Sam Bankman-Fried - Crypto, FTX, Altruism, & Leadership

1. 0:00 – 1:03

   Preview

   1. SBSam Bankman-Fried0:00

      When you start a company, there's enormous amounts of shit that looks like that, things that are, like, dumb or annoying or broken or unfair or not how the world should work. But it's how the world does work, and the only way to be successful is to do it, is to fight through that. Most of the times that we see a company grow really fast, really quickly and get really big in terms of number of people, it becomes an absolute mess internally. I do think that sometimes EA gets too narrow-minded and specific. I think this is one of the reasons why that people end up sort of fixating on one particular understanding of the universe, of ethics, of how things are gonna progress. I think stablecoins becoming an important settlement mechanism is pretty likely, and I think blockchains in general becoming a settlement mechanism and, you know, collateral-clearing mechanism seems decently likely to me. (music)

   2. DPDwarkesh Patel0:54

      Today on the Lunar Socie- Society Podcast, I have the pleasure of interviewing Sam Bankman-Fried, CEO of FTX. Thanks for coming on the Lunar Society.

   3. SBSam Bankman-Fried1:02

      Thanks for having me.

2. 1:03 – 1:56

   How inefficient is the world?

   1. SBSam Bankman-Fried1:03

   2. DPDwarkesh Patel1:03

      All right, first question. Does the consicutive- consecutive success of FTX and Alameda, does that suggest to you that the world has all kinds of low-hanging opportunities or was that a property of the inefficiencies of crypto markets at one particular point in history?

   3. SBSam Bankman-Fried1:16

      I think it's probably more of the former. I think there are probably just a lot of inefficiencies.

   4. DPDwarkesh Patel1:20

      And so I guess, another part of this question is if you had to restart earning to give again, what are the odds you'd become a billionaire, but you couldn't do it in crypto?

   5. SBSam Bankman-Fried1:26

      I think, um, I mean, they're pretty decent. Like, I- I- I... A lot of it depends on what I end up choosing and how- how sort of, like, aggressive I end up deciding to be. You know, there are a lot of pretty safe and secure kind of career paths for, you know, before me that definitely would not have ended there. Um, but I think that if I'd sort of, you know, decided to really dedicate myself to starting up some businesses, there would have been a pretty decent chance of it.

3. 1:56 – 5:00

   Choosing a career

   1. SBSam Bankman-Fried1:56

   2. DPDwarkesh Patel1:56

      So that leads to the next question, which is that you've cited Will MacAskill's lunch with you while you were at MIT as being very-

   3. SBSam Bankman-Fried2:01

      Yup.

   4. DPDwarkesh Patel2:01

      ... important in deciding your career. He suggests that you do earning to give at... by going to a quant firm, like, um, uh, Jane Street. In retrospect, given the success you've had as a founder, was that maybe bad advice, and maybe he should have advised you to start a startup or a nonprofit?

   5. SBSam Bankman-Fried2:15

      I mean, I don't think it was literally the best possible advice in that, like, you know, I mean, that was what? 2012 or something. Like, you know, "Think about starting a- a crypto exchange," would have maybe been a- a, you know... But- but I think it was definitely helpful advice, and I think that, you know, relative to not having gotten advice at all then, um, I think it- it probably helped quite a bit.

   6. DPDwarkesh Patel2:35

      Right. But then there's a broader question of, are people like you who could become founders-

   7. SBSam Bankman-Fried2:39

      Yup.

   8. DPDwarkesh Patel2:39

      ... are they advised to take lower variance, uh, lower risk, uh, careers that, um, in expected value are, uh, less valuable?

   9. SBSam Bankman-Fried2:47

      Yeah, I think that's probably true. I think probably people are advised too strongly, um, to go down safe career paths. But I think it's worth noting that, first of all, there's a big difference between what makes sense altruistically and personally for this, and, you know, to the extent you're just thinking of personal criteria, uh, that's gonna argue heavily in favor of a safer career path because you have much more quickly declining, you know, marginal utility of money than the world does. Um, so this is sort of, like, specifically for altruistically minded people. Um, the other thing is that, you know, when you think about, like, where or what is it that- that sort of, like, is advising people to choose a safer route. Um, I think people, you know, will- will often try and look to, "Oh, well, what was the career advice that they got? What was sort of like, you know, what were sort of these outward-facing factors that- that you can see?" But I- but I think often the answer has to do something with them and their family, um, or them and, uh, you know, their friends or- or something much more personal. And, you know, when we talk with people about what they're thinking about doing with their career, you know, personal considerations and the advice of people close to them weighs really, really heavily, um, on what decisions they end up making.

   10. DPDwarkesh Patel4:01

       So I- I didn't realize that the personal considerations, uh, were as important in your case as the advice you got from EA.

   11. SBSam Bankman-Fried4:08

       Oh, I- I don't think in my case, but I think that in- in the case of- of many, many people that I talk to, they are.

   12. DPDwarkesh Patel4:13

       So speaking of declining marginal consumption, I'm wondering if you think the implication of this is that over the long term, all the richest people in the world will- will be utilitarian philanthropists because they don't have, uh, diminishing returns from consumption.

   13. SBSam Bankman-Fried4:24

       I-

   14. DPDwarkesh Patel4:24

       They're just neutral.

   15. SBSam Bankman-Fried4:24

       I mean, I wouldn't say all will, but I think there probably is something in that direction where people who are looking at sort of how they can help the world are gonna end up being disproportionately represented amongst the most and maybe least successful.

   16. DPDwarkesh Patel4:38

       All right, let's talk about effective altruism. So in our-

   17. SBSam Bankman-Fried4:40

       Yeah.

   18. DPDwarkesh Patel4:40

       ... in your interview with Tyler Cowen, you were asked, "What constrains the number of altruistically minded projects?" And you answered, "Probably someone who can start something." Now, is this a property of the world in general or is this a property of EAs? And if it's about EAs, then what do you think it's about? I- is there something about the movement that drives away people who t- could take leadership roles?

   19. SBSam Bankman-Fried4:59

       Oh, I think it's

4. 5:00 – 7:06

   The difficulty of being a founder

   1. SBSam Bankman-Fried5:00

      just the world in general. I think, you know, even if you ignore altruistic projects and just look at profit-minded ones, we have lots of ideas for businesses that we think would probably do pretty well if they were run quite well, um, that we'd be, you know, excited to fund. And the missing ingredient quite frequently for them is the right person or team to take the lead on it. Um, and I think that in general, it just... it- it's kind of brutal starting something. It's- it's sort of brutal being a founder. And, uh, it requires a- a somewhat specific but extensive list of- of skills. Um, and- and I think that, you know, those things end up making it generally fairly highly in demand.

   2. DPDwarkesh Patel5:40

      What would it take to get more of those kinds of people to go into EA?

   3. SBSam Bankman-Fried5:44

      Yeah. I mean, I think part of it is probably just talking with them about, you know, have you thought about what you can do for the world? Have you thought about how you can have impact on the world? Have you thought about how you can maximize your impact on the world? And- and just sort of going down that path, I think a lot would be amenable, and I think a lot would be excited about sort of thinking critically and ambitiously about how they can help the world. So I think honestly just engagement is one piece of this. Um, and- and then another thing, I think, you know, even within people who are, um, you know, altruistically minded in looking at, like, what would it take for them to be more excited to be founders or- or to be better at, I think there are still things that- that you can do. And I think some of this is about empowering people and some of this is about normalizing the fact that when you start something, it might fail, and that's okay.... um, and that, you know, w- that's how most startups, and especially most very early stage startups... Obviously, this sort of changes over, over time. But, um, but that, you know, when you look at sort of early stage companies, um, you shouldn't be running them. Uh, you shouldn't be trying to build them to maximize the chances of having at least a little bit of success. Um, but what that means is that you have to be okay with the personal fallout of failing, and that we have to build a community that is okay with that. And I don't think we have that right now. I think very few communities

5. 7:06 – 10:42

   Is effective altruism too narrowminded?

   1. SBSam Bankman-Fried7:06

      do.

   2. DPDwarkesh Patel7:06

      Now, there are many good objections to utilitarianism, as you know. You said yourself that we don't have a good account of infinite ethics. Should we attribute substantial weight to the probability that utilitarianism is wrong? And how do you hedge for this moral uncertainty in your giving?

   3. SBSam Bankman-Fried7:20

      So, I don't think it has super large impact on, on my giving, partially because in order to, so you'd have to have sort of a concrete proposal for what else you would do and what that would imply that would be different, you know, actions wise. And I don't know that I've sort of been compelled by many of those. Uh, I do think though that there are a lot of things we don't understand right now, and I think one thing that you pointed to is, is infinite ethics. Um, I think another thing is... and I'm not sure this is quite moral uncertainties, might be physical uncertainty more so than anything else. But, you know, there are a lot of sort of chains of reasoning people will go down that I think are, like, somewhat contingent on our current understanding of the universe in a way which might, might not be right. And certainly, if you look at, like, expected value outcomes might not be right, I think, you know, say what you will about, like, the size of the universe and what that implies. But, like, you know, some of the same people who make arguments based on, "Well, here's how big the universe is," also, you know, think there's a, you know, think the simulation hypothesis has decent probability. Um, but I think very few people sort of chain, you know, chain through then, like, "Well, okay, what, what would that imply?" I, I don't think it's clear what any of this implies. I think, in the end, if I had to say like, "How have these considerations changed my thoughts on what to do?" The, the honest answer is that they have changed it a little bit, and I think the direction that they pointed me in is things with moderately more robust impact. And what I mean by that is there's sort of one way that you can calculate the expected value of, of, of an action, which is sort of pretty specific and pretty much like, "Here's what's gonna happen. Here are the two outcomes. Here are the probabilities of them." You know. There's another thing you can do though, which is to try and say like, all right, like, m-... it's a little bit more hand wavy, but it's something like, uh, "How much better is it kinda, kinda, you know, gonna make the world?" Like, "How much does it matter if the world's kind of better in, like, generic diffuse ways?" And I, I think typically, you know, EA has been pretty skeptical of that second line of reasoning, and I think correctly, because I think that usually when you see that deployed, it's nonsense. Like, usually, I think when, when certain people are, are pretty hard to nail down on, like, what the specific reason is they think that something might be good, um, it's because they haven't thought that hard about it, um, or don't want to think that hard about it. And that, you know, the, the, the much better analyzed and vetted pathways are the ones that you should be paying more attention to. That being said, I do think that sometimes EA gets too narrow-minded and specific about plotting out sort of, like, courses of, of impact. I think this is one of the reasons why that people end up sort of fixating on one particular understanding of the universe, of ethics, of how things are gonna progress. But that, you know, all of these things have some amount of uncertainty in them, and we jostle that. Um, some, some sort of, like, theories of impact and some models behave somewhat robustly under jostling, and some of them completely fall apart. I've become, like, a little bit more sympathetic to ones that are kind of, like, a little bit robust under thoughts about what the world ends up looking like.

   4. DPDwarkesh Patel10:41

      So,

6. 10:42 – 13:40

   Political giving

   1. DPDwarkesh Patel10:42

      in the 20, uh, May 2022-

   2. SBSam Bankman-Fried10:44

      Yup.

   3. DPDwarkesh Patel10:44

      ... Oregon congressional election, um, you gave $12 million to Kerrick Flynn, who, um, whose campaign was ultimately unsuccessful. How have you updated your beliefs about the efficacy of political giving in the aftermath?

   4. SBSam Bankman-Fried10:57

      Yeah. I mean, you know, it was, uh, a first ki- time that I sort of, you know, given that scale, uh, in a race, and, you know, did it because he was, you know, of all the candidates in the cycle, the most outspoken on the need for more, uh, pandemic preparedness and prevention. Um, you know, h- he lost, obviously. Um, you know, such is life. And I, I, I think that, you know, in the end, there are some updates. I, I think lots of sort of miniature updates on efficacy of various things. Um, but, you know, also, you know, I never thought that the odds were extremely high that he was gonna win. It was always gonna be, uh, uh, an uncertain, close race. There's a limit to how much you can update from a one-time occurrence. Um, if you s-... you know, thought the odds were 50/50 and it turns out being, you know, close in one direction or, or another, right? There's sort of a maximum of maybe a factor of two update that, that you have on that. And so, you know, I think that there were a bunch of sort of micro updates just on, um, you know, specific factors of the race. But I think on a high level, um, uh, I don't think it sort of changed my perspective on, uh, policy that much.

   5. DPDwarkesh Patel12:08

      But does it make you think there are diminishing or possibly negative marginal returns from one donor giving to a candidate because of the negative PR it creates?

   6. SBSam Bankman-Fried12:15

      At some point, yeah, I think that's probably true.

   7. DPDwarkesh Patel12:17

      Um, so, uh, c- continuing on the theme of politics, when is it more effective to give the marginal million dollars to a political campaign or institution to make some change at the government level, uh, you know, like putting in early detection, or when is it more effective to just fund it yourself?

   8. SBSam Bankman-Fried12:31

      It's a good question, and, you know, part of this is that it's not necessarily mutually exclusive. Um, but, you know, I, I think one thing here is looking at what is the scale of the things that need to happen, and how much are things like international cooperation important for it? When you look at pandemic prevention, you know, we're talking tens of billions of dollars, of, of, of scale necessary to, you know, start putting this infrastructure in place, so it's a pretty big scale thing, um, which is hard to fund, you know, to that, to that level individually. Um, and it's also something where, you know, we're gonna need to have cooperation between different countries on, you know, what their, uh, you know, surveillance for, uh, new pathogens, uh, look like and on, you know, vaccine distribution, right? Like, if you... You know, some countries sort of, um, have great distribution of vaccines and others don't. That's not good. It's both not fair and not equitable to the countries that end up getting hit hardest, but also in a global pandemic, it's gonna spread, and, and so you need to have global coverage. And, and so I think that's another reason that government likely has to be involved at least to some extent in the efforts.

7. 13:40 – 17:26

   FTX Future Fund

   1. SBSam Bankman-Fried13:40

   2. DPDwarkesh Patel13:40

      Let's talk about Future Fund.

   3. SBSam Bankman-Fried13:41

      Yeah.

   4. DPDwarkesh Patel13:41

      So, as you know, there are already many existing effective altruist, um, organizations that do donations. Uh, what is the reason you thought there was more value in creating a new one? What's your edge?

   5. SBSam Bankman-Fried13:50

      So, you know, part of it is I just think that there's value in having multiple organizations. Every organization's gonna have its blind spots, and, you know, you can help cover those up if you have a few. Um, and, you know, i- if Open Phil didn't exist, maybe we would have created an organization that looks more like Open Phil. But, you know, there is some extent to which they are covering a lot of what they're looking at. You know, we're looking at overlapping but not identical things, and, and so I, I think having that diversity can be valuable. But, you know, pointing to what are the ways in which we sort of intentionally designed it to be a little bit different from, uh, existing donors? Um, one thing that I think I've been really happy about has been the regranting program, um, that we've had. So we have, you know, a number of people who are experts in, in various areas, um, who we've basically donated pots to that they can regrant. And what are the reasons that we think this is valuable? One thing is just giving more stakeholders, you know, a chance to sort of voice their opinions in a way where we can't possibly sort of listen to everyone in the world directly and integrate all those opinions to come up with, like, the perfect set of answers. And so distributing it and letting them act semi-autonomous- autonomously can help, um, with that. Uh, but the, the other thing is that it really helps with large numbers of smaller grants. And so, you know, when you think about what, you know, an organization giving away $100 million in a year is thinking about, um, if we divided that up into $25,000 grants, right? Like, how many grants would that, that mean? That would mean, uh, what, like, uh, 4,000 grants, um, which is a l- a lot of grant to a- to analyze, right? Like, you know, if we wanna give real thought to each one of those, we can't do that. But on the flip side, sometimes the smaller grants are the most impactful per dollar, and there are a lot of cases where someone really impressive has a really exciting idea for a new foundation, for a new organization that could do a lot of good for the world, and needs $25,000 to get it started, right? To, like, rent out a small office, um, to be able to cover salaries for two employees for the first six months. Um, those are the kind of cases where sometimes a pretty small grant can make a huge change in the development of what might ultimately become a really impactful organization. But they're the kind of things that are really hard for our team to evaluate all of just given the number of them. Um, but the regranter program gives us a way to do that where, you know, if instead we have, you know, 10, 50, 100 maybe eventually regranters who are, uh, you know, going out and finding a lot of those opportunities close to them, they can then sort of identify those and, and, and direct those grants, and it gives us a much wider reach and, you know, also biases it less towards people who we happen to know, um, which is, which is good. We don't wanna just overfund everyone we happen to know and underfund everyone that, that we, we didn't happen to. So I think that's been sort of one initiative we've had which I've been pretty excited about, um, and, uh, you know, I think we're gonna, we're gonna keep doing. Um, and then, you know, I think another thing is what we've really tried to have a lot of emphasis on making the process smooth and clean. Um, there are pros and cons to this, but I do think that it sort of like drops the activation energy necessary for someone to decide to apply for a grant and, you know, fill out, um, all of the forms and things like that, and so we've really tried to bring more people in the fold, eh, you know, in, in terms of potential, uh, recipients.

8. 17:26 – 18:51

   Adverse selection in philanthropy

   1. DPDwarkesh Patel17:26

      If you make it easy for people to fill out your application and-

   2. SBSam Bankman-Fried17:29

      Yup.

   3. DPDwarkesh Patel17:29

      ... if it's generally, uh, you're finding things that maybe other organizations wouldn't, how do you deal with the possibility of adverse selection in your philanthropic dual flow?

   4. SBSam Bankman-Fried17:36

      It's a really good question, and of course that, that's a worry that, you know, Bob down the street might, like, you know, see a great bookcase that he wants and be like, "Oh, man, I wonder if I can get funding for this bookcase. It's gonna house, you know, house a lot of knowledge."

   5. DPDwarkesh Patel17:49

      (laughs)

   6. SBSam Bankman-Fried17:49

      Knowledge is good, right? Um, uh, I mean, obviously with... Not that one I think we would-

   7. DPDwarkesh Patel17:55

      Right.

   8. SBSam Bankman-Fried17:55

      ... detect pretty quickly. Um, and, and I think the basic answer is that, you know, we still have vetting on all of these, um, and so, you know, we, we have, you know, we do have, have oversight of them. But what we also do is we do really deep dives into both all, all of the s- the large ones but also into sort of samplings of all the small ones. You know, we do some oversight of all of them, but, but we will do really deep dives into randomly sampled subsets of them, which allows us to get a pretty good statistical sense of whether we are facing significant, um, you know, adverse selection in them.Um, you know, so far we haven't seen all these signs of it but we're gonna keep doing these analyses and, you know, see if anything, anything worrying comes out of those. But that, that's, sort of, a way to be able to, um, you know, have more trusted analyses for more scaled up numbers of grants.

   9. DPDwarkesh Patel18:49

      Mm-hmm. Um,

9. 18:51 – 23:00

   Relationship between different causes

   1. DPDwarkesh Patel18:51

      so a long time ago you wrote a blog post about how EA causes are multiplicative instead of, um, instead of additive-

   2. SBSam Bankman-Fried18:56

      Yep.

   3. DPDwarkesh Patel18:56

      ... and we talked about that a little while ago. Um, are, do you still find that that's the case with most of the causes you care about or are there cases where some of the causes, all you care about are negatively multiplicative? Like, an example might be economic growth and the speed at which it has

   4. SBSam Bankman-Fried19:08

      Yeah, I think it's getting more complicated. And I think that, I think specifically around AI you have a lot of really complex factors that point sometimes in the same direction, sometimes in opposite directions. And I think that especially if what you think matters is something like the relative progress of AI safeties re- safety research versus AI capabilities research, a lot of things are gonna have the same impact on both of those. And thus, confusing impact on, you know, safety, i- i- you know, a- a- as a whole. Um, so I, I do think it's more complicated now and I, I think it's not, sort of, cleanly things just multiplying with each other. I do think there are lots of cases where you see multiplicative behavior but I also think there are cases where you just don't have that. And that, you know, sort of, the conclusion of this is if you do have multiplicative cases, you probably want to be funding each piece of it. Um, but if you don't then you probably wanna be trying to identify the most impactful pieces and specifically moving those along. And, and, and so I think, you know, we, our behavior should be different in those two scenarios.

   5. DPDwarkesh Patel20:08

      If you think of your philanthropy from a portfolio perspective-

   6. SBSam Bankman-Fried20:10

      Yep.

   7. DPDwarkesh Patel20:10

      ... is correlation good or is it bad? (laughs)

   8. SBSam Bankman-Fried20:14

      Ah, I mean, like, I don't know. Expected value is the expected value, right? Like, and, and maybe here's, like, one way to think about this. Let's pretend that there is, you know, one person in, um, uh, Bangladesh and another one in Mexico. And we have, you know, one intervention that can, you know... We have two interventions, both, both 50/50, um, on saving each of their lives in particular, right? Some, some new, new, new drug that we could help, you know, release to combat some neglected disease. Um, and, and then there's the question of like, well, are they correlated? Like, are these two drugs correlated in their efficacy? And, and my basic argument is it doesn't matter, right? Because if you think about it from each of their perspectives, right? The person in Mexico isn't saying like, "I only wanna be saved in the cases where the person in Bangladesh is or isn't saved," right? Like, that's not relevant, right? They're like, "I, I, I would like to live," and the person in Bangladesh similarly says, "I would like to live." And, you know, you want to help both of them as much as you can. Um, and it's not super relevant whether, you know, there's a alignment or anti-alignment between the cases where you get lucky and the ones where you don't.

   9. DPDwarkesh Patel21:30

      Um, what's the most likely reason that Future Fund lives, uh, fails to live up to your expectations?

   10. SBSam Bankman-Fried21:35

       I, I think we just, like, kind of get a little lame. Like, we, we give to a lot of decent things but, like, all the cooler or more innovative things that we do just don't seem to work very well. And we end up sort of giving the same where- the same place, you know, that everyone else is, is giving, wherever that, that ends up being. And that, you know, we're not... don't turn out to be effective at starting new things. We don't turn out to be effective at thinking of new causes or executing on them. Um, and, uh, you know, hopefully we'll avoid that but it's always a risk.

   11. DPDwarkesh Patel22:06

       So, should I think of your charitable giving as a yearly contribution of a billion dollar or less or more? Or should I think of it as a $30 billion hedge against the possibility that there's gonna be some existential crisis that, um, requires a large pool of liquid wealth?

   12. SBSam Bankman-Fried22:21

       It's a really good question. I'm not sure. Um, you know, we're gonna start giving some, we already have. We've given away about 100 million so far this year. Um, and you know, we're going to start doing that partially because we think they're really important things to fund. Partially because we wanna make sure to start scaling up those systems and that process so that we're ready, um, and so that, you know, we notice opportunities as they come by and we have, you know, systems ready in place to give to them. Um, but I think it's something we're really actively discussing internally how concentrated versus diffuse we want that giving to be. And, you know, how much we wanna be sort of storing up for one very large opportunity versus how much it's gonna be sort of, uh, you know, mixture of many.

   13. DPDwarkesh Patel22:59

       When

10. 23:00 – 26:36

    Doing difficult things

    1. DPDwarkesh Patel23:00

       you look at a proposal and you think, "This project could be promising but this is not the right person to lead it-"

    2. SBSam Bankman-Fried23:04

       Yep.

    3. DPDwarkesh Patel23:04

       ... what is the trait that's most often missing?

    4. SBSam Bankman-Fried23:06

       Ah, super interesting. Um, there is... Uh, I'm gonna sort of, like, ignore the obvious answer set which are like, "The guy's just not very good," um, uh, which sure, fine. Um, and, and maybe look at cases where it's someone who, like, is pretty impressive but, like, I still think is not the right fit for this. Um, I, I, I think there are a few things. I think one of them is, how much are they going to want to deal with really messy shit? This is a huge thing. If you go to work for, uh, like... A- and maybe to give some example, like, when I was working at Jane Street, um, it's a really great place. Um, you know, had a great time there. One thing, which I didn't even realize was, you know, valuable there, until I saw the alter- you know, saw, saw sort of what, what things could look like outside was, you know, i- if I decided that it was a good trade to buy one share of Apple stock, um, on NASDAQ, uh, I... You know, there's like a button to do that, right? Um, if you, as a random, you know, citizen want to buy one share of Apple stock directly on an exchange, it'll cost you tens of millions of dollars and a year to get set up to be able to do that. Like, you gotta get like a physical colo, maybe? Like, in a, in, you know, Secaucus, New Jersey? L- like, like, you know, you, you have to, like, have market data agreements with these companies. You have to think about the SIP and about the NBBO and whether you're even allowed to lift on NASDAQ right then. Um, you have to build technological infrastructure to do it.... but all of that comes after you get a bank account, and let's even talk about that stuff. Getting a bank account that's gonna work in finance is really hard. Um, w- I probably spent hundreds, if not thousands, of hours of my life trying to open bank accounts. Um, and, you know, one of the things at early Alameda that was really crucial to our ability to make money was having someone very senior spend, um, hours per day in a physical bank branch, manually instructing wire transfers. And if we didn't do that, we wouldn't have been able to do the trade. Um, and when you start a company, there's enormous amounts of shit that looks like that, things that are, like, dumb, or annoying, or broken, or unfair, or not how the world should work. But it's how the world does work, and the only way to be successful is to do it, is to fight through that. And if you're gonna be like, "Ah, whatever," like, "I'm the CEO, I don't do that stuff," right? Then no one's going to do that at your company. It's not going to get done, you won't have a bank account, and you won't be able to operate. So, one of the biggest traits that I think is incredibly important for a founder, um, and for, like, an early team at a company, um, but that is not necessarily important for everything that you might want to do in life, is being willing to do a ton of grunt work if that's what's important for the company right then. Um, and viewing it not as, like, low prestige or, like, too easy for you or something like that, but as, "Whatever, this is the important thing, this is the valuable thing to do, so it's what I'm gonna do." Um, that's one of the, the, I think, core traits. And the other one is like, are they excited about this idea? Will they actually put their heart and soul into it, um, or are they kind of gonna be a little bit drifting and bored and not really into it and half-ass it? I think, like, those are two things that I really look for.

    5. DPDwarkesh Patel26:35

       How have you

11. 26:36 – 29:14

    The importance of focus

    1. DPDwarkesh Patel26:36

       used your insights about pitcher fatigue to allocate talent in your companies?

    2. SBSam Bankman-Fried26:39

       Uh... (laughs) Um, uh, so, uh, pitcher fatigue, um, is, uh... I haven't thought about this in a while. But, but, my thesis back then, which I still think is probably true, is that when it comes to, to pitchers, um, in, in baseball, uh, there's a lot of evidence that they get worse over the course of a game. Just the more innings they pitch, like, they get worse, and worse, and worse. Partially, it's just, like, it's hard on the arm. Um, but it's worth noting that the evidence seems to support the claim that i- it depends on, on the pitchers, but that, in general, you're better off breaking up their outings. That, like, it's not just a function of how many innings they pitch that season, but also extremely recently. And so if you could choose between someone throwing six innings every six days or throwing three innings every three days, probably you should choose the latter. Probably that's gonna get the better pitching on average and just as many innings out of them. Um, and for its worth, baseball actually has since then moved very far in that direction. Like, it, you know, an average number of pitches thrown by starting pitchers is down a lot, um, over the last five, ten years. Um, uh, how do I use that in my company? Uh, well there's a metaphor here, but I actually think I've gone the opposite direction, if anything. And, and, and here's sort of what my sense has been in terms of, you know, m- uh, computer work instead of, like, you know, arm... like, physical work is that, um, you don't have the same effect whereby, like, uh, you know, your arm is getting sore and eventually your muscle snaps and you need surgery if you pitch too hard for too... Like, that, that sort of, like, doesn't directly translate. There's a little bit of an equivalent of this, of people getting tired, right, and exhausted. But on the other hand, context is a huge, huge piece of being effective. Having all of the context in your mind of what's going on, of what you're working on, what the company's doing, makes it way easier to operate effectively. And if you could, for instance, have two half-time employees or one full-time employee, you're way better off with one full-time employee because they're going to have way more context than either of the part-time employees would have, and thus be able to work way more efficiently. And so, in general, I think our experience has actually been that, like, concentrated work is pretty valuable and that, like, if you keep breaking up your work, and whatever, it depends on the person, the context. But, like, in general, if you do that, you're never gonna be able to do as great of work as if you really

12. 29:14 – 31:54

    How SBF identifies talent

    1. SBSam Bankman-Fried29:14

       dove into something.

    2. DPDwarkesh Patel29:15

       So you've talked about how, uh, you, they experience relatively little when you're deciding who to hire. But in a recent Twitter thea- thread, you mentioned that mentorship is... or being able to provide mentorship-

    3. SBSam Bankman-Fried29:25

       Yep.

    4. DPDwarkesh Patel29:25

       ... to all the people who come on, that's one of the bottlenecks Yeah. ... to you being able to scale. Is there, uh, a trade-off here where if you don't, if you don't, um, hire people for experience and you gotta give them more mentorship and thus can't scale as fast?

    5. SBSam Bankman-Fried29:36

       It's a good question, but to a surprising extent, we found that the experience of the people that we hire has not that much correlation with how much mentorship they need. That much more important is how they think, how good they are at understanding new and different situations, um, and how good they are and how hard they try to integrate into sort of their understanding of, you know, let's say coding, their understanding of how FTX works. And, um, a- and so I, I think that we actually have by and large found that, like, other things are much better predictors of how much, you know, oversight, and management, and mentorship someone is gonna need than their experience at sort of similar looking roles.

    6. DPDwarkesh Patel30:19

       And how do you assess that short of hiring them for, them for a month and then seeing how they did?

    7. SBSam Bankman-Fried30:23

       It's tough. I don't think we're perfect at it. Um, but things that we look at, you know, do they understand quickly what the goal of a product is and how does that inform how they build it? You know, when you're looking at developers, I think we really strongly want people who can understand what FTX is, how it works, and thus what the right way to architect things would be for that, rather than sort of like treating it c- as, like, an abstract engineering problem divorced from whatever the ultimate product is. So being able to... And so that's something that you, you can try and ask people like, "Hey, here's, like, a high-level customer experience or customer goal."... right? How would you architect a system to create that? Um, so that's, that's one thing that we look for. Just an eagerness to learn and to, to, you know, adapt. Um, it's not trivial to test for that, but you can do some amount of that. You can try and give people sort of novel scenarios and see how much they break, um, versus how much they, they bend. Um, and I think that can be super valuable as well. Uh, and, uh, a- a- and, you know, also kind of like specifically searching for developers who are, you know, willing to deal with messy scenarios rather than wanting sort of a pristine world to work in. Um, because our company, it's customer-facing, it has to face some thirty per- third-party tooling. We've been a quickly growing company. All those things mean that, you know, we have to interface with things that are messy in the way the world is.

    8. DPDwarkesh Patel31:53

       Now before

13. 31:54 – 34:36

    Why scaling too fast kills companies

    1. DPDwarkesh Patel31:54

       you launched FTX, you gave detailed instructions to the existing exchanges about how to improve their system-

    2. SBSam Bankman-Fried31:59

       Yep.

    3. DPDwarkesh Patel31:59

       ... how to remove clawbacks and so on. Looking back, they left billions of dollars of value on the table.

    4. SBSam Bankman-Fried32:03

       Yep.

    5. DPDwarkesh Patel32:03

       Why do you think that was? Why didn't they just fix what you told them to fix?

    6. SBSam Bankman-Fried32:06

       Yeah. Um, it's a really interesting question. And my sense is that it's, uh, it's part of a larger phenomenon where... (sighs) What's the right way to put it? Like, so, okay, o- one piece of this is just, like, they didn't have a lot of market director experts. Like, they just did not have the talent in-house to be able to, like, think really well and deeply about risk engines. Um, and also there were cultural barriers between, you know, myself and some of them, which I think probably meant that they, you know, were less inclined than they otherwise would have been to sort of take it very seriously. Um, but i- ignoring those factors, I think there's something much bigger at play there, where, you know, many of these exchanges had hired a lot of people. They'd gotten very large. Um, and you might think that that meant that they got more able to do things because they had, you know, more sort of like horsepower. But in practice, most of the times that we see a company grow really fast, really quickly and get really big in terms of number of people, it becomes an absolute mess internally. There's huge diffusion of responsibility issues. No one's really taking charge. You can't figure out who's supposed to do what. And in the end, nothing gets done. And you actually start hitting negative marginal utility of employees pretty quickly, um, where the more people you have, the less total you get done. I think that happened to a number of them to the point where, like, yeah, I sent them these proposals. Where did they go internally? Who knows? You know, the, the, like, you know, vice president of exchange risk operations, but not the real one, the sort of fake one operating under some department with an unclear goal and mission or something like that, who, like, had no idea what to do with it and eventually just sort of, like, passed it off to a random friend of hers that she knew who was a developer for the mobile app and was like, "You're a computer person. Is this right?" And was sort of like, "I have no idea. I'm not a risk person." And that's how it died. And, and I'm not saying it was literally that happened, but something that sounds kind of like that probably happened, where it just, like, it's not like they had, like, you know, people who took responsibility. If they saw this was like, "Wow, this is scary. I should make sure that the best person in the company gets this and pass it to the TTO and, like, person who thinks about their modeling" and said, like, "Hey, is this thing scary?" And they looked at it, and they're like, "Wow, this might be a problem." I don't think that's what happened.

    7. DPDwarkesh Patel34:35

       Now,

14. 34:36 – 36:31

    The future of crypto

    1. DPDwarkesh Patel34:36

       there's two ways of thinking about the impact of crypto on financial innovation. One is the crypto maximalist view that crypto subsumes TradFi.

    2. SBSam Bankman-Fried34:44

       Yep.

    3. DPDwarkesh Patel34:44

       The other is that what you're basically doing is you're, uh, stress testing some ideas from, um, uh, s- uh, in a volatile, fairly unregulated market-

    4. SBSam Bankman-Fried34:53

       Yep.

    5. DPDwarkesh Patel34:53

       ... that you're actually gonna bring to TradFi, but, uh, this is not gonna lead to some sort of decentralized utopia. Um, so which of these models is more correct? Or is there a third model that you think is the correct one to be used?

    6. SBSam Bankman-Fried35:03

       So first of all, who knows? Right? Like, (laughs) I mean, you know, who knows exactly what's gonna happen? It's gonna be path dependent. Um, but, you know, if I had to guess, I would say a lot of properties of what is happening in crypto today will probably make their way into TradFi to some extent. I think blockchain settlement has a lot of value and can clean up a lot of areas of traditional market structure. Um, and, uh, and I, I think that, you know, composable applications are super valuable, um, and are gonna, you know, get more important over time. I think there are some areas of this where it's not clear what's gonna happen, and I think that when you think about how do decentralized ecosystems and regulation intersect, it's a little bit TBD exactly where that ends up. Um, and, uh, and, and so, uh, you know, I don't wanna state with extreme confidence exactly what will or won't happen, but I think some piece of this, of, like, seem pretty likely to me. I think stable coins becoming an important settlement, um, mechanism, is pretty likely. And I think blockchains in general becoming a settlement mechanism and, you know, collateral clearing mechanism seems decently likely to me. Um, and, uh, and more and more assets getting tokenized seems decently likely to me. Um, and, you know, there being programs written on blockchains that, you know, people can add to, that can compose with each other seems pretty likely to me. Um, and, you know, a lot of other areas of it, I think, uh, could go either way.

15. 36:31 – 41:45

    Risk, efficiency, and human discretion in derivatives

    1. SBSam Bankman-Fried36:31

    2. DPDwarkesh Patel36:31

       Let's talk about your proposal to the CFTC-

    3. SBSam Bankman-Fried36:33

       Yep.

    4. DPDwarkesh Patel36:33

       ... to replace Futures Commission merchants with, um, algorithmic real time-

    5. SBSam Bankman-Fried36:37

       Yep.

    6. DPDwarkesh Patel36:37

       ... risk management. Um, there's a worry that without human discretion-

    7. SBSam Bankman-Fried36:41

       Yep.

    8. DPDwarkesh Patel36:41

       ... you have algorithms that will be... that will cause liquidation cascades-

    9. SBSam Bankman-Fried36:45

       Yep.

    10. DPDwarkesh Patel36:45

        ... when they were not necessary. Is there some room for human discretion in, uh, in these kinds of situations?

    11. SBSam Bankman-Fried36:51

        There is, and the way I think about it is you have, you know, the, the way that traditional future market structure works is you have a clearing house with a decent amount of manual discretion in it-... connected to FCMs, some of which use human discretion and some of which use automated risk management algorithms with their clients. Um, and generally, the smaller the client, the more automated it is. We are inverting that to some extent where at the center you have an automated clearing house, then connected to, you know, potentially connected to FCMs, which could potentially use, um, you know, discretionary systems when managing their clients. The, the key difference here is that one way or another, initial margin has to end up at the clearing house, a programmatic amount of it, and the clearing house acts in a clear way. Um, and the goal of this is, first of all, to s- prevent contagion between different intermediaries. So, whatever decisions, whatever credit decisions one intermediary makes with respect to their customers doesn't pose risk to other intermediaries because someone has to post the collaterals of the clearing house in the end, um, whether it's the FCM, their customer, or someone else. Um, and so it, it gives clear rules of the road and lack of sort of systemic risk spreading throughout the system and contains risk to the parties that choose to take that risk on, um, you know, to the FCMs that choose to, you know, make credit decisions there. So, I, I think that, you know, there is a potential role for, uh, for manual judgment, um, but, you know, manual judgment, it can be really valuable and add a lot of economic value. It can also be very risky when done poorly. Um, and, uh, and I think that, you know, in the current system, each FCM is exposed to all of the manual bespoke decisions that each other FCM is making. And that's a really scary place to be in. We've seen it blow up. We saw it blow up with LME nickel contracts, you know, and we saw it blow up in other cases, you know, with, uh, uh, you know, few very large traders, um, who had positions on a, a number of different, uh, banks and, and, you know, ended up blowing out. Um, so I think that this provides a level of clarity and, and oversight, um, and transparency to the system so that people know what risk they are or are not taking on.

    12. DPDwarkesh Patel39:13

        Are you replacing that risk with another risk which is that if there's one exchange that has the most liquidity in futures, and, um, if that's what... if there's one exchange where you're posting all your collateral, so across all your positions, then, um, the risk is that that single algorithm that the exchange is using is gonna determine when and if liquidation cascades happen?

    13. SBSam Bankman-Fried39:32

        So, it's already the case that, you know, if you put all of your collateral with a prime broker, um, then, you know, potentially, whatever that prime broker decides, whether it's an algorithm or a human or something in, in between, is gonna decide what happens with all of your collateral. And if you're not comfortable with that, you could choose to spread it out between different venues. You know, you could choose to use one venue for some products, another venue for other products if you don't want to cross-collateralize, cross-margin your positions. You get capital efficiency generally for cross-margining them, even for putting them in the same place. Um, but the downside of that is that, you know, the risk of one can, can affect the other one. Um, there's a balance there. And, you know, I don't think it's a binary thing.

    14. DPDwarkesh Patel40:12

        Okay. Um, but, uh, given the benefits of cross-margining and the fact that less-

    15. SBSam Bankman-Fried40:16

        Yeah.

    16. DPDwarkesh Patel40:16

        ... capital has to be locked up as collateral, is the long-run equilibrium that the single exchange will win? And if that's the case, then in the long run, there won't be that much competition in derivatives?

    17. SBSam Bankman-Fried40:25

        I don't think it... I mean, you already could see that happening. I, you haven't... and I don't think we're gonna have a single exchange winning. Um, among other things, I think, y- you know, there are gonna be different decisions made by different exchanges which will, you know, be better or worse for particular situations. And I think, you know, one thing that people have brought up is, "Well, how about for physical commodities, um, you know, like corn or, or soy?" Um, you know, what, what, like, what would our risk model say about that? And the answer is it's not super helpful for those commodities right now 'cause it doesn't know how to understand a warehouse. And so, you know, you might wanna use a different exchange which had a more bespoke risk model that, you know, tried to understand, you know, how the human understand what physical positions, you know, someone had on. Um, I think that would totally make sense and, you know, that can cause a, a sort of split between different, um, different exchanges. Um, in addition, you know, we've been talking about the clearing house here, but many exchanges can connect to the same clearing house. Um, and, you know, we're, we are already as a clearing house connected to a number of different, uh, DCMs and so excited for that to continue to, to, to grow out. And, and, you know, in general, there are gonna be a lot of people who have different preferences over sort of different details of the system and, you know, choose different products based on that. I think that's how, how it should work and that, you know, people should be allowed to choose the option that makes the most sense

16. 41:45 – 42:41

    Jane Street vs FTX

    1. SBSam Bankman-Fried41:45

       for them.

    2. DPDwarkesh Patel41:45

       What are the biggest differences in culture between Jane Street and FTX?

    3. SBSam Bankman-Fried41:49

       I think, you know, FTX is... has much more of a culture of, like, you know, morphing and taking on a lot of random new shit. And Jane Street has... it's still like... I don't wanna say it's like an ossified place or anything. Like, it is som- somewhat nimble, but it is more of a culture of like, you know, "We're going to be very good at th- at this particular thing on a time scale of a decade." Um, and there are some cases where that's true with FTX 'cause some things are clearly part of our, you know, core business for a decade. But there are other things that like, you know, we knew nothing about a year ago and all of a sudden we have to get good at. And so I think that there is, um, uh, you know, been more adaptation and, um, uh, and it's also a much more public-facing and customer-facing business than Jane Street is, which means that there are lots of things like PR that are much more sort of central to what we're doing.

17. 42:41 – 43:44

    Conflict of interest between broker and exchange

    1. SBSam Bankman-Fried42:41

    2. DPDwarkesh Patel42:41

       Now, in crypto, you're combining the exchange and the broker.

    3. SBSam Bankman-Fried42:44

       Yeah.

    4. DPDwarkesh Patel42:44

       They seem to have different incentives. The exchange wants to increase volume, the broker wants to better manage risk, um, maybe with less leverage. Um, do you feel that in the long run, these two can stay in the same entity given the conflict of interest or potential conflict of interest?

    5. SBSam Bankman-Fried42:57

       I think so, and I think that the, uh, uh... there is, like, some extent to which they differ but there are I think more sense to which they actually want the same thing, and harmonizing them can be really valuable. And one is to provide a great customer experience. And when you have two different...... entities with two completely different businesses, but that have... Every order has to go from one to the other, right? You're gonna end up getting sort of, like, the least common denominator of the two as a customer. You're gonna get only things that are... Everything is gonna be supported as poorly as whichever of the two entities support what you're doing most poorly, and that makes it harder. Whereas, you know, by, by, you, use- you know, synchronizing them, um, it, it gives us much more ability to, uh, uh, provide a great experience on that.

18. 43:44 – 44:32

    Bahamas and Charter Cities

    1. DPDwarkesh Patel43:44

       How has living in the Bahamas impacted your, uh, impacted your opinion about the possibility of successful charter cities?

    2. SBSam Bankman-Fried43:51

       It's a good question. I think it's... I mean, it's the first time. You know, I think it's updated positively a little bit. I think, you know, we've built out a lot of things here, and that's been hopefully impactful, and I think, you know, it's made me feel like it is more doable than I previously would have thought. But it's also... it's, it's a lot of work. Like, you know, it, it, it's a large-scale project if you wanna actua-... And we have not built out a full city, like-

    3. DPDwarkesh Patel44:13

       (laughs)

    4. SBSam Bankman-Fried44:13

       ... we've built out some specific pieces of infrastructure that we needed. We've gotten a ton of support from the country and, you know, they've been very welcoming, and there are a lot of great things here. And so, this is way less of a project than just taking a giant empty plot of land and creating a city in it. That, that, that, that's way harder.

19. 44:32 – 46:10

    SBF’s RAM-skewed mind

    1. SBSam Bankman-Fried44:32

    2. DPDwarkesh Patel44:32

       How has having a RAM-skewed mind influenced the culture of FTX and its growth?

    3. SBSam Bankman-Fried44:37

       It's a good question, and I, you know, I think that what it means on the upside is that we've been sort of, like, pretty good at adapting and pretty good at understanding what the important things are at any time and at, you know, training ourselves quickly to be good at those, um, you know, even if it looks very different than what we were doing, you know, before. And I think that, you know, that's allowed us to, you know, focus a lot on product, to focus a lot on regulation and licensing, uh, to focus a lot on customer experience, o- on branding and, and, and a bunch of other things. Um, uh, and, and, and, and I think hopefully it means that we're able to sort of, like, take whatever situations come up, um, and provide sort of, like, reasonable feedback about them and reasonable thoughts on what to do, um, you know, rather than sort of, like, thinking more rigidly in terms of how, you know, previous situations were. On the flip side, you know, I, I think that it means that, you know, I have to have a lot of people around me who will try and remember what the most, you know, what the sort of, like, long-term important things are that might get lost day to day, you know, as we focus on, you know, things that pop up. And, you know, it's important for me to take time periodically, to take... step back and, you know, clear my mind a little bit and just think like, "All right. Let's try and just remember what the big picture is here. What are the most important things for us to be focusing on?"

    4. NANarrator45:59

       (instrumental music)

Episode duration: 46:10