2019 Advice to Entrepreneurs, Sam Altman & Jack Altman

hello hello i'm glad you had me do this um i think you had the hope that as your brother i would be doing softball questions i did um i have a secret agenda of trapping you into saying something silly in the grand tradition of former yc presidents i will try to say something completely innocuous and have to press hammer me for it i'm excited that's my goal and then we'll try to have time for questions at the end but i do have some questions so we will uh we'll get into it okay so yesterday was an exciting press day you announced transitioning yc to jeff allston you've done it for five years it's a really different place than when you joined so i would love to hear what are you most proud of and what do you most regret um i think the thing that i'm most proud of is we really built an empire uh in a way that i think has not happened many times in investing um we treated it like trying to build a company uh not make investments and we sort of had a a strategy and we ran it like we've got advised companies to build build a competitive mode and an advantage in a brand and a network and i think we did that really well and i'm proud of the uh the team of course and the founders that we fund most of all but i think we came together with the strategy and built something that has created a ton of value and will do so for a long time in the future um there's always like lots of regrets but i think the biggest one uh to a much lesser extent in the same way that the president of the us is sort of the effective leader of the party or the real leader of the party i think the president of yc is sort of the unofficial leader of the startup movement and i think there's a lot of things that have gone wrong in the startup movement and i feel some degree of personal responsibility for that and the one that bothers me most of all is i think the motive as starting a startup has gone from this sort of outsider thing to basically the equivalent of going into banking or consulting when you graduate an elite college i think the motivation of founders has gotten off and i feel some personal responsibility that we weren't able to um do everything we could to grow the startup ecosystem uh and also really imbue the culture that uh and and sort of the right motivations for starting the startup mm-hmm yeah and we'll get into this more um but um i guess also sort of so that's five years back zooming back 15 years when you started your company silicon valley was a very different place how do you feel about the state of startups is it what you expected it would be in 2019 is what you hoped for no um i think the saddest thing or the thing that i feel most uh that i wish were most different is i think people are starting uh insufficiently ambitious startups and and are operating on an insufficiently long time frame i think the biggest opportunity in startups is people who are willing to take a very long time frame work on very hard intensive uh capital intensive resource intensive talent and has it whatever problems and i think i would have expected as sort of startups became canonized and sort of professionalized that you would see people willing to do harder things and take more risk and sort of have more manhattan project like startups or apollo project-like startups and i think we've actually seen less and less and that i think is not what the world needs right now it's also sort of not what i find personally interesting one of the things i always tried to do at yc was push towards more of the hard tech startups and i think the community has gone away from that that said uh to i'm much more positive than negative as a whole um i think it's it's remarkable it's incredible um what has happened about how easy it has become to start a startup how often people start things that look like toys and go on to become these like monsters like airbnb um and if you think about how much the world has changed in the last 10 or 15 or 20 years and how much of that change has been driven by startups and how dependent we all are on services that didn't exist 20 years ago companies that didn't really exist 20 years ago uh i think that's incredible so i'm i'm met much happier than if we didn't have the whole startup movement um but i would like to see it become more ambitious again do you think that this has been sort of a necessary part of the tech innovation and deployment sort of phase like carlota perez talks about or do you think that there's something that yc or other sort of institutions could be doing differently to sort of change the course uh i think it's really up to entrepreneurs to change the course i think yc and other people can continue to talk about what could be different um but it really comes down to like what the best entrepreneurs in the world decide to work on and we can try to convince them kv and others can try to convince them we all should um but i think one thing that's happened that is a little bit what's the word i i think a lot of the discovery phase of how to start a startup has been figured out um it doesn't feel like a frontier anymore at this point it feels like there is a playbook and if you want to be a founder um there's this set of things you can do it's very well explained online by many different people and you can follow that and if you do it a few times and get lucky at some point it will probably work and that is a tempting thing to do in the way that it used to be tempting to be a banker or consultant and so that siren song i think people away from sort of more uncertain you know the part of the map where it says warning there's dragons here um but those are that's like the fun area to work at for me at least and i think for some other people as well and i think what will happen is startups in the sort of the traditional sense or the sense now that has become commonplace will become a more normal career path uh and then there will be some people doing things like opening ir plenty of other companies that sound uh non-standard that makes that scarier is in some ways what you're saying is doing a sort of normal startup feels safe and doing some of the bigger things is scary you you you've had this sort of willingness to be wrong a lot um i don't know that everyone necessarily has that and i i wonder if that's what it takes well i think that's again like i think most people want to do something that is like a relatively uh well understood path that other people are on as well and i think uh it is kind of cool that like starting a startup has been so figured out that it is no longer considered a super risky career option but it's something that like tens of thousands of people apply to yc to do every year um and you know then so that'll be what a lot of people do and that's fine that's good and then there will be some people who are like i'm gonna go do this other thing and start this very other kind of company or you know really take a long-term horizon on on research but i think that's all okay i think i think it is important to me personally and to other people i've met uh to be willing to not necessarily be wrong but have other people think you're wrong for years and then turn out to be right um and that was that was what it felt like when i started a company i loved that feeling uh and i think that is still the way that most of the value in the world gets created i think when everyone thinks i'm doing something that makes obvious sense or is definitely gonna work um or is what i should be doing i always get quite suspicious how can entrepreneurs build the sort of resolve yeah to not care for or to be unaffected for years if people think they're wrong yeah so i think this is actually uh the normal advice here is this is something you're born with or not i don't think that's true um i think people have to build up their own confidence slowly over time by uh having a deeply held conviction having people tell them they thought it was wrong it didn't make sense and then being proven right and then it happens again and again and again and each time it happens you trust yourself a little more uh you are willing to take a little bit more uh risk you're willing to like take a little bit of a longer term bet have people think you're wrong for longer and eventually uh you you build up the confidence that uh you'll be you will be wrong sometimes but when you're right and everyone else thinks you're wrong um those are the opportunities where the most value gets created and it's worth it to be wrong some significant percentage of the time if you can be like breathtakingly right and against the crowd occasionally and another thing i think there that is helpful is to find a crew like one of the things that is hardest about being a startup ceo is not having peers uh and one of the things that yc got right accidentally is we gave ceos peers the other people in the batch and i have been lucky to have sort of a uh crew of um conspirators to work with that even when sort of like most of the world thought i or they were doing something wrong we could sort of like rely on each other and i think it's it is useful to find um people who have that sort of shared spirit i have one more question this topic related to this in toic and it's related to something that um vinod has talked about that i think is very powerful which is entrepreneurs approaching big projects like climbing everest and that there's this difference between a zero million dollar company and a zero billion dollar company and that you don't need to know the exact course but more or less you should have this big plan to me in a certain way that seems not at odds but different from a type of yc advice that i hear a lot which is solve a very niche narrow problem and then sort of trust the process that you'll expand from there how do you think about that yeah i don't think they're really at odds i think in both cases you're saying like you want to go to the top of the mountain you can see the top of the mountain you can see the first base camp and you can't see the things in between you can't see the path in between and what we always tell founders is uh like have the ambition have some sort of vision for how the company can be huge know what you're going to do first um and then be flexible in tactics and where you expand but like keep going up don't go to the side or don't sort of stop when you get to base camp and that was always how i ran yc which is that uh like we had a strategy that we talked about all the time and the strategy was that over time um capital and advice would commoditize and the only way to build a durable advantage in investing was going to be to build a network effect which is what we tell any company like network effects are the best competitive advantage when you can have them and so now when people ask like oh like people almost they're like this thing that happened at yc is so cool like was that an accident it was like no of course we planned it and like what does it mean that you as an investor like are so surprised that we planned it i think investors run their own operations most of the time extremely differently than the advice they give to their founders they view decisions as sort of these one-off like oh we're going to like work on this deal but there's not really a cohesive strategy most of the time there is not a long-term plan to build enduring value in fact most of the things that investors tell founders to do build a team like you know have a succession plan like it doesn't it usually doesn't happen kv an exception um but it's surprisingly rare and that point has really been driven home for me given the apparent and i think genuine surprise uh of other investors reacting to the fact that like we had a strategy all along and that we were trying to build something of enduring value and not just something that sort of is good at stock picking and i think there's a real lesson there if anyone wants to start a new investing firm which is you can play a game on the people on plane um and i want to come back to this video of time but i want to make sure that i ask a couple questions about open ai too um because you have a new job now i do well i've had it for a while but now i haven't yeah i can focus on it more so open ai you've got sort of like a mission statement yeah like what what is the mission of open ai we want to build safe artificial general intelligence and figure out how to maximally spread the benefits among humanity i think that building true ai something that is genuinely smarter and more capable than humans in every way will be the most important thing humanity ever does and i think if we don't have wide input into how we want this transition to go and a wide sharing of benefits and make sure that it's sort of like human positive uh that's really bad and the upside of doing all that is really good like we can solve any problem we can imagine um just curious of like going into this i know it's been going for a while but going into the project i'm curious what is your sort of level of conviction right now that ai will sort of wake up um and and when would when do you feel that that will happen you know the the the wake up question is harder because consciousness is so poorly defined but my level of conviction that ai will eventually surpass human intelligence unless we destroy ourselves first is 100 it will happen um the timing question is much harder and i don't think it matters that much these huge debates in the field about are we 10 years away or 50 years away or mostly in that range i would say and and like again for this most important moment in human history to be that close whether it's 10 or 50 years away i don't think it's super relevant and i think speaks to a problem of time horizons in the tech industry because that is the blink of an eye and everything is going to be different on the other side of that and we should all i think at least decided for myself like once you believe that it's hard for me to work on anything else uh and and the fact that there's so much debate about the time frames when everyone agrees that that they're that relatively short um is surprising and i think indicative of this other problem what would you say is sort of different in your motivations now in starting something like open ai versus you know in motivations you had during looped or yc if those were similar like just to try to find what the thing is that would kind of inspire more of these types of things um so looped was not that was like that kind of just happened i didn't have like a lot of i didn't give a lot of force i was just like hacking on this project i really liked that it became a company um yc i definitely was like okay how can we build this sort of engine of innovation to do lots of things uh and you know it'd be easy for me to tell the story which is like i always knew ai was the one that was going to work the best and there's some truth to that because i don't remember this when i was a kid i was like really into sci-fi and ai yeah you do yeah and i always thought that would happen someday but you know when we started working at openai i didn't know it was going to work and there were like lots of other projects that i also had high conviction in my work um and then but when it became clear to me that i thought we actually had a plan to get from here to agi uh then it was clear that that was what i wanted to work on what are the um what are the other areas where there are projects like open ai that could be happening now that aren't so i think there should be like much more investment into nuclear energy i think there should be much more investment into carbon capture i think we will figure out a number of real-time diagnostics of humans that look at a molecular level all the time and sort of answer daily weekly monthly just diagnostics to everything going on in your body i think that there will be synthetic biology should be a revolution in terms of how a lot of things get made at massive scale there's a very long list of this stuff and i agree with you not not much of it is happening or not enough of it is happening and why doesn't it like why don't these projects happen so i had this observation i wanted to go over yc like one of the things that i tried to do was really push towards more hard tech companies and in the year that i started the year before there had been like many thousands of photo sharing startups that had gotten started uh because everyone wanted to be the next instagram it was sort of as that was ryze was starting and there were four nuclear fusion startups in the world and that seemed to me like wrong or or at least like a misallocation of risk reward and i think one is that uh a lot of people have this thing which is i need to first like start a company so that i can make 100 million dollars in three years and then i'm gonna do what i want and actually i think a lot of those people if they would relax that very short time constraint would go on to make a successful software company and then could do something else but this thing of like i gotta do this thing very quickly to solve this problem and then i'm gonna go start a company that seems to usually not work and so i think people have this feeling which is like i gotta just i can't take that much risk it's like it's the special time i can start a software company make all this money and then i can like be free to do whatever i want and i'll have already succeeded and i won't be stressed but then they're not willing to make a long-term commitment to that either and even easy companies no company is easy but even things that people perceive as easy companies it still takes like 10 years and so you've still got to like really commit to do it well um and i actually think there's never been an easier time to start a hard company as your first company i think in the current environment it's easier to start a hard company than it is to start an easy company easy companies are boring it is very hard to get the 10th or the 20th employee to join it's very hard to like differentiate there's hundreds of shots on goal for every good idea um and people don't care that much and if you are starting like you know a nuclear fusion company even if the odds are against you people really want to help you have this tailwind people want to join you in the quest and that is a super valuable thing especially in a time when there's so much clutter and so much noise um and so i i think that if you're going to if you're gonna start any company it's like a 10-year commitment and if it works and it's not it's not easy to start the easy company something i remember i can't remember who from the paypal crew said was that it there used to be fewer of these sort of good startups and so it was they would get more concentrated teams do you think that's a real phenomenon huge problem huge problem um [Music] it is so hard right now to hire good people to be the early employees at a startup because they all just want to start their own so if the mission doesn't matter if it's not sort of this big quest that people want to join then they're like well why would i be an employee and sort of not get the exposure the learning the status the glory 10 times 20 times the equity which i think is a huge mistake um why not just like go start my own and again i think yc has some responsibility here yeah um all right i'm gonna ask one last question then we'll save a couple minutes for q a um my last question is if you could you know wave a one change anything about the tech startup entrepreneurship ecosystem what would you change we talked about this a few times but if i could if i could do one thing it would be to get everyone in the ecosystem to take a much longer time horizon i think this world where people start a company and plan to run it for four or five years join a company and only plan to stay for one or two that's not how important gets done in the world and having people willing to make a very long-term commitment to a hard problem and from the outset rather than say i'm going to do this thing as like a stepping stone to go start this other company or my real the company i'm really passionate about or i'm going to work here for a year and then start my own company or work here and then there and then there i think that is so value destructive for a certain class of problems that if i could pick one thing it would be to get people to make a long-term commitment to high-risk high-reward problems and that's easier said than done and i'm in a place where it's easier for me to see that say that because i've already been as successful as i need to be but that's still the one thing i would change awesome thank you um yeah questions um i don't know if there's a mic set up yeah i guess there's might going around in the back um sam you mentioned that um you joined open air full-time and you made it your sole focus when you realized there was a plan to get from here to agi can you just share your thinking around that plan yeah so i think the most important thing that i didn't realize initially is that the amount of compute available per to sort of train one neural network is growing about 10x per year and we have visibility to seeing that continue for the next six or seven years and most of the progress so far in machine learning has been this scaling it up like if you if you make a chart of the landmark result of each year it's basically uses 10x more compute than the previous year and you know with another six seven eight orders of magnitude we will be able to approach uh the human brain and scale and i think we have to figure out a few more big ideas along the way for the algorithms but not that many uh it's mostly like if we can just scale it up and i think now we know how to do that new hardware and other things that will go a very long way and then i think in the big areas of intelligence of sort of like unsupervised learning reasoning reinforcement learning we artificial life we have sort of multiple attack vectors on each that all seem pretty promising hi sam nice to meet you um you made two really interesting points i thought um one around entrepreneurs not really doing the hard stuff today and then the other thing about people not sticking on terms of long long time horizons i wonder what you think um in terms of how much of the people not doing hard stuff is due to the fact that it's harder to get funding for hard stuff to a certain extent i think there are certain exceptions coastal being one of them but among many firms in the valley it's easier to fund the easy stuff and the second question i have is with respect to the time horizon piece um do you think of the standard contract was a six-year vesting schedule as opposed to a four-year vesting schedule that might change things yeah in terms of funding for hard stuff i think it used to be true it used to be almost impossible and i think we owe a great debt of gratitude to elon musk who's done more than anybody else to change investors willingness to fund hard stuff also to coastal ventures and founders fund and a handful of other firms that sort of have stepped up i think today though it's really gone the other way and i want to get this message out i think what we hear consistently from investors is like we want to fund that stuff right talking to yc and when we started funding these companies at yc they had a hard time raising money they now have a much easier time raising money i think people can sort of sense that all these things we've been talking about the importance of concentrating talent for example and the difficulty of doing that without a big mission so i actually think it has become much easier to raise at least the first two rounds for a hard tech company it does get more difficult when you need to raise hundreds of millions or billions of dollars but i suspect that will change too in terms of vesting yeah i definitely think six is a lot better than four but i suspect the right answer for a lot of companies is eight or ten sam uh what are the three largest mistakes you see founders make that founders in this group are probably making like the the most common things you see as mistakes that you would advise them to change um in no order one very common mistake is to not delegate enough to an executive team um really great executives if they're good enough they could go run a company if they wanted to and i think people really value autonomy and you can give them a vision but you have to get in my experience you have to give them huge authority much more than most founders find comfortable in in what they do uh and in my experience i have been on average always surprised at the upside even though i've had plenty of individual failures and i think operating this way is uncomfortable to most founders especially most first-time founders especially when the people are more experienced and many ways more talented than than you so this is like a systemic mistake that i think a lot of founders make another is not being willing to commit internally externally publicly privately to a big vision i often see founders that come in with a big vision a lot of ambition and then they kind of the advisors around them like most advice is usually terrible but it's particularly bad in that it encourages people to have a smaller vision and not be so ambitious because people think it's scary or it's you know it sounds too crazy and so letting advisors push you off of what you actually believe when really like all the best founders i know have incredible unwavering confidence in their vision even if they change course a few times on how they get there is a really common mistake and that affects everything it affects how you hire people how you retain them how you keep a sense of momentum in the company how you talk to the press it's a huge issue um and then related to that another thing that i see founders make it's the other version of this if you lose in short-term momentum if there's like a period of time where the company because you get distracted on something or you're sort of make a wrong bet and and you lose short-term forward momentum um that's very hard to recover from and i think most founders until they've done that once don't understand how expensive the mistake that is and if they get it wrong if you do is momentum you have to like do this like all-out effort to fix it and the way you fix it is not what most people want to do which is like give motivational speeches it's just like get any win one at a time and uh that can destroy a company if you don't do that quickly one more yeah we can do one more so we had a an interesting conversation about the abysmal absence of engineering ethics education and the catastrophic implications on our political system of the unintended consequences of uh social networking over the last 10 years would you mind expanding briefly on the word safe for ai specifically oh yeah yeah agi specifically for agi specifically um you know we got a lot of criticism can i go a minute over time yeah okay um we got a lot of criticism recently for not releasing a full language model that we trained and i was very surprised by the criticism because what we said is that like we think this is something that if the world does not adapt to can have a catastrophic effect on political system and sort of manipulation of what people think you know we can now generate texts like people can photoshop images and the world doesn't know about that yet so we're gonna do this controlled release and i didn't think it was particularly controversial decision because that is like a baby baby baby baby step towards what's at stake with agi and the community had this incredibly negative reaction all research has got to be open it's got to be released this is like fear-mongering to talk about this the same journalist who would sort of say like um facebook was so irresponsible for not thinking ahead of time about the implications of what they were doing uh was saying like i can't believe open ai is just not releasing their research here and that has caused me a lot of pause because compared to what's ahead of us that was not even a warm-up and i think what we mean by safe agi uh is not just like how do you not have the robots come after us but how do you not cause harm to humans along the way how do you make sure that this is used to make people's lives met better not net worst and i think the technology industry has generally failed to be thoughtful enough about this to the credit of the industry this is a complex the internet is a complex system with a lot of emerging behavior but the thing that scares me the most about safety around ai is that it is so hard to reason about it's happening so quickly the answers are so unclear that and we don't seem to be in a place where we can have [Music] a reasonable discussion because it's so loaded with emotion on all sides um that i worry the world is just not ready for it and will not it people will not move fast far enough fast enough to really think about what's at stake here and the things that we have to do many of which are uncomfortable to the industry and particularly sort of the ai research community and what policymakers and we decided to do so i think that's going to be extremely difficult and i feel less optimistic that we are societally ready for that conversation than i did five months ago awesome well i think that's our time sam thank you for doing that

Channel: Khosla Ventures

Views: 23,984

Keywords:

Id: NAaRhXQCt9o

Length: 30min 10sec (1810 seconds)

Published: Thu Aug 01 2019