Why Extreme Risk Takers Are Winning — With Nate Silver

27:50And and although I'm kind of more in the river mindset, I have experience in both. and part of what the book's trying to do, like, you know, my pitch to people in the village is like, I can help you to understand how these people on the river think. Because I'm one of them and I talk to them and I think I have very honest conversations with them. But they can be trying to take advantage of you or can be trying to run circles around you. And they are very capable of lots and lots of lapses and errors and judgment, as well as just things that are kind of self-serving the way that different technologies are regulated, for example.

28:25and we will get into that in a moment also for listeners towards the end of this conversation we're going to ask Nate for his like three key poker tips I'm excited for that I recently joined like a once a month weekend poker game so this will help me but we're going to hold off on that for a moment because we do want to talk a little bit about sports betting and then get into the VC stuff the sports betting section of your book is fascinating the first part that I found so interesting was, or I would just say the most interesting part to me was the fact that the casinos don't really have like such a sophisticated operation setting lines and they're waiting for professional bettors who've like their own formulas or crunching numbers to make their bets.

29:05And then they sort of adjust the line based off of what these pro bettors are. So effectively, if you get in to a bet pretty early on, like basically, or immediately once it goes up, you might have better odds than if you wait for for sure no look you might have some like smart kid um who's 25 years old who uh you know whatever phd in economics or math or not even a phd probably maybe but more likely just a ba in math and is getting paid um a hundred thousand dollars a year to go work for a casino somewhere and you know the college basketball lines must go up and he He used like a computer program to estimate Duke versus North Carolina or whatever else.

29:49And like maybe he's talking to his colleague on the other side of the desk. But no, I mean, they are not that sophisticated in terms of setting lines. What they're good at is understanding how to account for information and the actual act of what is called bookmaking. Right. Maybe, Alex, they think that you're smart. So the line starts out with the Kansas City Chiefs favored by two points. when you bet on the Chiefs, they'll move the line to Chiefs minus three. Whereas for me, they think I have the profile of a recreational better. I bet on the Chiefs every week. So when I make that bet, they're not going to change one thing.

30:25They'll want to keep the line there to encourage me to bet more potentially. So that act of gaining information is a traditional process of bookmaking. And it's a market discovery, almost like an auction type format, a price discovery format. In your book, you talk a lot about sports betting, but the one thing that you sort of stay away from, or maybe it's not applicable, but I want to get your opinion on is the fact that this is potentially dangerous for the sports. I mean, I think probably the Shohei Otani thing happened after your deadline, but we've already had Shohei potentially implicated in a sports betting scandal, or definitely implicated.

30:59We don't know if he was involved, but who knows? We may never find out. We've had an NBA player suspended. I'm sure there's been, And, you know, a lot of these platforms won't let college games be bet on. Do you think that sports are getting a little bit too close to the gambling platforms? Because, you know, we're going to just end up seeing scandal after scandal of people being unable to resist. Like gambling can be in a compulsion. And especially if you have a way of trying to swing the outcome, it can really change the balance of what might be one of the last pure things. Yeah, no, look. In the United States or in the world, which is like sports played by the rules.

31:34I think the leagues are getting a little bit greedy in some ways. So, for example, in the NBA, Jonte Porter is a player who was suspended for betting on his own prop bets, right? He would bet that he wouldn't get as many rebounds as the over-under line had and mysteriously would find ways to miss rebounds and have injuries that required him to be taken out of the game. So why is anyone betting hundreds of thousands of dollars on this obscure player? Like, that doesn't make sense. So having like, I call it the kind of in and out menu where you only have like six items, but they're all done pretty well.

32:10Like that's more rational where if you want to bet on, you know, big time sports, the point spread, the money line, which is just the odds that each team will win or the over underline. Like that's probably fine when you have things like player prop bets, when you have things that are based on news, there are many casinos that will let you bet on like who is taken first in the NFL draft. If you work for ESPN and you're an NFL reporter or you work for an NFL team, you're going to have a lot of inside information about that. So I think it's been a little bit of a money grab in the short term, and there's not enough focus on making it more sustainable.

32:51I mean, look, I live right in New York City. I go to Madison Square Garden a lot. At one point, there were literally three different ads for three different sportsbooks at New York Rangers games on the rink around the ice, the sideboards. Usually, people care about category exclusivity in such a narrow venue, but they so didn't care that they were... I mean, I think it was overbought, and you've seen some consolidation where actually DraftKings and FanDuel are doing reasonably well. They have a bit more market share relative to the others. but it's always been a tough business. And the reason it's a tough business is because like, it's not that hard.

33:29I mean, it's hard. Don't get me wrong. It's very hard to beat the sports books, but when they offer this gigantic menu of hundreds of different bets on every single game in thousands of leagues around the world. And by the way, and there are nine legal sports books in New York and however many gray market offshore ones. So I can look for the best line out of thousands of options. it's not that hard to find a line that I can beat by 4 % on average, which is what you need to beat the house. You have to beat it by, you know, they take about a 4%, 4.5 % cut of every bet. So I think it's always been, in Vegas, sports betting has always been seen as an amenity, A-M-E-N-I-T-Y, meaning like you come in, there are certain types of customers that want to go in for the NCAA tournament or for an NFL weekend or, you know, big soccer match and bet a few hundred bucks and gamble and watch on the big screen.

34:23And like, and sports books are fun, kinetic environments. If you go to like, I was there for Copa America. I was just in Las Vegas. And like, it's fun to see kind of the Colombian fans and the Argentinian fans and all the people who have action on the game. Like that's a fun piece of like eye candy for the casino. But sports betting has never been a major profit center. It's like one or 2 % of overall Vegas revenues. And I think it's been like a little overbought in a way that isn't fully sustainable. Okay, one last thing on sports betting. I don't know if you guys started this at 538 when you were there, but this was something that's always kind of annoyed me about the analytical approach to sports.

35:00Is that like sometimes the sites will be like this team has a 70 % chance of winning or this player has like a 30 % chance of hitting a home run. Like you watch baseball on Apple TV right now and it's ridiculous. Like they adjust the win probability with every, every percentage. And for me as like a massive sports fan, I've always thought this is the, the word chance is so wrong here. Cause it's not a chance. Like it's not like you're like putting a ball in the roulette wheel and allowing it to come where it is. It's actually something that's determined by human outcome, human intervention here.

35:35And I just think chance is the wrong word for this. What's your response to that? We can get into some existential questions here. I think it like literally depends on the sport. to some degree. In a sport like hockey, for example, which there's a lot of luck in hockey, generally speaking, teams don't have that large an edge. Like literally, if the Zamboni made a certain patch of the ice slick in between periods, and maybe the puck bounces and goes in the goal, and like otherwise, it doesn't certainly. But look, that's actually a very deep, Alex, existential debate about the kind of nature of uncertainty.

36:10Look, epistemologically, you don't know, right? It might be true that it's not literally a matter of chance, although some things actually are kind of random. Like, this is part of what chaos theory said, is that some systems are irreducibly complex, like weather systems for example. Now, I don't know if a baseball game falls in that category, but there is probably some element of literal chaos, I suppose. Okay. I feel like I could do a full show on this, but I'll accept the argument. All right. One more thing about on this line, politics. So polymarket, you know, watching it predict like who's going to be the nominee, who's going to win the election.

36:56You can bet on anything there. It's a betting site on sort of news outcomes. I mean, I think you're on the board as well. I'm an advisor to polymarket. That's correct. Yeah. An advisor. Are they getting to are these prediction markets getting to be better than actual polling in terms of predicting an outcome? So I have my own polling-based model at my newsletter, so I'm a little bit partial to it. However, one thing about polling is that it takes a few days to incorporate new information. You actually have to contact people on the phone or online and have them take your poll, and you go back and do your weights and run your numbers, whereas they get to react to things in real time.

37:32So, for example, in the disastrous debate for Joe Biden, was it June 26th or 27th or something like that, You could see reflected in real time prices at Polymarket and other markets changing on the chance that not just Trump would win, but that Biden would drop out. You could see Bitcoin reacting in real time because Trump is seen as bullish for Bitcoin. So, you know, these were not people who were waiting for information to come in. There are people who were making their best guesses in real time. And that skill is underrated, too. Look, I'm a guy who loves to be rigorous. The book is very rigorous.

38:09It involved three years of reporting and writing. The model we build is thousands of lines of code. But the skill that people in the river have to make decisions based on incomplete information is often vital. Because if you wait for complete information, then someone else has already made that bet first, right? There's no more profit in that bet anymore. You have to be willing to make judgments when the data is noisy. and that's kind of what makes the actual skin in the game risk-taking communities actually have the potential to make profit and makes like academia or the village then too too slow most of the time okay that's interesting so you're a river guy but you're also a polling guy yeah and the answer that i'm hearing you give right now is maybe the river will eventually have an edge on polling i think this is in politics or am i reading this wrong this is changing too in part because um you know the markets are becoming more like the politics marks are becoming more like uh more like sports betting markets which again are are are at least pretty smart once you have price discovery um i think the reason why is that everything is kind of becoming eaten by politics now and so you have more and more financial firms hedge funds investment banks so forth who think they have to like understand political risk because political risk can affect things like different classes of stocks, how different industries are regulated, interest rates, whether foreign countries, if you're investing in some foreign exchange, is this country going to be stable or not?

39:44Will the leadership be good for the economy, good for equities? So you have more and more investment, I think, and this is somewhat firsthand knowledge based on going around and talking to people about consulting opportunities and so forth. You know, I have noticed more sophisticated conversations happening in the last few years. Some of these companies will trade in different ways, but like prediction markets are a part of that. I mean, the critique was always that when you just had kind of hobbyists, that it didn't get that far because you only have like one election every four years. Building an election model is hard.

40:17People are very partisan. But you have kind of more institutional investment now in building models of things involving political risk or other types of exotic derivatives where there's actually adequate liquidity and actually adequate opportunities to trade, which I didn't think was as true, say, five years ago, 10 years ago. that's crazy nate because i mean we know that money has uh influence on politics are we going to end up getting to a place where like i mean i guess this was always the case but like companies or financial institutions that may end up placing big bets on candidates in one arena just work hard behind the scenes to manifest that uh that outcome that they're trying that they're financially linked i mean there are a couple things here one is that there have been accusations before in 2012 for example there were accusations that um mitt romney traders were deliberately buying up romney stock at prediction markets because they thought that would create favorable buzz and therefore that's like free earned media for example but remember that's at a point 12 years ago where there's less volume so you have to spend less to influence or manipulate if you want to use the term that market yeah look you do get into some gray areas here where if you You have like, you know, traditionally organizations like the CFTC, which regulates commodities exchanges, have not loved betting on real world events in part because they were worried that it could create moral hazard, for example.

41:53You know, if you're betting on some insider trading case or something like that, then if you have access to like manipulate the markets. However, the fact is that, like, you know, if I were talking to CFTC, this risk is priced into lots of commodities and so forth anyway. That when when the when the Donald Trump price increases, it's probably good for for maybe defense contractors or Bitcoin or or whatever else. I don't want to speculate too much. That's not the part of it that I do. But like, yeah, but like, you know, political risk has huge effects around the world. If there's some, you know, dictatorship that emerges in what had previously been a capitalist country, then that can have profound effects on the global economy.

42:36Things like, like, for example, if you're investing in NVIDIA or semiconductor supply chains and understanding the China-Taiwan conflict is going to be like very important potentially. And so these things are things people have to price anyway. And the argument for prediction markets is that is that they give you a more efficient and explicit way to do it. Okay, let's talk about VC and founders. Yeah. We're like 40 something minutes in. We have to do this. It's the show mandate. But VCs and founders, they think about risk in really different ways, which is another thing that you highlight in the book that to me was surprising.

43:10So can you talk about like the different type of risk that a VC and a founder tries to take on and how does that influence what they do? Yeah, I mean, the main difference is that most VCs want founders who are going all in on an idea with a very long time horizon. So for 10 years or often more. I mean, these companies can have, I mean, the J curve means you can often take 10, 12, 15 years before your fund is actually making money, even if you've made good investments. Right. I love how you call these founders hedgehogs. Hedgehogs. This comes from. Basically, get in that hole and keep burrowing.

43:42So this comes from the typology from Isaiah Berlin, where a hedgehog versus a fox, where a fox is crafty and kind of roots around for different things. And it's kind of like a hunter-gatherer type, right? Whereas a hedgehog very much like digs in and stays berries in the hole and is very much all in. So for a founder, I mean, being a founder is, I think, legitimately often a very risky thing to do. it's pretty rare for a company to become the next google or spacex or whatever else um by definition you're pursuing an idea that probably is not adequately addressed in the market usually the market's smart it's for a good reason that your idea wouldn't wouldn't work potentially um and so and so you have all these uh vcs who are foxes right they'll say yeah actually we don't want to have too much domain knowledge you talk to like michael moritz right he used to be a journalist and then became one of the best investors on on earth um right from sequoia um he's like yeah as a journalist you kind of parachute into a place and you have to again use incomplete information to make decisions quickly um and you don't necessarily want to be a domain expert sometimes domain experts get blinded or biased by the forest for the trees a little bit, you want to be like a good generalist.

45:06Whereas for a founder, you want them to solve a problem in a different way and devote all their resources to it and be a true believer in themselves, right? You know, Paul Graham, another VC I talked to says, you need founders to be optimistic. It's kind of irrational in some ways. You would think you want a founder who's a perfect assessor of risk and very well calibrated. But if you're not like a true believer and you're idea then it's very hard to sell other people on it um and kind of in a world where things are getting more and more quantified and analytically driven um maybe you need maybe there's more of an edge now from being a a true believer and saying i don't care what the barriers are i don't care what the conventional wisdom is i'm just going to kind of plow through things now to be clear this often ends in disappointment or ruin.

46:00And if you're a VC, then you can collect 15 hedgehogs into a fund. And then that turns out to be a very nice thing. I mean, somewhat contrary to the conventional wisdom. Before I started this, before I knew anything about VC, I thought, well, actually, it's like a poker tournament where you lose money most of the time. Like, that's not true. If you talk to like Mark Andreessen and Andreessen Horowitz or friends like that, or Sequoia, they'll say, actually, we make money from maybe 60 % of our investments return something and then 20 % return a lot. And that's, it's actually a pretty nice business if you are in a, in a, in a top tier firm.

46:36Um, but, but that's from aggregating risk. And, and that's, again, that's where you realize expected value is if you can literally kind of run the experiment a hundred times, you can be almost guaranteed to make money. We run simulations in the book where it's like, if you really are, returning 20 % ROI per year on average with a lot of variance, you're almost guaranteed to make money. And there's enough empirical evidence that like, oh, like a certain A16Z fund will have a down cycle and only make 11 % per year for five years. It's still better than you can normally do in the S &P 500 or things like that.

47:15And so it's a really good business to be able to find these dogmatic founders who are taking huge risks and then have a piece of all of them and hedge that risk. Right. And you talk in the book about how a Silicon Valley marries risk tolerant venture capitalists with risk ignorant founders. Basically, founders go all in and it's win or lose. I mean, they'll make a salary probably, but basically they could do all this work, destroy themselves physically for nothing. And whereas a VC, that's okay as long as they pick enough. The VC mindset is such that when you had FTX blow up, I guess it was Sequoia said, yeah, you want high variance bets and so we would make the same bet again.

48:04They lost a lot of their, I would say, Sheen, Sequoia did with the FTX bet. But also I think that you made a good point which is that one of the things that makes Silicon Valley unique, and it goes back to this uh quality of being willing to be embarrassed is they're willing to be embarrassed whereas like it's not necessarily like in the more villagey status you know safety type of uh occupations and fields like that actually matters a lot but for silicon valley it's just another day yeah and look you see lots of behaviors in silicon valley some of which are maybe asocial or antisocial behaviors but everything from like tolerance for psychedelic drug use to tolerance, which may not be a good thing for like difficult kind of, you know, asshole founders.

48:49And look, I think within itself, like Silicon Valley, it's kind of conformist relative to itself in certain ways, but it really doesn't care about what the village thinks or what the rest of society thinks, or at least it shouldn't. I mean, I think you have like, you know, obviously certain VCs and founders are more politically sensitive. But, you know, you talk to Mark Andreessen, he thinks the rest of the world's becoming more conformist. And that's great for me, because it means the returns to being like, different are, are higher now. If there's more of a penalty, a social penalty for, for nonconformity, then I don't care about the social penalty.

49:30And I'll, I'll make money in investing cool things and kind of make my own universe out of this. But, but isn't that a little bit, isn't that bullshit? Like a little bit from, from Andreessen, because, you know, as I'm reading the VC chapter, and I think you even put this in, like, I started having these questions like, all right, so VCs, if you're in a position like Andreessen Horowitz, you're probably guaranteed at least a 15 % return as you spread your bets. Okay, so that's safety. VCs, they're herd followers, right? So like, they are pack animals in a way that like, you would typically associate with the village.

50:01And then you ask like, all right, well, are VCs disruptors? Or are they incumbents? And in some way, you know, they are the incumbents making money off of the disruptors. So I'm curious what you think about that. Like Andreessen can talk a big game about time to build and change the world. But with this whole like raise your hand up and fight type of mentality, like aren't they the man just as much as like Wall Street institutions are or the traditional village institutions are? For sure. Or the comparison I'm sure they won't like that I make in the book is that it's a lot like Harvard or Stanford.

50:32because a lot of these guys have become anti-big universities. But like if you're Harvard and you're guaranteed to be able to recruit, you know, an 85 % yield rate from the best students around the world, it's really hard to fuck that up, right? If you're a startup, then it's such a positive signal to have Andreessen Horowitz invest in you that you'd kind of have to be crazy to turn them down. And so, I mean, Andreessen even told me that it's a self-fulfilling prophecy at some point. I respect the fact that Silicon Valley understands the importance of making high upside bets and understands the importance of having a long time horizon, being willing to project out 10 or 15 or more years.

51:14I do think sometimes that some of them are spoiled winners. I say that slightly affectionately, but they have a lot of things going for them. They control a greater and greater share of the world's wealth. And I understand as someone who himself often gets annoyed by the village, I understand the impetus to be annoyed by it. But going back to being a little bit more apolitical, I think might be a better look. And maybe in the long run, because there is some risk of regulation. I mean, certainly when it comes to things like AI in particular, then government's going to have an opinion about that.

51:56in crypto, government's going to have a lot of say over how crypto works and what the upside potential of those businesses are and those platforms are. So I think they are a little bit asking for fights. And again, I have this tendency too. I'm competitive. I like to pick fights, but it's not always leading to decisions that I think were good decisions at the end of the day, when you kind of wake up and look back on things later on. Yeah, and look, I mean, I appreciate VC. They find a lot of interesting things. We talk about them on the show. They're fun to speak with. But with some of the top-end ones, there does feel like there's some of the same sanctimony that you would find like on a Princeton campus.

52:39And that kind of annoys me. Yeah, it's like the horseshoe theory of like, you objectively have it really good. If you're like a tech billionaire, you get to spend all day meeting interesting people and investing in cool businesses and throwing cool parties anywhere in the world. I mean, if you have infinite money, then that's a pretty nice life. And I think I understand where the competitiveness comes from, but we are seeing more signs of a tech backlash. And so far, they're still making lots of money. But I do think AI in particular is a sector where there is a pretty big appetite for regulation.

53:19is actually one of the few things that both Democrats and Republicans agree upon is they actually are concerned about AI. Not as concerned as like the effective altruists or whatnot, but the notion that like, hey, it's going to take over your job. It's going to lead to discrimination. It's going to lead to whatever else. You're going to cede your agency to the machines. Like that is intuitively scary for most people. And I think they have to be careful about, I mean, you know, sports betting is a much lower stakes industry, But like, you know, VCs should have a long term outlook. And that includes the kind of political environment in which they're operating, because you could have a shift toward more regulation for sure.

54:04Totally. OK, so we'll take a break now and we're going to come back. I think we'll try to talk about a handful of the following. Sam Bakeman-Fried, effective altruism, AI, existential risk, nuclear stuff, and Nate's tips for poker. So I'm going to try to get to at least three. We'll be back right after this. And we're back here on Big Technology Podcast with Nate Silver. He is the author of On the Edge, The Art of Risking Everything. You can see it if you're watching on the video here. I've plowed through it. Really great book. Let's talk a little bit about effective altruism. And Sam Bakeman-Fried.

54:38in this for, we'll start with SBF, then we'll zoom out a little bit. I guess like effective altruism takes this like, uh, expected value type of thing to the extreme. And with like SBF, there's a moment in the book you meet with him as like, uh, the whole FTX thing is going down and he's such a compelling character. It's like the third book I've read that has like a lot of SPF. And I feel like my appetite for reading about him is just like endless, but there's an interesting moment where like basically sam says if he can flip a coin i'm gonna get details of this wrong but flip a coin three times and two times the world is twice as good as what it is and one time the world is obliterated he would do it every time or even less than that right if you if you win 51 out of 100 flips and you make the world twice as well off then then he literally would take that coin flip and i think he's actually not even being tongue-in-cheek about that.

55:32I mean, he told Carolyn Ellison that in court testimony that I obtained. He said similar things to like Tyler Cowen, The Economist. Yeah, it's philosophy known as utilitarianism, taken to a quite logical extreme. And I think this speaks to, look, I think effective altruists tend to be, A, I think SBF is not representative of the whole group, certainly. I think they are very smart. They ask questions that other people don't ask, which I always think there's value in doing. But I do think that... So explain what they are from your perspective. I mean, sorry, I know this is huge. Try to do it in like a minute because everybody has so many different definitions of them, but you actually went through the reporting process of like meeting and learning about EA and then you can go back to like the weakness.

56:19So effective altruism is a brand is the way I call it created by Will McCaskill and Toby Ord, who are professors of philosophy, I think, at Oxford. and it kind of is what it sounds like. So the original idea was how do you more effectively do altruistic things? For example, give more money to charity, right? If I give money to like the American Red Cross or something, you can calculate that actually X percent of it is wasted on administrative costs and then the way they distribute it is for things that may or may not be good, right? And so the original idea was that if you give money to things, for example, like mosquito nets in Africa, were found to prevent loss of life and prevent severe disease at a pretty low cost.

57:06You can save a life for$5 ,000 or something on average. So it began, I mean, I'm deliberately simplifying the story here because the origin story gets a little bit convoluted, but it's kind of like this short version of it. It began as a way to do more effective philanthropy. But then people began to apply it to all sorts of other things like animal welfare. For example, you know, if I decide to eat a chicken or some chicken nuggets, right? What is the utility for the chicken's life versus my pleasure and my nutritional value and like eating those chicken nuggets? So I know some EAs, for example, who won't eat chicken, but will eat beef because you can create a lot more portions out of a giant cow than you can out of a chicken.

57:57and cows are also kind of dumb and therefore they are valued less in some EA systems. So it gets very weird very fast. And then kind of where the interest in things like existential risk and AI come from is that if you think that there's a chance that something could destroy civilization, like unaligned AI or like nuclear weapons or like bioweapons, for example, where climate change, although climate change is not traditionally an area of EA focus, we can talk about that, that has a very high expected value gain or loss based on what you do. So yeah, it's kind of taking this gambler's expected value maximization mindset and applying it to all sorts of real world problems that it's not traditionally applied to.

58:45And I'd say with very mixed success. Right. And the point is that this is sort of making its way into the business world. and if expected value is sort of like you're a North Star and you think about like, do I want to risk the company on this behavior that may have like a two-in-one chance of working, but a one-in-three chance of doom, or sorry, two-of-three chance of working, one-in-three chance of doom, most founders would never do that. But like SBF, for instance, seems to be open to making this type of decision. Yeah, I mean, literally he told me that if you're not willing to risk completely ruining your life, that you're a wimp was a term he used and not undertaking enough risk.

59:31And believe me, that is atypical. But, you know, you talk to, like, Vitalik Buterin, who founded Ethereum, and he's like, yeah, I thought there was only a 10 % chance that it would work out like this, right? But if you make 100x your money in that 10 % of the time, then the expected value is positive. I mean, my issue with someone like Sam is more that I think he's miscalculating a lot of things. If you talk to people who were in his orbit, then, you know, he was a person who might make a good first estimate, but then not refine it and become very hedgehog like and get very dug in and also not thinking about the harm that would occur to to other people.

1:00:07But, you know, I'd be the first to tell you that, like, you know, I like to work on problems that are kind of closed ended problems. So we were talking before about sports and how is it really random or whatnot. You can get into some philosophical discussions there about sports or elections. What's really random? What's not? But the fact is that you can represent these things reasonably well with the model. Sometimes they're little fudge factors that you might round up around down the model based on things. It's a little bit hard to quantify, but your experience tells you are important. But these are kind of closed world problems that you can have repeated trials of.

1:00:43I worry that sometimes the EAs are like operating in a world where they assume everyone has like 200 IQs instead of us regular stupid human beings and often have like a lot of short sighted realizations. Because when I look at like work even in sports or something like that, there are lots of bad election models out there and bad sports models. And that's with relatively good data and closed systems. So, you know, you can very easily make very bad models if you're not careful. Exactly. And that to me is like the huge hole in expected value that, you know, we've been talking about through this whole conversation.

1:01:23and I think you explicitly like talk about this in the book that like basically expected value can be good for things that are quantifiable but the problem is that people will take it and say like you know this has an expected value of this but they're fudging it they're really I think you called it improv and to me it sounds a lot like a religion so I just what I I maybe maybe this is from the book maybe I just wrote this as a note but there are strict values up until there's something that has self-interest. And then you can squeeze it in and sort of fit your expected value calculation. And this is some examples you gave that effective altruists, they say that future lives are important, but they also are okay with abortion, which is an ideological inconsistency, not really, I'm not taking a position on it.

1:02:10They're okay with not having kids, although they value future lives. They defend immigration restrictions because they say that like if we're anti-immigration then like it means that there's a chance that like more pro-immigration politicians will come in and allow more immigration like it reminds me exactly of religion where like you have this tenets and you have these self-beliefs and you make the tenets work to advance your self-beliefs it feels like the same exact thing and I know religion is sort of like a pejorative in these type of worlds but it's that's what it sounds like and that's a huge huge liability when it comes to expected value.

1:02:46There are some similarities, right? Just like in some churches, you tithe or tithe, I'm not sure how to say that word, but give 10 % to the church. Some EAs believe in earning to give, where you go work for a hedge fund or something, and then you give 10 % of your income away for that. Look, one of the ironies in the book is I think people who are operating in the world of like vc or poker in some ways the profit motive grounds you a little bit um and kind of gives you maybe more of a bs detector and that whereas if you're a true believer like the people who worked for spf i think were actually very nice conscientious people um but that may have made them less skeptical than they should have been of like of like sam himself um so and don't get me wrong like i think a lot of eas have done a lot of smart things with respect to like poverty reduction and things like that i think overall having more focus on existential risk is a big win for for eas i mean there's been a paradigm shift in in the awareness of the conversation about ai and and what risks and opportunities that it might pose potentially so look i think they've done good for the world on that but i think it's a small community um it is fairly demographically homogenous right it's you know a lot of kind of 30 something mostly mostly white people a lot of men very concentrated in some areas like oxford or silicon valley or things like that maybe who don't have that much necessarily life experience so i think it's a useful movement but like yeah there are things that are i mean i think religion is a little bit too disparaging of it um but they proselytize right they they they want to spread the word um they tend to be pretty good at at pr um when they have when there's an ea book coming out it has like big like book release parties and and things like that and yeah there i mean i think there are there are definitely some parallels.

1:04:52Yeah. I mean, like, look, my first introduction really to AEI, I was learning about it through SBF. I'm more impressed with it now than I was previously, but there are still like some flaws. So, but it is very interesting movement and we'll be doing more about it on the show. Let me ask you this then. And this is sort of what it all sort of builds up to, which is that EA is extremely influential in the AI field. Yes. And just not even not even just da just this thinking this similar like thinking river style thinking thinking and expected value and then this sort of hypothetical that you had with sbf of like flip a coin 50 100 times 51 times the world is twice as good uh 49 it's it's destroyed um and a lot of people who think in this expected value mentality will be like all right i'll flip that every time maybe not a lot of people maybe with sam is the outlier but then you think about ai and it's like oh is ai that exact calculation where ai if it works out well will make the world twice as good and if it works out poorly it might destroy the world and do you want people with this type of thinking to be running ai projects yeah so sam altman who believe me i think is not in the same category as spf i talked to different sources like he got i mean he also got ousted by ea align folks So, um, but, but Sam Altman, and I talked to him, I think in August, 2022 for the first time, that was the main interview, which was a good time because it was like when he was a little bit less guarded right before GPT.

1:06:18So they kind of knew what was coming with 3.5 and four, and it wasn't quite public yet. So he was, I think in an optimistic and unguarded mood. And he's like, yeah, I think that AI could end global poverty. Like he meant that very literally, like it could end global poverty. We'd have so much economic growth and plenty and automation of tasks that used to be burdensome for human beings that you could end poverty, he thought. But he also thinks that it could also destroy the world. I mean, I think the chance that he would put is not as high as some people. But he's like literally willing to make not the 50-50 bet that Sam made, but maybe a 90-10 bet.

1:06:55Maybe a bet where there's a 10 % chance of really bad things happening because he's convinced there's a 90 % chance of really good things happening. I guess that kind of gets more into expected value terms. But then you get into questions of like, what right does he have to make that bet for all of humanity? And there are parallels. I mean, when the Manhattan Project was experimenting with atomic weapons, there was one experiment they thought had some chance of like actually igniting the Earth's atmosphere on fire. It was very unlikely, but they figured it might be a one in 1000 chance of like literally incinerating the earth.

1:07:36Right. And like, you know, who has the right to take that sort of chance? At least the Manhattan Project is kind of at least nominally an offshoot of the US federal government, which is an elected company. um but but yeah and look when you have people who are kind of taking risks on behalf of everyone else then they often wind up being villainized by society i mean that's an area where again the potential for like i think political backlash is is fairly high and um and you know it's also like if you have these gpus that have lots and lots and lots of compute i mean one good thing about AI risk is that we, there are a finite number of entities that we know have enough compute for now to, to build top, uh, top tier models, large language models and other models that might not be true as things get faster and faster in five years and 10 years, which could potentially be more dangerous.

1:08:34But, um, but I think they, I mean, for one, in one way, it's reason to be optimistic is that like, you know, something doesn't quite add up to me where you have these like smart, rich like nerds in San Francisco and Silicon Valley making bets for all the future of humanity. Like something doesn't quite add up there. I mean, I would think humanity might have a tendency to want to have something to say about that maybe. Right. And the people that are closest to this are the ones that are most concerned, right? It's like you met people who are so afraid of, I think, AI risk that they won't save for retirement.

1:09:11You made people whose P-dooms were like in the 30 percent and they felt they were considered moderates in their community. And yours is between 2 and 20 percent, which is interesting. But it struck me that you took this fear of existential AI risk pretty seriously. Look, the fact is that the relevant experts in the field are pretty unanimous. this is at least a risk that we want to monitor, right? So like you just said, Alex, the fact that the people who are closer to it are more worried is worrisome. Also the fact that we don't fully understand the technology. I mean, people call it like a giant bag of numbers.

1:09:54How large language models do what they do is a little bit of a mystery and considered like a little bit of a miracle. I mean, people now kind of take for granted um because maybe we're at like a a little bit of a plateau kind of post 3.5 four we'll see where we are in a year or so but like you have this machine that's basically like a can almost pass the turing test right you can like ask it a question about anything and some of the answers are stupid but some are quite ingenious and brilliant it can draw pictures i mean there's this like thermostatic increase where you just get used to this for now but like i you know few experts in the field would have expected anything like this to occur 10 years ago right um you know the reason why open ai was originally a a a non-profit is because it was not clear it was enough of a long shot bet that like it wasn't clear if it would work oh you just kind of throw a bunch of like compute through this transformer architecture and maybe it'll keep scaling up and produce a miracle where you have like a machine that has comparable linguistic intelligence to human beings and all of a sudden And it does, or at least it's pretty close.

1:11:03I mean, that's kind of a miracle. And the fact that it's not fully legible, what it's doing exactly, I think ought to be frightening. All right, Nate. Before we leave, I'd love to get your three tips for poker players. What are the three things that anybody who wants to be good at poker needs to know? I think the most important basic tip for poker is be aggressive. if you're new in activity there's a tendency to to take the middle ground so if you if you play like a backyard game with people who have never played poker before they tend to call too much when they should be be raising or folding that mechanic of like take more decisive action is a good decision making framework across any type of risk taking i think um you know number two is is being cool under pressure and partly that involves not trying to be a hero right um executing basic strategy um and kind of you know because when you are operating in an environment where all of a sudden you're playing a ten thousand dollar pot in the world terms of poker or something you're actually operating on different physical and brain chemistry um like literally You have a stress response that kicks in, an adrenal response, and you're kind of in a flow state or a zone state, which some people will panic and go on tilt.

1:12:28And you have to be able to avoid that. That's something that can be learned a little bit with experience. And number three is being really observant. one of the things I learned just playing a ton of live poker over the past three years is that things like physical tells, getting a read on someone, having an intuition for what someone's doing through their just general demeanor, like that's a real thing. I used to think this was like exaggerated kind of BS, but like, it's a real thing. I mean, human beings making decisions under stress are not very good at concealing information. So whether it's like a micro expression or something like that or or a betting pattern or a speech pattern or or or involuntary you know looking at someone's like uh neck muscles for example or the way they're hunched over what they're doing if they look away that can be a sign of strength and so just being attentive and the importance of attentiveness is is a big lesson i think for poker players and why don't i throw out a fourth one that i learned from your book which is that you got a bluff and you have to bluff yeah Yeah, poker is a game, the whole premise of the game is based on bluffing.

1:13:33There are lots of fun card games that don't involve bluffing per se, although they may involve deception. But the only reason in poker, in theory, to ever pay off someone when they have a good hand is because they are capable of bluffing. And even until you get to the very top 100 best poker players in the world, believe it or not, most poker players bluff too little and not too much. Yeah, I'm always like, I want to hold on to those chips, but I'm learning after reading the book. Don't do that. No, you have to. You have to take risks to win a poker tournament. It's unavoidable. And so that's part of why it's an attractive metaphor for life, et cetera.

1:14:12The book is On the Edge, The Art of Risking Everything by Nate Silver, who's with us here today. It's out this week. I encourage you to go grab it. Definitely one of the more interesting books I've read in a long time. Nate Silver, thanks so much for joining us. Of course. Thank you, Alex. All right, everybody. Thanks so much for listening and we'll see you next time on Big Technology Podcast.