for early-stage investing. Our founders in general have not liked their late-stage investors."}, {"start": 812.34, "end": 838.58, "text": "And I think if we can shake up that ecosystem, that will be a big positive for all the founders that we fund and that we don't fund. And we have really been able to do that in the first six months. So it's been a new experience. It's run by a fairly separate team inside of Y Combinator. But I think we'll be able to support a lot of really cool companies. This is a question that you may have had to..."}, {"start": 838.58, "end": 868.56, "text": "So think about before, but how do you deal with the sort of signaling risk and those kinds of issues now that you're potentially an actor in any late stage financing of a sort of YC successful company? So this was such a difficult issue that we almost didn't raise the fund. We would rather not make money. We would rather not support our companies that were already doing well than hurt companies that were struggling."}, {"start": 868.56, "end": 897.57, "text": "And so we spent six months debating what a late-stage fund might look like that did not violate the relationship we wanted to have with our founders. And we came up with two rules, both of which hurt our returns a little bit but preserved the community and in the long term will, I hope, ensure that we generate great returns. One is that below a $300 million valuation, we do our pro rata in every company across the board. It's automatic. It doesn't matter how good or bad we think you are, we just do it."}, {"start": 897.57, "end": 925.1, "text": "So there's no signal to look at. No investor can say, well, YC likes this company or doesn't because we do it every time. And we just accept investing in the companies that we think are bad as a cost of doing business and maintaining our relationships and community. The second is that we set the threshold where we can make independent decisions above the number where signal still really plays. So we looked a lot at"}, {"start": 925.1, "end": 953.89, "text": "Where our companies tend to raise money based off of really hard numbers, sort of growth capital instead of just a pitch. And we found that in those, the signal didn't matter as much because people were really doing something much more mathematical. And we found that was around $250 million valuation. And so to leave ourselves a buffer, we said that $300 million was the valuation above which we would make decisions to lead or not lead rounds."}, {"start": 953.89, "end": 984.61, "text": "And at that stage, we're not, I don't think, viewed as a big provider of signal because most investors just don't think they know what we're doing when it comes to growth investing. So if anything, we're like a little bit of anti-signal. But I think time will prove otherwise. But in any case, it's beyond the stage where I think we cause problems for our founders. And from the point of view of the tension that you intimated, you might sometimes see with,"}, {"start": 984.61, "end": 1014.51, "text": "Series A investors. So as Mosaic, Mike and I, when we pitched our firm, firmly saw ourselves as the folks that would invest after someone had been through YC or being potentially funded by angels or seed funds. And there's lots of folks in the room that we work with like that. I can't imagine a scenario where I would say, don't go to YC unless we felt that,"}, {"start": 1014.9, "end": 1044.18, "text": "That would sort of block us from investing. And even then, if we felt it was the right thing to do for the company, I think we would encourage the entrepreneurs to do it. Do you see yourselves as not just disrupting the later stages of the venture ecosystem with the continuity fund, but also at the earlier stages with Series A funds? No."}, {"start": 1044.18, "end": 1074.1, "text": "We don't. We don't plan to ever be a Series A firm. I think one of the things that has worked well for YC is we know what we're good at, but we also know what we're not good at. And Series A investing is very different. What we're good at is, I mean, this is sort of a derogatory term for it, but what we're good at is spraying and praying, lots and lots of investments. What we're not as good at and what takes a huge amount of work."}, {"start": 1074.1, "end": 1103.26, "text": "is where you choose five to ten companies and work incredibly closely with those as a Series A board member. And that is critical. If it were not for really good Series A board members, I think YC's returns would be far, far lower. So I think we'd be doing our companies a huge disservice and we'd be bad at it to try to be a Series A investor. It would also cause, obviously, huge signaling problems. That's why we only do the late-stage investing."}, {"start": 1103.26, "end": 1134.06, "text": "In addition to our core program. When you get to the later stages, those investors tend not to be intimately involved anyway. And I don't think they're as helpful, so I'm happy to out-compete those guys. But the men and women that get involved as truly hands-on partners with these companies after YC spend far more time than we do with these companies and provide them just incredible services. And we don't ever."}, {"start": 1134.06, "end": 1166.16, "text": "I don't think we'd ever be good at that. We have no desire to go try to take on that business. If you sort of imagine a world three, five years from now, how many companies could YC be funding? And I'll give you more credit than saying you're spraying and praying because if out of a funnel of 20, 30,000, you're distilling it down to 250, that's already... 350. 350, you're already picking."}, {"start": 1166.35, "end": 1197.7, "text": "One in 100. But how big can this become? My rough guess is that we should be 50% to 100% bigger than we are today. I think that's kind of how behind the curve we are of all the good companies in the world. And then the startup ecosystem worldwide maybe grows at 20% to 30% a year. So, you know, quite a bit bigger. Two, three X would not be crazy. What would be great?"}, {"start": 1197.7, "end": 1227.95, "text": "And we've tried a first version of this with the fellowship. We have an evolution of that we're going to try out later this year. What we'd love to do is figure out a way to keep the core program at roughly the same size, but have a giant funnel above it where we can, in a much lighter touch way, reach 10,000 companies a year, something like that. And would they benefit from the network in the same degree as all the other participants that, absent the three months?"}, {"start": 1227.95, "end": 1258.29, "text": "Well, if we couldn't find some way for them to benefit from the network, it wouldn't be worth it. They wouldn't get enough value. It would have to be different. And we certainly haven't invented it yet. I think people think that we do a lot more really advanced planning and we have a lot more answers than we actually have. But we know what we don't know. And we're not afraid to admit it. And we have no idea how this is going to work. But it's important to do. And we'll figure it out as best we can along the way."}, {"start": 1258.29, "end": 1286.91, "text": "As you think about some of the areas that you'd like to sponsor and take these bets on, what I'd like to do is now sort of move the conversation into discussing some of those and bring in Mike Chalfan, my partner, who's been spending a lot of time in our firm on AI, which I know is one of the areas that you're very excited about. Thank you. Thanks, Simon. So maybe if we could just start by just talking about the origins of..."}, {"start": 1286.91, "end": 1316.21, "text": "Your interest in AI before we sort of dive into some of the gnarly issues. Well, maybe even before that, I'll say sort of how we think about new areas. We're not really a trends-based investor. But startup to us does not mean a software company. Startup means a company that is designed to grow very quickly. It may fail, but that's at least the goal. And so what we think about is we want to fund any company that has some plausible shot."}, {"start": 1316.21, "end": 1343.66, "text": "At being a 10 plus billion dollar company. And that is such a strict rule that we don't allow ourselves any other rules or we'd have nothing to fund. And so we have to be willing to look at companies in any area. And if we're not an expert in that area, but we think it might be super valuable, then either one of the partners has to go become an expert or we have to hire a new partner. And I think this is."}, {"start": 1343.73, "end": 1372.32, "text": "Not the only model, but it's a model that's worked really well for us. We don't constrain ourselves to the thing that is working today or that worked last year, because we don't believe that's necessarily the thing that's going to work next year. And we don't let ourselves have the excuse of saying, well, we don't know about that, so we're not going to invest. If we think it's important, we figure out how to get reasonably good at it. And through this process, we have decided to really pay attention to some new areas."}, {"start": 1372.32, "end": 1401.01, "text": "Biotech, Energy, and AI are three that I have spent a lot of time focusing on. Hardware is something that other partners in the firm have done a lot with. EdTech has been another area of expansion. More and more companies selling to government. But this is kind of how we think about the world. Our first choice is always to support something by funding a startup. But if there is technology that we think is really important, or if there is a societal issue that we think is really important, and it does not fit in a startup,"}, {"start": 1401.01, "end": 1430.94, "text": "We will find some other way to support it. We fund nonprofits. We have something called YC Research, which is a nonprofit research lab. And if there is something that we think is important that cannot or should not be done inside of a for-profit company, then we do it this other way. And so on the should not side, we get something like AI. On the cannot side, we get something like our basic income research. But AI in particular is something that I, probably like many people in this room,"}, {"start": 1430.94, "end": 1460.24, "text": "I've been interested in since I was a little kid. I think many people that learned to program a computer for the first time realize that someday this is going to be really smart and maybe be smarter than a human. And so I studied AI in college. My advisor was Andrew Ng, who now runs AI at Baidu. And at that time, it was sort of depressing because we all had fervent conviction that AI was going to be a big deal."}, {"start": 1460.24, "end": 1489.7, "text": "And nothing was working. But I stayed close to the field. And in the last four years, or three years, it's just been incredibly exciting to watch. And it has been absolutely transformative. We started a group called OpenAI. Elon Musk, and me, and the former CTO of Stripe, and a guy from Google Brain. And that group now has about 25 people."}, {"start": 1489.7, "end": 1516.77, "text": "It is a nonprofit. The goal is to build general super AI for the benefit of humanity and try to do that in a way where it is not a single for-profit company with a single AI that controls the world and hopefully doesn't go off the rails, but a widely distributed AI where each human has an agent that becomes part of them to some degree. And we sort of try to just distribute this power and make all the humans better."}, {"start": 1516.77, "end": 1548.66, "text": "And it's super exciting to work on. Very, very exciting. I mean, just there's sort of positive and negative aspects to it. And you started on the positive, so maybe we can explore that a little bit more first. The way you describe it is very personalized, and I guess people couldn't envisage an assistant or something like that. But if you could flesh out a little bit how you see"}, {"start": 1548.66, "end": 1579.26, "text": "Everyone on the planet having their own AI and what that would look like beyond the sort of how we perhaps think about assistance today. I think that will certainly be one area where everyone will have a superhuman personal assistant where you can speak to something in true natural language. It will know what you want and it will go figure out a way to do it. But I think things are about to get pretty crazy. We're going to, we are not that far"}, {"start": 1579.26, "end": 1607.57, "text": "From a world where computers get better than humans at most non-creative tasks. And the creative tasks, you know, painting, writing novels, I think even that will come someday. You know, last year we got to a place where I think we saw the first types of doctors become less good at computers with AI radiologists. We'll see that I think in a lot of other areas."}, {"start": 1607.57, "end": 1637.12, "text": "Certainly self-driving cars are now, I think, to a point where it is not a question of if but when they become far better than human drivers. And I think when is probably only a few years away. Foxconn now has robots that are so good that for better or for worse, I think for better, they just eliminated a huge number of low-skill menial labor. And so you can see a world where we have basically"}, {"start": 1637.12, "end": 1666.34, "text": "Effectively, infinite resources and wealth. This was actually another mission statement considered for Y Combinator, which was to create infinite wealth. And then the question is, what are the new jobs for humans? I believe we'll figure this out. I think humans have limitless demand and they will figure out new types of social structure and new ways to get status and fulfillment and happiness and new jobs to do."}, {"start": 1666.34, "end": 1693.04, "text": "But they'll probably look really different. And if you were sitting a few hundred years ago on the other side of the Industrial Revolution, it was probably very hard to imagine that a lot of human jobs are going to get eliminated, and the thing that was going to replace them was going to be software engineering. And I think today, someday we're going to have software engineering get eliminated as a job, because you'll just be able to say to a computer, build me this exact thing, and it will do it."}, {"start": 1693.17, "end": 1722.69, "text": "And I think it's as hard to imagine the jobs that will come next as it was then to imagine software engineering. So I think this is an important question. I don't pretend to have the answer. We're doing a study, also in YC research, where we're going to try giving about 1,000 people a basic income for five years. And we'll say, we're going to fast forward to this world, let's imagine it, where we don't need everyone to work, and we have enough resources where we can give everyone money, and you can do what you want. And how will people spend their time?"}, {"start": 1722.69, "end": 1752.27, "text": "And how will they find happiness and fulfillment and social status and everything else that people care about? And I think it's going to be really different. I think it's important to start studying this now. Because even if the jobs don't really change for 10 or 20 or 30 years, human society does not adapt very quickly. And the sooner we can start seeing what happens and how people adapt, I think the better off we'll be. And we can hopefully learn how to"}, {"start": 1752.34, "end": 1784.24, "text": "Make a future where people are excited and engaged and happy. So it'll be interesting to see what goes on there. But I think my view of this is like, let's not be afraid of AI eliminating jobs because the jobs that it will eliminate first will be the worst jobs. But if it does that, it will do it because it can do it much less expensively than humans. And so the cost of the quality of a great life will come down."}, {"start": 1784.24, "end": 1813.94, "text": "And we'll be in a world where we can choose, we'll have to choose, but we can choose to just be much better than we are today about redistributing wealth. And we don't need everyone to work. And people can have a great life. And we'll be able to afford it because the cost of a great life will have come down so much in a world where we need the basic income in the first place. So let's not be afraid of eliminating these jobs that I think are not jobs that people are excited about. And let's let them find better, more productive, more fulfilling ways to spend their time. But I think that requires some foresight and some planning and some willingness to have a..."}, {"start": 1813.94, "end": 1841.65, "text": "A conversation that is not popular. That's a big, big, big topic. Big, big vision as well. And I guess one of the things that marks out YC from perhaps some other sources of capital and ideas and coaching. Very interesting. Just to sort of switch a little bit to the negative, because I guess as you described it, one of the motivations around OpenAI was..."}, {"start": 1841.65, "end": 1871.92, "text": "To avoid concentration of AI. So just to sort of tease out a few things there. One is it seems like at least you and one of your co-founders at OpenAI have slightly different concerns or maybe Elon Musk has additional concerns around sort of a bad actor obtaining AI and using it for negative purposes. And then there's also this sort of concept of"}, {"start": 1871.92, "end": 1898.67, "text": "I think we're both very afraid of the negative future and very excited for the positive future. And the reason that we're spending our time and money on this is that we believe in the next not too long of a time period."}, {"start": 1898.67, "end": 1926.22, "text": "We have the opportunity to push it towards the positive future. And we really want to try to do that. It definitely is scary to me to think about the single AI world. So one approach that some people have to AI safety is to say, hey, multiple AI agents will fight with each other and will have a lot of conflict. And that's very dangerous. And that the safe approach is to have a single AI."}, {"start": 1926.22, "end": 1956.02, "text": "That rules the world as a benevolent dictator and stops all other AI projects and makes this utopia for humans. And there is an argument that you can articulate that that is a good path. The thing I have never been able to be comfortable with on that approach is let's say we are successful in creating this single AI that is a billion times smarter than the next most intelligent agent in the universe, the smartest human."}, {"start": 1956.02, "end": 1985.39, "text": "And we think that we built in some controls and we think we built it to like humans. But we somehow put a bug in there or there was something that we didn't understand and it decides to go the other direction. And then you have this massively powerful agent that is a billion times more intelligent than the next most powerful agent. I don't see how we have any hope of controlling it or getting it to do what we want it. Like if it goes off, I think things get really unsafe."}, {"start": 1985.78, "end": 2012.58, "text": "Also, there will presumably be some group of people in the world that kind of control the AI, and if that group of people decides, I think the quote is absolute power corrupts absolutely, if they decide something the other seven whatever billion of us don't like, that seems bad. So I'm definitely afraid of a world that has a single concentrated AI. I like better a world with a lot of AIs, and in the same way that there are occasional bad humans."}, {"start": 2012.58, "end": 2043.25, "text": "But we as the collective are able to overpower them because we're all of roughly the same skills. I think if we have a lot of AIs and most of them want what we want and want, then they will be able to stop an occasional bad AI. So I think that's why we call OpenAI, OpenAI. We want this to be open technology made available to everyone. There's another problem that we're starting to think about now, which is if you have humans"}, {"start": 2043.25, "end": 2073.81, "text": "And AI. And they're separate. And they want the same thing, which is to sort of be the dominant species on the planet. You have inherent conflict. Whenever you have two groups that want the same thing and only one of them can have it, like being the number one or the dominant species or in control, you have some amount of conflict. And so the world where we have that conflict and either we're trying to enslave AI, the AI is trying to enslave us, those both seem bad."}, {"start": 2073.87, "end": 2103.44, "text": "So I think, Elon thinks too, some people on the team have different opinions, that some version of a merge, and this can range from the simplest chat interface where everyone has an AI companion that works together with the human, and the humans do what the humans are good at, the AI does what the AI is good at, and they kind of become this merged sense of self. Or Elon's version of this is that we just need full-on cybernetics and brain implants."}, {"start": 2103.44, "end": 2135.02, "text": "Other people think we should go the other way and upload ourselves into the computer. But whatever it is, some sort of merge, so it's not us versus the AI, but us and the AI as sort of one combined entity, seems to me like the most likely happy path. Okay, interesting. We can talk about crazy stuff later. No, no, yeah, this isn't the crazy stuff. So the, I mean, one of the things that that raises is sort of regulation and"}, {"start": 2135.02, "end": 2159.9, "text": "If you look at it already today, it's hard enough to approve a new asthma drug or something relatively benign, let alone how we combine people and AI. So just to sort of peel that, prise that apart a little bit, first of all, you started talking publicly about regulation and AI some time ago. In the last sort of couple of years, has your feeling for..."}, {"start": 2159.9, "end": 2189.58, "text": "How regulation could work. Yeah, I still think AI should be regulated in the future. I think this is all still pretty far away. But whether it's 10 or 20 or 100 years away, in human history, that is the blink of an eye. In the history of biological life, that is really the blink of an eye. And so a lot of the debate at dinner parties in Silicon Valley and events like this is, is AI 10 or 100 years away? And I don't care. What I think is interesting."}, {"start": 2189.58, "end": 2219.54, "text": "I think it is remarkable that we're not debating, is AI going to happen or not? But is it going to take 10 or 100 years? Neither of which is a long time frame. We are that close to the end of the line of humans being the most intelligent thing in the universe. And so for me, there's this question about, a lot of people talk about the regulation question of a timing question. Well, maybe we should regulate AI, but it's too far away. And I actually agree. It probably is too far away."}, {"start": 2219.57, "end": 2247.38, "text": "It's not that far away and no one knows how to regulate it. So I think we should start thinking about what to do there so that when we get closer to this, whether it takes 5, 10, 25 years, someone at least has an idea because today no one has any idea. And one of the other aspects about thinking through regulation whenever it's applicable is sort of who determines what the common good is."}, {"start": 2247.38, "end": 2277.73, "text": "Are there inadvertent consequences of a given definition of a common good, even if we can agree one? If we can unpack that a little bit, that might be helpful. This is why we like the model of extension of individual human wills. I don't like the model of a single dictator deciding what the human good is. That has gone off the rails most times. That's been tried in human history. But if we can instead extend individual humans and make them better, the humans can still decide collectively"}, {"start": 2277.73, "end": 2306.93, "text": "What the common good is. When you think about the sources of data that you can gather to power OpenAI and how that's different from the source of data available to today's would-be AI champions, could you just talk about that a little bit, please? I think the data advantage is significantly overstated."}, {"start": 2307.15, "end": 2339.47, "text": "The internet is a giant source of data. Anyone with a connection to the internet can get as big of a data set as they please. Many YC companies are also willing to contribute their data to OpenAI. But we don't have a problem there. Moreover, I don't think about the amount of data that a human needs to train. It's not actually that much. We can basically learn like two bits per second. And so I've been alive 31 years times, I guess it's like..."}, {"start": 2340.3, "end": 2370.99, "text": "300 million seconds per year times two bits per second. It's not a crazy amount of data that it takes to train a human to be pretty smart. So some of our AI researchers believe that when... Do you guys know the Paul Erd\u0151s book? This idea that God has this book of the most beautiful, shortest mathematical proof for every theorem and when you die you get to see it and occasionally you get a glimpse of the page."}, {"start": 2370.99, "end": 2401.98, "text": "Because you find the most beautiful the shortest proof for a given theorem. And so some of our researchers believe that when we get the AI algorithms from the book for building AI, when we discover those, first of all, you'll be able to print them on a t-shirt. And second of all, they won't require that much data to train. And so I think the advantage is overstated because I don't think it's going to be as hard."}, {"start": 2401.98, "end": 2429.9, "text": "I don't think it's going to require as much data as we think it does. And then given the importance of the mission, is there anything you could talk to us about regarding the governance of OpenAI that maybe is underappreciated in the public domain? We are just now trying to figure out what the governance model will be. Today it's just Ilan and I on the board, but we now want to figure out how to expand that."}, {"start": 2429.9, "end": 2459.87, "text": "You know, probably something like seven other people, not from Silicon Valley, not from the world of technology, perhaps. Maybe some of them will be. And we also want to figure out a plan for how, as we get closer and closer, this becomes something kind of democratically voted on by the world. We don't know what that structure is going to look like yet. We're beginning to talk about it. And then from technological safeguards, we've just hired our first two safety researchers."}, {"start": 2459.87, "end": 2489.57, "text": "Which I'm delighted. I think we got two of the best people in the world for that. And then I think probably a minor point, but there's been a lot of attention paid to the individuals who are at the heart of starting the project or foundation. But could you talk about the corporate sponsors and whether they have any sort of any input? No, they don't. We're pretty careful about that. So we've had a number of companies that want to donate money."}, {"start": 2489.57, "end": 2518.18, "text": "We really appreciate the support, and we try to be clear with them that you can donate money, but you can't have any access. And they all understand. They're all okay with that. Okay, fantastic. And then just sort of stepping back, we're talking about something very, very long-term, and again, as you said, imminence is not part of the OpenAI. It's not a consideration. Yeah, but again, this idea that we should only work on things that are going to happen in the next two years is crazy. Yes."}, {"start": 2518.18, "end": 2547.94, "text": "I think a lot of people don't work on stuff like this because it's not going to happen in two years, and that's a huge mistake. Yeah, so what I wanted to ask about is you've talked quite a fair bit about long-termism and your preference for long-term thinking. It would be great to understand partly how you think that changes how you evaluate problems and opportunity, but also when you look at and you speak to companies at YC,"}, {"start": 2547.94, "end": 2578.8, "text": "How can a startup factor long-term thinking more into its planning? Well, honestly, I think a lot of the problem comes from investors. I think most investors, although they talk a huge game about being very long-term oriented, are simply not at all. They need results to raise their second fund. The fund's life is maybe only 10 years, and they want to go buy their vacation home, and they don't want to wait that long. And so I think..."}, {"start": 2578.8, "end": 2608.5, "text": "It starts actually with the investors. It is up to the investors to convince the founders to take a super long-term view. Sometimes you don't have to. You get a Zuck or a Larry Page and they take a 30-year view anyway. But I think one of the things that YC has done really well is encourage our founders to take very long-term views. We invest our own money. We don't have LPs. We do on the continuity fund, but on the early stage side, we don't have LPs that we have to answer to. We don't have to show any short-term things. And we're happy to..."}, {"start": 2609.01, "end": 2638.53, "text": "We want to make a lot of money, but we're happy to wait a long time to do it. And we understand that these things take a long time to compound. So I think the way that this changes the most is investors deciding that they're willing to sign up for things that are going to take a long time and pushing founders to take a long time. I think Silicon Valley is pretty good at this, better than most other places in the world, but still not good enough. And certainly if you look at our most successful companies."}, {"start": 2638.53, "end": 2667.34, "text": "They're on path to take a very long time. And I think investors can push this through. I didn't plan this, but I'm happy I became an investor at a relatively young age. So I made my first investments when I was 24, 25. And I realized that if one advantage I had was that I could go invest in things that would take 50 years, and I would still benefit from that."}, {"start": 2667.34, "end": 2696.72, "text": "And I think that's been a mindset that has been good for me. And does that play into the filter that you take as YC in terms of the 350 companies that you admit into the program? Do you skew it more towards folks that have, whether it's AI or other more long-term ambitious? Honestly, for us, it's so much about the founder. If we see an incredible founder,"}, {"start": 2696.72, "end": 2726.48, "text": "You know, a Tracy of PlanGrid, a Brian of Airbnb, a Patrick of Stripe. They can be doing anything that they want. And no matter how bad of an idea we think it is, we will delightfully throw our money at them. And, you know, on the hope that they're smarter than we are and they see something we're missing. And this has worked out for us many times. And so, yeah, sure, we love it if a great founder is working on an area we care about. But if it's an area that we love and a bad founder, we still won't fund it."}, {"start": 2726.48, "end": 2756.72, "text": "So it's far more important to us to look at the person and say, can we see this founder growing in to be one of these iconic CEOs of these super impactful companies? And we'll talk a little bit about energy and synthetic life going forward, but just while we're on this subject, how do you screen for that? And speaking honestly, did you know when..."}, {"start": 2756.72, "end": 2783.39, "text": "Brian and Joe and Nate walked in, you know, these are the guys, five minutes with... In that particular case I did, there are plenty of other companies that have gone on to be successful where I didn't think I was not blown away in the first meeting. But in that particular case, me and everyone else that met them, I won't take special credit for this, you know, like they had been, we can tell the story later if we have more time, but it was this crazy, crazy situation. So those guys were..."}, {"start": 2783.39, "end": 2813.86, "text": "That was an easy one. And generally, what is the kind of question that you ask or the emotion or intuition that you feel that helps you filter that way? You know, you're looking for evidence that the founders will do extraordinary things in the future. And the best way to look at that is evidence of doing extraordinary things so far, either in the beginning of that company, if it's already been running, or in those founders' lives before the company, relative to whatever kind of"}, {"start": 2813.86, "end": 2841.17, "text": "So we're as impressed by a founder who is born in rural Pakistan with nothing and makes it to the US to start a company as we are with someone that was born in the US with a little bit more luck and is able to start their business by the time they're 18. But as much as we can adjusting for circumstances, we look for evidence that founders have done really extraordinary things. And I think that"}, {"start": 2841.17, "end": 2871.41, "text": "It shows a level of creativity and determination, which is very important to be a good founder. The general traits, none of these will surprise any of you. Determination, communication, intelligence, vision. This one is a little bit more difficult to define, but ideas we haven't heard before. Some evidence of the ability to have new ideas that everybody else doesn't have, that's really valuable too."}, {"start": 2872.37, "end": 2902.34, "text": "And so taking that into some of these other areas that are part of the research activity that you mentioned, can you spend a few minutes talking about both energy and synthetic life? Sure. When we attended the demo day earlier this year, there was already some companies that were embarking on those areas. I know it sounds like you've taken it personally. Yeah, so again, if we can do something as a for-profit company, we would rather do that than do it in the..."}, {"start": 2902.34, "end": 2931.47, "text": "NYC research. And so on the energy side, we have been able to fund the two areas we've concentrated on generation are solar and nuclear. Solar, those companies are less hard, difficult technology and more new business models. So one company is called Bright that sort of puts solar panels on homes in Mexico, which I think makes a ton of sense. On the nuclear generation side, we have companies like Helion in Seattle working on fusion."}, {"start": 2931.47, "end": 2961.09, "text": "And Oklo working on very small, very inexpensive fission reactors in Mountain View. We also funded some companies on energy storage, energy transmission as well. On synthetic biology, we have funded dozens at this point of companies doing work here. And that has been a really amazing field to watch in the same way that AI was sort of quiet for decades and then had amazing breakthroughs in a period of a few years."}, {"start": 2961.09, "end": 2990.86, "text": "We've seen the same thing in synthetic bio. In the last few years, you see this explosion of companies that are now able to create new organisms, like Genco Bioworks, which was one of the first synthetic bio companies that we funded, to editing existing organisms and improving them in different ways. So that, I think, will be a very important vector forward. I'm conscious that there's, I'm sure, lots of questions."}, {"start": 2990.86, "end": 3018.62, "text": "Thank you very much. I just wanted to ask for a bit more detail on the basic income."}, {"start": 3018.62, "end": 3047.62, "text": "pilot than how clean it is? And in particular, how much money are you giving to people to replicate the world that you think may arrive in time? So the first thing that we're doing with the basic income study is a pilot. This is a much smaller scale. It'll last close to a year. And it is to test our methods, how we're going to study, how much money to give, how to distribute it, how to track results. And so that is really kind of the design phase of the study. That will be with 100 people."}, {"start": 3047.62, "end": 3076.94, "text": "And we'll try different numbers. And our hope is that we only need to run one pilot. But if we have significant learnings or we learn that we're really wrong, we'll have to redesign it and run another pilot before we're ready to start the full study. But if it goes well, a year from now we could be ready to start the full scale study. And the numbers that most people throw around are something on the order of $18,000 per year."}, {"start": 3077.2, "end": 3106.7, "text": "I think we would do 500 people that we gave it to and 500 people that were the control group. And again, lots of questions we're trying to figure out now about how to distribute the money, how to track progress. But the plan would be to roll that out probably still in Oakland, but maybe in two different cities at the same time next year. Christian. Thanks for hosting and thanks for being in London. So on the AI and the regulation piece, so something that, can I?"}, {"start": 3106.7, "end": 3135.09, "text": "It concerns me in a way as to who gets to decide how we regulate it. If it's the nation state, each nation state's going to have their own priorities. So effectively, China and the US and India will regulate differently. If it's the UN, that'll take a long, long time. And then I kind of start running short on actually who would actually do it. So in your mind, when you think about regulation for AI and not going down the benevolent dictator route, what entity actually creates that regulation? And just to add to that, should it be a commercial entity?"}, {"start": 3135.09, "end": 3161.71, "text": "We honestly don't know. This is actually something we're doing research on, is who should regulate it and how. What I think would be really cool, and just a cool sort of statement, is if you could do something democratic on the internet. Because the internet is powerful at this point. The internet can make consensus decisions. The internet can force them to some degree. And so to see the community basically"}, {"start": 3161.71, "end": 3189.65, "text": "Volunteer to be self-regulated and then have the internet as a whole. And I don't propose to know the specifics here. But to get input from people around the world that are part of this shared thing that is now, I think, more powerful than any single nation state, I think that would be kind of a great moment in the history of technology. And so we're doing some research because it's a really hard problem to see if we can come up with a solution to throw out to the world for consideration."}, {"start": 3193.42, "end": 3223.42, "text": "Hi. So it kind of related to that. I was sort of thinking, you mentioned that one of the benefits you had of starting your investment career at the age of 24, 25 was that you could take a 50-year view on things. And that sort of let me sort of thinking about, you know, what we're sort of setting in motion here, having implications that go well beyond 50 years. And if you thought about the people who built the great cathedrals across Europe, when they embarked upon the project, the probability of it finishing in a lifetime was close to nil."}, {"start": 3223.42, "end": 3250.64, "text": "You know, you see in Sagrada Familia in Barcelona is 250 years into a 400-year project. But the foundation behind it was religion. And I sort of question, you know, similar to this, that, you know, when I look across to the U.S. and see the prospect of who might be in the White House, that the social fabric that is sitting behind what we're going to set in motion is quite disturbing. I think this is a really..."}, {"start": 3250.64, "end": 3277.58, "text": "Who is actually doing that research and where do you get those skills from that can kind of project what sort of world it is that we're setting in motion? Are they from the technology sphere or are they from other spheres? I think it's important to have a wide variety of backgrounds."}, {"start": 3277.58, "end": 3307.41, "text": "I think you want an organization of any sort, but certainly one like this where you're putting into place an organization that is going to hopefully outlive all of you. Religion is one way to get it to survive for a long period of time, but a deep-seated belief about something that's really important for the future is another. I think to do that, it's very important to have"}, {"start": 3307.41, "end": 3337.06, "text": "Diversity of backgrounds and life experiences on the team. But to have a deeply shared vision for the future. And so you want people that are engineers and philosophers and political thinkers and a lot of other things as well. And you want them to come together and bring all of these different viewpoints about the world and skill sets. But to share this deeply held high conviction belief about this."}, {"start": 3337.06, "end": 3374.86, "text": "It's a version of the future that's absolutely critical. And if this group doesn't do it, it's not going to get done right. And I think that's how you create these things that outlast any individual, even Trump. Ellie, we have someone, a lady at the front here. It's been amazing just hearing about the scope of the opportunities that you see to be exploited, and maybe following on from the last question as well. I wonder how you think about bad things to be combated. So we're talking about religious extremism, but I mean, nationalism."}, {"start": 3374.86, "end": 3404.77, "text": "Environmental challenges, plastic in the oceans. Do you see them as part of the opportunities as well or do you think about them differently? No, absolutely. I mean, I think many of these things are both incredible risks and incredible opportunities. And again, where we can address them with a for-profit company, we'd like to do that. And where we need to do it some other way, we'll do that as well. But we're trying to build, as are all of you, the future we want to live in. And there are a lot of bad things to combat and a lot of great things to build. Sometimes they're the same."}, {"start": 3404.77, "end": 3434.53, "text": "So I think the energy stuff that we talked about earlier, if you could pick one thing to help the poorest half of the world a lot and all the rest of us still a large amount, I think cheap, clean, safe energy would be the number one thing. Energy obviously wrecks havoc on the environment and health, but also it's a huge economic cost and it's responsible for a ton of war and a lot of other things that you could fix if you could just fix the energy problem."}, {"start": 3434.53, "end": 3465.02, "text": "Things that I'm always struck by when I read about different problems in the world is how many of them reduce to energy. And so I think if that's one problem, or if you could get that to work, you would solve a lot of problems that seem unrelated. And that would be really powerful. I think some of the crazy stuff happening right now on the internet can also be addressed with better services, although I don't think that'll be enough. I think we also just need government action and regulation and policy."}, {"start": 3465.02, "end": 3492.56, "text": "I think one of the worst things that has happened on the internet has been the complete disregard for basic human decency that you see. There are things that people are willing to say from one computer screen to another that they would never ever say face-to-face or most people wouldn't. It seems like we have some deep-seated evolutionary pressure where it is socially unacceptable to be"}, {"start": 3492.56, "end": 3521.46, "text": "That rude or that crazy face to face. And there's probably some like, if you did that, you got kicked out of the tribe and didn't get to survive very long. So we seem to have that program into us not to do and it's gone missing on the internet. And it makes the internet honestly not a super fun place to be a lot of the time. And there's a question of can we build new technology to bring that back? So maybe, this is pure speculation, maybe in VR, if instead of typing back and forth,"}, {"start": 3521.46, "end": 3551.42, "text": "We feel like we're standing in front of another human and we feel like we have the latency is low enough and the fidelity is high enough that it's not this uncanny valley but it feels like we're standing face to face with another human. Maybe we can like trick our brains and get that evolutionary pressure to be decent people back and maybe we'll feel sort of like more connected and we won't say well we should blow up that group of people to the same degree. So I do think there are technological solutions to a lot of this but I don't think it's the only."}, {"start": 3551.42, "end": 3581.22, "text": "It's the only answer. So I guess a lot of us founders in the room probably grew up, even in our last startup, sort of reading Paul's essays and sort of almost ingraining into our DNA the mindset around done is better than perfect and get the product out. A lot of us now are probably running companies that kind of fit into some spaces you talk about where it's this new generation where it's a lot of harder problems to run. Done is better than perfect might not be a mantra you can really run away with because there's regulations, there's government, there's..."}, {"start": 3581.22, "end": 3610.96, "text": "There's ethics, there's consumer privacy. So are you seeing this sort of change in mindset for entrepreneurs coming into the YC program? Certainly a little bit, although it is important. There's a big mistake, a big danger zone here, where you can get into this mindset and say, well, I am trying to cure cancer, and thus the normal rules of a startup don't apply to me. I don't have to be frugal. I don't have to move quickly. I don't have to build a"}, {"start": 3610.96, "end": 3640.67, "text": "Product that works. I don't have to take safety seriously. There are a lot of these things where you can get really wrong. And I think it is super important to remember that no matter what you're doing, you still have to win in the market. And absolutely, if the stakes are higher, you have to be a whole lot more careful and a whole lot more thoughtful and move a whole lot more slowly. But that doesn't mean not to move forward at all. And I think most startups still make the mistake."}, {"start": 3640.67, "end": 3671.42, "text": "Of giving them an excuse, giving themselves an excuse to not have to execute relentlessly. And it's very important not to fall in that trap. That said, obviously you'll have to adjust it down quite a bit if you're, you know, working with human health. And just on that topic, if I can just add a question that follows on. Could you just talk to us a little bit about the sort of thought behind the investment in Asana, which I guess is part of that thinking about how best to operate? Yeah, I have been,"}, {"start": 3671.42, "end": 3702.46, "text": "Really interested in how to coordinate groups of people for a long time. I think going back to the thing we were talking about, about groups of 20 humans and not being too rude, I think we're very good at organizing small groups, but we are not very good generally as humans at organizing groups of lots and lots of people to a common goal. And corporations have been one answer to that. Religions have been one answer to that."}, {"start": 3702.46, "end": 3728.35, "text": "It's sort of an amazing thing that it's possible at all. And as YC gets bigger and bigger, and we funded thousands and thousands of companies, some version in the future where we funded thousands and thousands or tens of thousands of companies, and there's millions of employees at them, and we want to be able to coordinate that whole group to some degree. Thinking about the technological tools that will make that possible has been really interesting to us. And I think Asana has"}, {"start": 3728.35, "end": 3762.86, "text": "A very interesting way of doing that that kind of happens to just fit my own personal mental model and workflow. Okay. Andre. Just building on that last point, you mentioned coordinating large groups of people and gave religions and corporations as the example. Many of the problems that you've been talking about today relate to global problems as opposed to nation-bound problems. How does technology and some of the startups you're looking at allow us to coordinate on a global level?"}, {"start": 3762.86, "end": 3791.89, "text": "And what does that mean for some of the, you know, arguably archaic notions of democracy, localized capitalism, localized fiscal rules as the sort of norm? Well, I mean, good and bad, I think that's already happened. I think people choose the communities they want to be part of on the Internet, and the Internet does not, like sort of accepting China, the Internet does not have nation-state boundaries. And so people..."}, {"start": 3791.89, "end": 3821.71, "text": "Choose the communities that they want to be part of. You see this on Reddit, where Reddit is not only one community, but it's made up of tens of thousands of very distinct communities that people, no matter where they are in the world, can join and leave and do whatever they want. I think we'll see a lot more of this. And I think that that trend and one other, which is the ability of local"}, {"start": 3823.54, "end": 3853.36, "text": "The ability for local efforts to have non-local effects. If China burns a gigantic amount of coal, it affects the whole world at some point. I think those two things together, if I had to guess, would lead to a weakening of the traditional nation-state boundaries and more towards some interconnected world. But I'm not sure about that because there are other forces pushing very hard in the other direction. And I could see an alternate world where it goes hard in the other way. And to give the US example,"}, {"start": 3853.36, "end": 3884.06, "text": "You know, right now we're mostly one country, but we have 50 states. And I could see a world where the 50 states become really powerful and the U.S. becomes less powerful and people kind of move and get increasingly polarized. So I think it's a really important area to think about. Certainly the work we're doing at YC, I think, will lead us towards more, you know, we're all kind of one world and one species. And that's what I'd like to see, but I could see it go the other way."}, {"start": 3884.06, "end": 3925.68, "text": "Should we do maybe one or two more questions? I think so. I think great. John. It's not that different. He's only 30 now. A blessing and a curse that I've had is that I cannot make myself focus on something I don't find interesting. And I am pretty good about not doing things I don't want to do."}, {"start": 3925.9, "end": 3957.41, "text": "I don't have one of these super rigid, you know, today I do, like, Mondays I do this, Tuesdays I do this, Wednesdays I do this. It's rough. There's, like, a very rough thing about, like, you know, kind of Mondays I go into OpenAI, Tuesdays we do the main YC batch and other things on other days. But beyond this very rough thing, I try to just work on whatever I think is important and interesting. And I try to have, like, very unstructured time. I try to not have a lot of meetings. They always fill up."}, {"start": 3957.41, "end": 4010.51, "text": "Last minute, the day before. But I try to leave a lot of time open to go work on whatever I think is sort of the most important and interesting stuff. A lot of that is advising startups. That's probably the biggest single time block. And the second biggest is probably just running YC. At this point, we're big enough that there's a lot of just sort of operational work to do there. One last question. Yeah. Yeah."}, {"start": 4012.18, "end": 4043.94, "text": "So in addition to OpenAI, another group that YC Research is sponsoring, they call themselves HARC, which is, I think, the Human Augmentation Research Center. And OpenAI is obviously the AI side of the equation. HARC is the IA, intelligence amplification side, where we just figure out how to use software or drugs or whatever to make humans better humans and hopefully kind of meet in the middle at some point. So with that, from..."}, {"start": 4043.94, "end": 4055.79, "text": "Mosaic and Mike and myself, thank you very much for spending time with us and for those of you in the audience. Very interesting."}]}