Numerai founder Richard Craib: Situational Awareness's mistake was volatility, not leverage

Speaker 1: Google. So we can dig into this more. There's a whole bunch of deeper questions about what is the actual efficacy of Meta spending on AI, how much are they spending on tokens, how much are they spending on headcount, all these things matter. But we'll dig into it another time because we have Richard Craig from Numerai here with us in the TV Penult Show. Richard, how are you doing? Hey. Doing great. How are you? Thank you so much for taking the time. I'm so glad that you had a chance to hop on the show on such short notice. Very excited to talk to you. But could you kick us off with a little bit of background for anyone who's unfamiliar with your career, your business, and why it's relevant to talk to you today?

Speaker 3: Yeah. So I am a hedge fund manager. I started my hedge fund Numerai in San Francisco in December 2015. Okay. Before that, I studied mathematics. And I was very, very interested in AI from a very early age. And in 2012, some of the key breakthroughs happened with self driving cars and things like that. You could really see that this would be a really good thing to kind of bet on in the future. But Numerai, we're a quant hedge fund, so we trade stocks using AI. We don't simply invest in AI companies. Although pretty much every company is in some way entangled with AI, so it has a lot to do with us.

Speaker 1: And in terms of like the fund structure, how similar or different is it to what we're talking about with situational awareness where there's like a few LPs, maybe some institutions invested? How broad is the LP base? How concentrated are the bets? How big is the investment committee? How are investment decisions made?

Speaker 3: So, yeah, I mean, we're trying to make, alpha, which is a kind of a technical term. Like, we can't take any market risk. We can't take any factor risk. We can't, go go off on anything we think is is you know, some something's gonna be big. Okay. We can't invest in that except for through our strategy, which is a low volatility strategy. Mhmm. The fund has been growing a lot. Assets are up a 100% a year for two years now. Congratulations.

Speaker 1: Well,

Speaker 3: remember that's a lot of that's AUM growth. But the point of the company is to make this kind of thing called alpha, which I feel like is a term that gets abused a lot. And we are trying to make sure that our investors can make a better portfolio by holding a piece of us. Yes. So

Speaker 1: help me understand Alpha in the context of situational awareness because I've heard a lot of people throw around, it's just levered beta. It's beta. But when I think about situational awareness and the ideas and the thesis that was contained in that PDF that was released almost two years ago, that felt contrarian. It felt like alpha in the sense that you can understand the future in a way that other people can't. Is that not alpha?

Speaker 3: Yeah, it is. Kind of isn't in a naive sense. It is. It's obviously the situational awareness, you know, blog posts that came out was sort of super visionary and fun. I mean, read it a whole thing. We had a mutual friend. I think we were at Joey's wedding together in Vienna. And I said, Joey, can you please introduce me to Leopold? This is so cool. So there's a place for that type of fund where you're going to take a big bet with a lot of risk. And those types of funds can obviously produce extraordinary returns. But the trouble is the driver of the return is the risk more than the alpha. Okay. So the alpha is there, but the driver of the return is is the risk. And so when people retreat to a more classic example of alpha,

Speaker 1: what are we talking about? What what what is the purest source of alpha?

Speaker 3: Yeah. So I mean, the the here's the naive way to have alpha. Right? If you if you if you have if you beat the market, you have alpha. Okay. So natural. Why why what's what's so hard about that? If if my portfolio is up 20%, but the market's up 9%, I have alpha. But that actually is still kind of missing the point in actually a very deep way. And so the problem is there's not just one factor. You have to look at do you beat the market? You just have to look at do you beat all the factors

Speaker 4: that there are in the market? So there's a factor called momentum,

Speaker 3: there's a factor called technology, there's a factor called US technology, there's a factor called momentum times beta. Yeah. Okay. All of these things in the, as a term of art in quantitative finance, you call that risk premia. Sure. They're risks and they might have a premium but they're actually not the types of things you would call alpha. And so a fund like Numerai, we're trying to hedge to things, so many things, thousands of different risks so that if someone ever looks at our portfolio and wants to say, do you have an offer or are you making money just from risk? They can't argue that. It's very hard to say that. So that's the goal. And that's the goal of all hedge funds, that is Citadel or a Millennium. They're trying to make this kinda alpha.

Speaker 1: How much should I if I'm if I'm trying to understand if I'm if I'm looking at a fund's returns over a number of years, how much should I be able to identify alpha purely by uncorrelated results with the broader market? Like, is if the market goes down 10%, the fund goes up 20%, then the market goes up and the fund goes down. Is that is that giving me, like, a sense of alpha that at least this fund is searching for that, or is that just uncorrelated returns and they're just throwing throwing darts So in the if you're uncorrelated

Speaker 3: from all the factors, you have alpha. Okay. I think there's a sort of sleight of hand that, discretionary investors, hedge fund managers tend to pull, which is they say, well, look, we believe in The US, we believe in AI, we want to take that factor risk because that's part of our return and which a sophisticated investor would say, we can take that risk ourselves without paying fees. Yeah. There is nothing stopping any LP of situational awareness from buying anthropic shares themselves. Right? There's there's not if if and there's nothing stopping them from going long micro

Speaker 1: Micron

Speaker 3: or something. It's really, like, almost childish to think that they a sophisticated investor wouldn't be able to pull those trades themselves. And so to take a personal example, here's a fun thing. So I bought, it's a good day to say this, I bought some Amazon options.

Speaker 1: Okay.

Speaker 3: Yeah. I bought some cool options on Amazon. I'm a hedge fund manager. You know, I don't really trade very much, but I just think it's a cool company, and it's gonna benefit from AI. And I and one of my friends said it would be a good thing to buy. That's it. Okay? Now my call options are up, whatever, 300% today. Yeah. Whatever it is. Does that mean I'm on a generational run? Does that mean I'm a genius investor? No, it just means a gamble I took paid off and it paid off actually probably appropriately for the risk I took. So in the hedge fund industry, you know, that the banality of that is extreme. No one is no one at numerize talking about how I bought Amazon options and made money. It's like so boring. Okay.

Speaker 2: But so so so e even the timeline, like, the the the length that Leopold was, you know, it was, a year and a half. He was on this, like, generational one as as people say. Is is there no difference between that and just Yeah. When does a generational run start? How do you define Is it eleven months and thirty days?

Speaker 3: I think it takes a generation, actually. Oh. I think that's how you mentioned the investment performance of someone over a very long time. Think you're right there. So that's that's why it's kind of a funny term. Okay. So wait. But

Speaker 1: what do you think about the idea that certain managers, might it might be possible in theory to, like, copy them, but they are getting paid to do something that you might emotionally not have the resilience to do. I'm thinking of this Jeremy Giffon post. He says, People really miss that buy and hold means the ability to buy and hold, not that you should buy and hold. And he quotes this screenshot from the Financial Times that says, since 2010, Warren Buffett sold his entire holdings in 63 positions with an average hold time of four years and three months. Combs and Weschler, two other managers, exited 48 stocks holding for just two years and ten months. And so even though we all know buy and hold, buy low, sell high, Psychologically, it's hard to do. So that's maybe what some hedge fund managers are getting compensated for. Sure. I might have read situational awareness, agreed with Leopold, but do I have what it takes to actually go and buy on margin and do all this crazy stuff and and like and not and not paper hands.

Speaker 3: I love that, discussion. I think there's a sort of feeling that he he was the one with the courage, to take all this risk or or something. Mhmm. But even that falls flat on a hedge fund manager, I'm afraid. Okay. Because it's like, you know, it's like someone someone goes all in with pocket sevens, and he he doubles money. Mhmm. If if you learned anything about his courage or his skill, really, he's it it's a bit like, it's a bit foolish, really. So it's basically like the I also used to actually think this. I was a young hedge fund manager. I feel like maybe aging out now, but I was a 28 year old hedge fund manager. And I did also have this perception that surely these big hedge funds are just so risk averse that it's almost like for psychological reasons they refuse to take enough risk. And I've come to learn that they're taking nearly the exact amount of risk that you should take because these things are mathematical. There's a right amount of risk to take if you have a certain sharp ratio. And those numbers are smaller than you think if you want to run money for a very long time. Yeah. Right? So, know, Leopold was up 400% in H1 of twenty twenty six. Well, Warren Buffett was up 5000000% by basically being sensible for a very long time. So do you want to do the 400 and then lose everything or do you want to do the 5,000,000 that it takes thirty years? So I think that's kind of like the orientation you have to think about. So it's not that some people are unwilling to take risk, it's that they are already taking that risk. It's just a small part of their portfolio. So there's no doubt that in the markets, numerized AI models that are trading thousands of stocks at all times that we intersected with situational awareness. And we bought some of their positions and sold some of their positions. Sure. And maybe had similar alpha if you constrain our portfolio to just look at that segment of the market. Sure. But we didn't do it at the wrong size, you could say. Yeah. To that.

Speaker 1: You actually ran the numbers on or you had Claude run the numbers on how likely a catastrophic down down drawdown would be based on a certain leverage ratio. What does that math say is the optimal amount of leverage, the optimal amount of risk? Is it all based on the timeline? If you think, okay, well, I want to be the next Warren Buffett. He's been investing for sixty years. I need to work backwards from that number. And then, is that different than if I say, look, I actually believe the singularity is going to happen in 2030, And so I only need to survive another four years, and then I'm done because it's whatever the singularity is.

Speaker 3: Yeah. Well, that's in some ways almost the problem. You really have to have a long horizon for investing. You know, the market is not pricing that the world ends in 2030. Yeah. In fact, the market's pricing, I mean, it's just not like on every level. Yeah. If the world was going to end, know, the volatility of the stock market would be as high as situational awareness is funded. Sure. So it's like you know we basically we have to have this orientation of long term and if you don't have that orientation kind of all bets are off. I mean if you're gambling and you're playing poker and it's late at night and you you only have thirty minutes left to play before you have to catch a flight Yeah. Okay. Let's up the risk. Yeah.

Speaker 1: Yeah. I mean, there there is an element. I I'm not saying that this is what's happening there, but there is an element where you're like, if I believe that the world is ending in four years, I wanna spend the next four years on a yacht. And so I need to get as much liquidity as possible in the short term, sort of like the the you know, you have you have what what what does that mean? You have, like, six months to escape from an underclass? Yeah. Sort of like the hedge fund version of that, I suppose. But you clearly don't, you know, believe in that. What are your beliefs about the future technology, the financial markets broadly? Like, how are you feeling? Is that even something that you interrogate? Or are you preoccupied with other sources of alpha that are less like macro and long term trend based?

Speaker 3: Well, any long orientation. So if you're long the stock market, the problem is all of our investors are already long the stock market. You don't have to tell them, you know, it's a good idea to buy stocks. And so, you know, if I was running a fund and maybe this personal trade of buying Amazon, if I had sold that to an investor, they'd like, well, bro, we already had Amazon and we already sized the risk appropriately. So all you did was add risk and not any reward. Really, it's that kind of tension that forces all these issues. But, you know, I also want to say, I mean, I think one of the first things I said in my post is I think there's a place for these kinds of funds. Okay, so I've invested in funds like this. So this was a fund. It was basically the a the the crypto version of this was a fund called Polychain. Yeah. And it's good. And I was a big LP in that. In fact, the first LP and the biggest, I think the biggest one for a while. But I guess I never got that enthusiastic that if someone had made a thousand percent of the past two years or whatever it was, that this would be some permanent property of that manager. You see what I mean? Like if there's no amount of looking back on a track record of a manager that's just buying risk will basically be able to overcome the realities of investing, which is that it's always going be hard to make money. So let me defend the situation more, not just that these funds are cool because I invested in one, but I think they're cool because many investors want a high return on the risk that they're taking because they already have their portfolio entangled in all these investments. So they have, you know, bonds and stocks and all of this stuff. And this marginal 500,000,000 that they're going to give to a small fund, they want that fund to kind of kick gas with that money. So that's where I would say, I don't think it's right to say Leopold did something wrong in a certain sense. It was very likely to go badly, but I don't think it's intellectually wrong for any LP to have invested in that. In fact, maybe I would have invested in it or something if I could have understood it a bit more and so on. But obviously there were a lot of red flags and the fact that the volatility was obviously much higher than a 100%. You can't make 4% in six months without having that kind of volatility. Then you're basically saying, we're literally doing a coin flip here. And I just hope that all the investors knew that. And by the way, if you're a huge investor and you put 500,000,000 in and you lost it all, on the very day you lost it all, you might still be up on your portfolio. Sure. Because your mother's stuff went up. Yep. And so that's how investors should be managing. It's on their whole portfolio. Yeah. So I want I don't wanna say anything bad about anyone who invested in that fund. I'm sure there are some of the best wealthiest investors out there, so I'm sure they're fine.

Speaker 1: And, I mean, from the letter that he sent today or yesterday, it seems like the the the Citadel deal put a lot of new cash on the balance sheet, cleared a lot of things out, and that there might not be permanent capital impairment as he put it in the letter. So there he is living to fight another day. What do you make of this idea that the optimal number of blowups for a hedge fund manager is one. Not zero. Not five. But do you agree? We

Speaker 3: had one we had one losing year so far with Vidurai. Right? So we are not are not, you know, if you're in the business of being an investment manager, you're almost putting yourself on the hook to be like, yeah, one time we lost money. Yeah. In fact, about 50% of days, we'll lose money. Right? So you you really don't wanna be, you know, negative about that.

Speaker 1: Mhmm. Yeah. Sorry. Continue.

Speaker 3: No. So I think that's that's that's one very important piece. I would say though the key thing that I think matters is not just leverage, which I think people are saying, oh, you use leverage, you've lost money. We use leverage. I'll bet you we use way more leverage than situational awareness. Interesting. We're levered long. We're not levered factors. So we're levered. If you were to plot our performance on a graph, you would see and you would say show me the unlevered performance. It would basically look like the x axis. You wouldn't even notice it. Sure. It's so much kind of risk constrained that it requires leverage to get to appropriate vol. Got it. So with with more leverage than situational awareness, we have one tenth of the volatility and risk is volatility squared. So if they have 150 squared units of risk and we have 10% squared, it is a whole different ballgame. So I was I'm mostly critical about the the the high volatility. I think that was the mistake, not the leverage.

Speaker 2: Mhmm. Interesting.

Speaker 1: Where do you think the fund goes from here? It feels like he's set up for a second act. The fund is not liquidated, not private only, not converted. How do you think this evolves? Like, what is the lesson to learn?

Speaker 3: I think think, first of all, I don't think that we should have a culture of thinking people are intelligent anymore because AI is smarter than all of us. Okay, so if one of the reasons is you think there's like a super genius investor out there that's clearly got something like a kind of same type of reason you might have invested in SPF. He's clearly on, you know, doing so well. Don't do that. Don't do that hero worship thing. Just take out your pocket calculator and calculate the risk of ruin. And then you'll probably avoid a lot of these types of mistakes. But I think, I really believe in AI and I'm a singularitarian. I mean, I'm seriously a believer. I'm not moderate about it. But because of that, guess what? AI is in the stock market. AI is in Citadel. AI is in Numerai. So if someone comes up and says, Oh, my thesis is we're going to buy AI stocks because AI is going to be big. Okay. The stocks are in the market. The market is an artificial intelligence. So don't take it lightly, man. Yeah. Like, yeah.

Speaker 1: What is the process psychologically for dealing with a career where 50% of your days are going to be bad days? There's a lot of people in most careers I feel like can show up and have a have a a day that's successful. Like, they move the needle forward. They didn't make backward progress. And yet with the job of a fund manager in your role, half the time you're gonna wrap up your day and you're gonna be worse off than the day before. Right?

Speaker 3: I think well, you see my, terminated hand in the back. Right? Yeah. I think you need to lean on math and AI. You need to basically say, you know, if the volatility, if you know what volatility means and you know what your volatility is, can there be anything to shake you if you if you can just rest on that mathematics? I mean, so when our fund had a down year, we lost a bit more over the whole year, a bit more than one standard deviation of volatility. Okay, but guess what? That happens to practically any investment. Yeah. You're guaranteed to lose one standard deviation, one unit. You're guaranteed to lose your volatility. So the question is, were we running a 100% volatility when that happened? No, we were running a survivable amount. So we didn't lose everything. But yeah, it's, so I think there's some, there's really a lot of comfort to be had in math and AI and just kind of trusting that things will work out. All you need to do to be Warren Buffett is just get 9% alpha per year and add that to the S and P, right? So if you have a hedge fund that's making 9% alpha per year and they can repeat that they're going to make 5000000% return and all the people taking too much risk are going to lose it all over and over and over again. And so, you know, that's how you how you have to think about it. Last question and we'll let you go.

Speaker 1: Retail froth. There's an interesting arc here and I wanna know if you think it's it's it's real or the shape of it, but it feels like when when the situational awareness PDF drops, no one's thinking about bottleneck stocks or or memory. The like, the smart money gets in. I would say everyone in Silicon Valley was already thinking about that. Yeah. Yeah. Yeah. Yeah. But that's like maybe the smart money and then it gets bigger and bigger. And and then pretty soon, it's like retail meme stock. And it feels like one of the knock on effects of that is just more volatility, like the last little leg up can be more tumultuous. Is that the way to think about that? Is is it important to to think about, like, the meme stockification of a particular asset class as a factor and then how to deal with it?

Speaker 3: Yeah. Know, our fund was operating with high leverage during twenty twenty one's, like, meme stock rally. Sure. And it is a a phenomenon. And Cliff Asness, a famous quants in QR, talks about how the Internet sort of making the market less efficient because there's so many people who almost get the same message at the same time with the same media and then have the same trading account systems like on Robinhood or whatever that can instantly express that. But I think, you know, I don't think that is something to lose sleep over. I mean, it's very easy to handle that type of thing. The the what happened to Leopold wasn't something extraordinary. It was banal. It was you had higher volatility and you had a standard deviation drawdown.

Speaker 1: It's gonna make for a bad movie then. Well Well, it'll be a great movie, I'm sure. This movie won't have any math in it, but it's probably math It should. It should. It should. Well, thank you so much for taking the time to come chat with us. Have a great rest of your week, great weekend, and hope to talk to you soon. Thanks so much. You. We'll talk to you soon. Goodbye. Let me tell you about CrowdStrike.