On the Edge: The Art of Risking Everything

by Nate Silver

In a world forged not by the toil of human hands but by the computations of machines, those of us who understand the algorithms hold the trump cards.

But risk-taking is an understudied personality trait, and the academic literature is divided over the extent some people are generally more risk-taking as opposed to taking risks within specific domains.

Sometimes the entire trajectory of history can turn on nearly random events, like the poorly designed “butterfly ballot” in Palm Beach County, Florida, that caused some Floridians to mistakenly vote for Pat Buchanan and probably cost Al Gore the 2000 presidential election. Play out thousands of poker hands, watch hundreds of sporting events when you have money on the line, or invest in dozens of startups, and you’ll quickly learn that between the vagaries of fate and our uncertain state of knowledge about the world, getting things even somewhat right is hard enough. Probabilities are usually the best you can do.

Politicians and political parties, by contrast—especially in a highly polarized two-party system like the United States—don’t subscribe to this way of thinking and really don’t want you to think that way about elections either. Instead, they see their victories as being morally righteous—not reflecting contingencies like butterfly ballots or the Electoral College or what the inflation rate happens to be,[*4] but rather as embodying the “right side of history” or even God’s will. They see every election as uniquely, existentially important—not drawn from some probability distribution of possible outcomes, as expected value assumes, but its own special snowflake. They also don’t want to permit a whole lot of room for nuance or complexity or pluralistic and probabilistic thinking—it’s hard enough to keep your coalition together, so you don’t want people on your “team” arguing with one another. And they view the whole idea of betting on politics as cringey and morally suspect.

The natural companion to analytic thinking is abstract thinking—that is, trying to derive general rules or principles from the things you observe in the world. Another way to describe this is “model building.” The models can be formal, as in a statistical model or even a philosophical model.[*6] Or they can be informal, as in a mental model, or a set of heuristics (rules of thumb) that adapt well to new situations.

As Sarah Constantin puts it, decoupling is “the ability to block out context…the opposite of holistic thinking. It’s the ability to separate, to view things in the abstract, to play devil’s advocate.” Decoupling has been found by the psychologist Keith Stanovich to correlate with performance on tests of logical and statistical reasoning, a type of intelligence that is valued in the River.

Then there’s the “personality cluster.” These traits are more self-explanatory. People in the River are trying to beat the market. In sports betting, the average player loses money because the house takes a cut of every bet. So if you follow the consensus, you’ll eventually go broke. Investing is more forgiving; just putting your money in index funds still has a positive expected value. Still, professional traders are trying to do better than the market-average return. So part of the job of people in the River inherently involves being critical of consensus thinking, often to the point of being contrarian.

Finally, I put risk tolerance in this cluster because—whether they’re degens or nits in other parts of their lives—being willing to break from the herd and go against the consensus is certainly not the safest professional path. Entrepreneurs usually have high levels of openness to experience and low levels of neuroticism, the “Big Five” personality traits that correlate best with risk tolerance.

Even as a Riverian myself, I have quite a few criticisms of the River, and I think it could use critiques that hit the target more often. So here’s a quick attempt to outline what I think are steelman versions of them. A steelman argument—a favorite technique of EAs and rationalists—is the opposite of a strawman argument. The idea is to build a robust and well-articulated version of the other side’s position, even if it’s one that you disagree with. Let’s begin with the River’s critique of the Village since it’s the one I’m naturally inclined to sympathize with.

Riverians think that partisan position taking often serves as a shortcut for the more nuanced and rigorous analysis that public intellectuals ought to engage in. They think these problems were particularly apparent during the COVID-19 pandemic and that the Village often adopted nakedly partisan positions—from endorsing public gatherings for the George Floyd protests after weeks of telling people to stay at home, to pushing to discourage Pfizer from making any announcement of the efficacy of its COVID-19 vaccine until after the 2020 presidential election—under the guise of scientific expertise.

Riverians worry that Villagers are stifling competition by increasingly focusing on equity of outcomes rather than equality of opportunity. Riverians usually hold the classic capitalist belief that the free market does a better job than central planners in sorting out winners from losers. Furthermore, they believe that market competition benefits society as a whole

by producing technological innovation and economic growth and improvements in the standard of living.

But the Village, not unreasonably, thinks it’s because competitions are often rigged in the River’s favor. By any objective measure, Riverians are powerful incumbents rather than the disruptors they sometimes claim to be, and they benefit from existing social hierarchies; you don’t have to be super-woke to notice that much of the River is very white, very male, and very wealthy.

The Village also questions whether recent technological innovations have in fact benefited society. Silicon Valley may talk a big game about rocket ships to Mars and lifesaving medical technologies, but one of its biggest categories of investment is social media, which has been blamed for everything from the revival of nationalist governments to depression among adolescents.

Most pointedly, it sees Donald Trump and the Republican Party as having characteristics of a fascist movement and argues that it is time for moral clarity and unity against these forces.

So they view the Riverian inclination to poke holes in arguments and “just ask questions’’ as being a waste of time at best, and as potentially empowering a wake of bad-faith actors and bigots.

“Chess is not a game. Chess is a well-defined form of computation. You may not be able to work out the answers, but in theory there must be a solution, a right procedure in any position. Now real games,” he said, “are not like that at

all. Real life is not like that. Real life consists of bluffing, of little tactics of deception, of asking yourself what is the other man going to […] do. And that is what games are about in my theory.”

Game theory is the mathematical study of the strategic behavior of two or more agents (“players”) in situations where their actions dynamically impact one another. It seeks to predict the outcome of those interactions, and to model what strategy each player should employ, to maximize their expected value while accounting for the actions of the other players.

Obviously, people are not always optimally strategic, or even rational at all. But I think it’s a good approach to life to give other people some credit instead of treating them as nonplayer characters, or NPCs, the video-game term for filler characters who have no agency of their own and whose behavior is simple and predetermined. I think of game theory like Frank Sinatra thinks of New York City: “If I can make it there, I’ll make it anywhere.” If you can compete against people performing at their best, you’re going to be a winner in almost any game you play. But if you build a strategy around exploiting inferior competition, it’s unlikely to be a winning approach outside of a specific, narrow setting. What plays well in Peoria doesn’t necessarily play well in New York.

The prisoner’s dilemma is one example of a Nash equilibrium. No player can improve their position by unilaterally changing their strategy. No matter what Isabella does, for instance, Wyatt is better off snitching.

Batting Average in Batter vs. Pitcher Battle Verlander throws fastball Verlander throws curveball Betts guesses fastball .260 .240 Betts guesses curveball .180 .350 What’s the Nash equilibrium? It can be solved with a little algebra—and as you’ve probably guessed, it involves randomization. It turns out that Verlander should throw the fastball about 58 percent of the time and Betts should guess fastball about 89 percent of the time. With this mix, Betts hits for a .252 average. That’s better than he could do with a “pure strategy” where he always picked the same move.

This may be the most important poker-related insight from game theory. You can adopt a strategy that seeks to take advantage of an opponent’s mistakes. But that enables your opponent to take advantage of you if they catch on. For instance, if you’re playing against an old guy who you assume rarely bluffs, he’s going to have some very profitable bluffs if he’s able to switch gears.

That’s what Coates’s book The Hour Between Dog and Wolf is about. The title comes from the French phrase l’heure entre chien et loup—the twilight hour where it becomes difficult to distinguish a dog from a wolf. Metaphorically, the phrase refers to the physical transformation we go through when we’re confronted with a substantial amount of risk. Especially when we encounter novel, uncertain situations—“when a correlation between events breaks down or a new pattern emerges, when something is just not right, this primitive part of the brain registers the change long before conscious awareness”—and we switch from being predictable, domesticated beings to wild creatures living off of cunning, hormones, and animal instinct.

But people in the River are also good at abstraction, skilled at taking data points and drawing out general principles from them. Sometimes—certainly not always—this translates to being ethically principled as well.

When you deviate from the optimal strategy in poker, it can potentially come back to bite you. If you try to exploit someone, you’ll risk being exploited yourself. So poker players often think about what the right play is in the abstract, assuming their opponents are also trying to make the best play. In game theory, there’s a sense of reciprocity—treating others as you wish to be treated.

Prop bets are perfect bets for Peabody, highly technical questions that benefit from elaborate modeling. To estimate the chance of a field goal being scored in the fourth quarter, for instance, you’d ideally want a probabilistic simulation of the entire game—way beyond the capabilities of the average bettor, but Peabody has been working on models like these for years.

You can reduce your risk through careful preparation, and that’s essential. But at 22,000 feet, you can’t eliminate risks entirely. Vescovo didn’t beat himself up about the accident. He didn’t think it was a risk he could have prevented—sometimes those 2 percent chances happen.

Successful risk-takers are conscientiously contrarian. They have theories about why and when the conventional wisdom is wrong. There is an oft-neglected distinction between independence and contrarianism. If I pick vanilla and you pick chocolate because you like chocolate better, you’re being independent. If you pick chocolate because I picked vanilla, you’re being contrarian. Most people are pretty damned conformist—humans are social animals—and Riverians are sometimes accused of being contrarian when they’re just being independent.

Maybe conscientious contrarianism resonates with me because it reminds me of what it’s like to argue with people about politics on the internet. It might not appear this way when you look at my Twitter replies, but my goal when I do this, in line with my background in forecasting, is to seek accuracy and hopefully be proven right by subsequent events.[*16]

So be a conscientious contrarian—look for flaws in people’s incentives rather than their intelligence—and then seek out a place where your own incentives are well-aligned with your goals.

It understands that a relatively remote chance is worth taking if it has a sufficiently high payoff. Indeed, some of the venture capitalists I spoke with for this chapter sounded a lot like poker players. “I always say expected value is the probability of success times the degree of impact.

someone who shared their interest in “embracing variance” and “increasing interestingness.” And the truth is, they’re right. As you’ll see, I don’t agree with everything Silicon Valley does. But I do agree with 80 or 90 percent of it. I say that because I tested myself. Andreessen’s “Techno-Optimist Manifesto” includes 108 statements that assert beliefs that Techno-Optimists hold, such as “Techno-Optimists believe that societies, like sharks, grow or die.” I went through each statement and marked down whether I mostly agreed or mostly disagreed with it. On balance, I agreed with 84 percent of the statements. The exceptions were mostly things I found overstated (“We believe everything good is downstream of growth”), sometimes to the point of being cringe inducing (“We believe in the romance of technology…. The eros of the train, the car, the electric light, the skyscraper”).

Probabilistic thinking works best over repeated trials. If I get all-in with a pair of aces in poker and my opponent makes her flush draw, I’ll feel good about that because I’m going to play thousands more poker hands and I’ll make money from her in the long run. Potentially world-altering technologies are in a funny category, though. If you bundle them into a portfolio of long-shot bets, as VC firms do, then you can take a probabilistic outlook—and if you’re a top VC firm, you’ll be all but guaranteed to make a profit over the long run.

Attitudes of Foxes and Hedgehogs[*12] How Foxes Think How Hedgehogs Think Risk-tolerant: Think in expected-value terms and are willing to act on it. Do not necessarily see themselves as risk-taking; however, their ability to take calculated risks distinguishes them from most people. Risk-ignorant: Not necessarily natural risk-takers. However, because they may misestimate or overestimate their abilities, they may sometimes make decisions that others would regard as incredibly risky. Probabilistic: Think like poker players, i.e., willing to make bets based on incomplete information. Deterministic: Think like chess players, i.e., that outcomes are highly certain once the position is worked out.

Multidisciplinary: Incorporate ideas from different disciplines, read widely across the political spectrum. Specialized: Often have spent the bulk of their careers on one or two great problems. Adaptable: Find a new approach—or pursue multiple avenues at the same time—if they aren’t sure the original one is working. Stalwart: Stick to the same “all-in” approach. Prone to confirmation bias. But will double down when others would quit. Self-critical: Sometimes willing, if rarely happy, to acknowledge mistakes in their predictions and take the blame for them. Stubborn: Mistakes are blamed on bad luck or idiosyncratic circumstances—a good idea had a bad day. Tolerant of complexity: See the universe as complicated, perhaps to the point of many fundamental problems being inherently unpredictable. Order seeking: Have an engineer’s brain. Expect that the world will be found to abide by relatively simple governing relationships once the signal is determined through the noise. Empirical: Rely more on observation and time-tested strategies than theory. Principled: Expect that solutions to many day-to-day problems are manifestations of some grander ideological theory or struggle. May believe that civilization stands at a pivot point.

because political polarization makes it harder to know who to trust, there’s no choice but to be a fox. “There’s no longer experts. There’s only perceived expertise. And even that’s not verifiable anymore,” he said when I brought up Tetlock’s work. “I think expertise is a lost art form today. So you have to be a little bit of a fox and be able to just scurry around and calibrate what you’re getting.”

“Why are second-generation kids never that successful?” asked Social Capital CEO Palihapitiya, who moved with his family from Sri Lanka to Canada and worked at a Burger King to help support them. Instead of the entrepreneurial mother or father, who had “only one curve that they were optimizing for, which is the what-have-I-got-to-lose curve,” the child starts “with the exact inverse curve working against them, which is the risk of embarrassment,” Palihapitiya said. “No matter what the parent says to that child, that person is operating from a perspective where the perception is that they have an enormous amount to lose.”

Bitcoin operates in much the same way; it’s a focal point. As with the case of Grand Central Terminal, it’s not a completely arbitrary choice. Bitcoin was the first cryptocurrency, and being first counts for a lot when you need a focal point. There are also some attributes that give it an aura of permanence; there are a fixed number of Bitcoin that will ever be created—21 million—and Bitcoin’s protocols are notoriously hard to change. Still, it benefits from the fact that some cryptocurrency, or at most a couple of them, has to be the gold

standard. If I want to transact in NateCoin, you want to transact in Bitcoin, and our mutual friend will only accept Dogecoin, none of us are getting any business done and we all lose the game.[*14] Focal points also help to explain the art world—they’re why an Andy Warhol silkscreen of Marilyn Monroe can sell for $195 million even as there are many starving artists who would be thrilled to sell their work for $1,950. The art world is “an envy-based economy,” as New York magazine art critic Jerry Saltz described it to me. “In my world, ninety-nine point nine percent of us are struggling. And we all pay attention to the one percent of the one percent of the one percent.”

But in this final portion of the book, we’re visiting the part of the River where people instead think about open-ended, so-called grand-world problems: everything from where best to spend your charitable contributions to the future of humanity itself. Grand-world problems are much easier to screw up, and the consequences for screwing up are much bigger. If a Riverian type can mess up as badly as SBF did on a congressional primary—a small-world problem—think about one with potentially civilization-altering technology like AI in their hands.

However, the reason some Riverians have become obsessed with grand-world problems is because the Village and the rest of the world screw them up all the time, too, in ways that often reflect political partisanship, an endless array of cognitive biases, innumeracy, hypocrisy, and profound intellectual myopia.

Singer Stream Yudkowsky-Hanson Stream Key Figures Peter Singer, Will MacAskill, Toby Ord Robin Hanson, Eliezer Yudkowsky, Nick Bostrom, Scott Alexander Favorite Subjects Animal welfare, global poverty reduction, effective giving, existential risk Artificial intelligence, futurism, prediction markets, cognitive biases, existential risk Political Orientation Center left, progressive, relatively well aligned with U.S. Democratic Party Libertarian, eclectic, suspicious of major parties Ethical Philosophy Usually but not always explicitly utilitarian Sometimes utilitarian leaning but less committed to any one ethical theory Personality Type Academic, reserved,[*9] well-spoken Nerdy, quirky, provocative misfits Altruistic? Yes in theory and often but not always in practice Not necessarily; think more rigorous evaluation of evidence and probabilistic thinking can be used toward a variety of ends What They Call Themselves Effective altruists Rationalists

First, there’s instrumental rationality. Basically this means: Do you adopt means suitable to your ends? There is a man who has eaten more than thirty thousand Big Macs. Now, this might not be a reasonable and prudent thing for him to do. But if this man’s life goal is to eat as many Big Macs as possible, you could say he’s instrumentally rational because he’s done a bang-up job of

The second type is epistemic rationality. This means: Do you see the world for what it is? Do your beliefs line up with reality? If Big Mac Man thinks his diet is exceptionally healthy, but he’s dying of major organ failure caused by his lack of balanced nutrition, he isn’t being epistemically rational. (Although in fact, he has low cholesterol and is otherwise in good health, or so his wife claims.)

EAs and rationalists are trying to get people to overcome their biases and be more impartial. That’s an extremely thankless task, particularly as the movement becomes more prominent and clashes more often with the Village. Politics, after all, is in some sense about encouraging people to be more partial.

“It is a profound and necessary truth that the deep things in science are not found because they are useful; they are found because it was possible to find them,” Oppenheimer said. Long ago, humankind ate fruit from the tree of knowledge, then began to scale the branches of scientific and technological achievement. It’s those of us who go where the action is who move humanity forward—and who might bring about its demise. Although some might prefer to live ignorantly in an eternal paradise, we are irresistibly drawn toward the path of risk—and reward.

I’d urge you to at least accept the mildest version of doomerism, this simple, one-sentence statement on AI risk—“Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war”—which was signed by the CEOs of the three most highly-regarded AI companies (Altman’s OpenAI, Anthropic, and Google DeepMind) in 2023 along with many of the world’s foremost experts on AI.

Is it really rational to retaliate when you’re doomed already, and the only benefit is in the psychological satisfaction of getting revenge on your enemy?

“Tversky really thought that these biases in human nature were, for lack of a better term, mistakes. You know, errors,” McDermott told me. “It’s like an optical illusion. Once I show it to you, you’ll see the error of your ways and you’ll correct them. And what evolutionary psychology says is, hey, it’s not everything, but there’s a set of behavioral predispositions” that might account for seemingly irrational behavior.