What this covers
Nate Silver joins Sean Carroll to examine how people make high-stakes decisions under uncertainty, drawing on decades of experience forecasting elections, analyzing sports, and playing professional poker. The conversation centers on a recognizable personality type—Silver calls them "Riverians"—analytically skilled, competitive, and comfortable with risk, who populate poker rooms, venture capital, finance, and rationalist circles. Silver argues that this cluster tends toward expected value maximization and quantification, yet pure EV reasoning justifies catastrophic bets that intuition rejects. The discussion traces how decoupling—the ability to separate what will happen from what you want to happen—defines good forecasting but carries hidden costs.
Silver walks through concrete failures of pure rationality. Sam Bankman Fried endorsed bets that could destroy civilization if they offered even marginally better odds, a position Silver pins to the St. Petersburg Paradox: rational agents must incorporate risk of ruin and common sense alongside probability. The Riverians' tendency to quantify everything can metastasize into engineer's disease, where oversimplified models ignore real-world fragility. Separately, Silver examines when expected value thinking actually applies—poker and venture capital benefit from repeated trials and portfolio diversification, but irreversible life decisions demand different rationality. He discusses prediction markets, the difficulty of election forecasting (too few trials to verify skill), why people compress all probabilities toward zero or fifty-fifty, and how poker trained his intuition for probability by making the long run visceral rather than abstract. The conversation also touches on technology's ambiguous trajectory, utilitarian thinking's blind spots, and Silver's own forecast errors, particularly his confident misjudgment of Donald Trump in 2016.
Silver argues that a distinct personality cluster—analytically skilled, competitive, and risk-tolerant 'Riverians'—drives high-stakes decision-making, and that good probabilistic reasoning means decoupling forecasts from desired outcomes while tempering pure expected-value maximization with risk aversion and common sense.
- Expected-value reasoning alone justifies catastrophic bets (SBF's St. Petersburg logic), so rationality must incorporate risk-of-ruin and common sense.
- Decoupling—separating objective forecasts from preferred outcomes—is a hard but essential skill that defines good forecasters.
- A recognizable analytical, competitive, risk-tolerant personality type recurs across poker, finance, VC, EA, and rationalism.
High P-doom estimates extrapolate excessively from very limited empirical evidence.
- Carroll estimates P-doom at about 10^-3 or less, arguing the large numbers others cite come from extrapolating from very tiny bits of data; Silver, as a Bayesian, shades down post-LLM-pause but defers somewhat to serious people who have thought more about it, noting one learns not to give full weight to one's gut.
“for me it's more like 10 to the minus three or less. I think that these are crazy big numbers that people are getting by extrapolating from very tiny bits of data.”