Max Read has a recent sharp piece in the New York Times Magazine, "Gambling Is the Culture," that I want to pass on, along with some thoughts it set off in me. His argument is that gambling money no longer just pays for our culture. It has become the culture. Sportsbooks, offshore casinos, crypto, options-trading apps and prediction markets like Kalshi and Polymarket are merging into one speculative sector. That sector now underwrites sports broadcasting, podcasts, the streamer-influencer economy and increasingly the news itself, through data partnerships with CNN, Fox and Dow Jones.
The more interesting part of the essay goes below the money. Read traces it all to what the philosopher Ian Hacking called the "probabilistic revolution": the slow replacement of causal, particular explanation with statistical inference over distributions. Statistics was born in gambling (French mathematicians arguing about dice), grew up in 19th-century state censuses and mortality tables, and now lives in Silicon Valley. Venture capital buys exposure to a distribution rather than picking winners. Recommendation algorithms estimate conditional probabilities without knowing anything about you. Ad auctions price your attention billions of times a day. And large language models are, in Read's framing, the purest expression of the statistical style: probability distributions over text, with no beliefs and no facts, which nonetheless work astonishingly well.
The style's success justifies its spread into the self. Engineers "update their priors" on Thai food, venture capitalists call a walk "+EV," and party conversation turns to one's "p(doom)." Read notes that people once pictured the mind as clockwork, then hydraulics, then a computer. The newest self-image is the mind as a stochastic inference engine. He treats this as a cultural affectation of the tech elite, and a costly one. Constantly calculating your personal risk profile, hedging every bet and avoiding commitment to keep options open is, he says, exhausting.
Here is where I want to push back, and then agree with him for a different reason.
The brain really is an inference engine. This is not a Silicon Valley fad. It goes back to Helmholtz's "unconscious inference" in the 1860s, and over the past twenty years it has become one of the central frameworks in neuroscience, in the predictive processing and active inference work of Karl Friston, Andy Clark (Surfing Uncertainty), Anil Seth (Being You) and many others, as long-time MindBlog readers know. On this view the brain does not passively receive the world. It continually generates predictions about incoming sensory signals and about the body's own internal state, and updates a hierarchy of models according to prediction errors weighted by their estimated reliability. Perception is a controlled hallucination; action is a way of making predictions come true. So the people at Read's Silicon Valley parties are, in a sense, describing their brains more accurately than the clockwork and computer metaphors ever did.
But what the brain's inference is for matters. The predictive brain did not evolve to maximize expected payouts. It evolved for allostasis, keeping a body alive by anticipating its needs before they become emergencies. Its deep imperative is to reduce uncertainty about the states it must stay within, not to seek out variance. And it does this cheaply. Most of our everyday competence rests on stable, high-confidence priors: habits, routines, commitments, trusted relationships and shared cultural expectations. These let us stop recomputing the world from scratch. A commitment, in this light, is a metabolic economy. It pins down a prior so that resources can go elsewhere.
That gives us a neurobiological reading of Read's exhaustion at the deli counter. He describes himself and his pregnant wife weighing Listeria risk against her craving for prosciutto, and finding that individually tailored risk drained them as much as blanket medical rules constrained them. Of course it did. Holding every option open and recalculating every choice is just what a predictive system is built to avoid. Bounded agents depend on settled expectations, many of them borrowed from other people. Offloading those settled expectations onto institutions, traditions and trusted authorities is not epistemic laziness. It is how a finite brain affords to function at all.
The gambling industry exploits the same machinery from the other side. Wolfram Schultz's classic work showed that midbrain dopamine neurons signal reward prediction error, the gap between expected and received reward. Uncertain, variable rewards keep that signal firing in a way reliable rewards don't, which is why slot machines, loot boxes, infinite feeds and parlay bets are so compelling. A system that evolved to learn from surprises can be captured by an environment engineered to deliver surprises forever. The irony is that an industry built on the vocabulary of rational Bayesian updating works best on users whose updating machinery has been hijacked.
Finally, Read's best point is about insurance. One of the earliest institutions built on probability theory was insurance, which pools risk collectively. Shifting probabilistic calculation away from institutions and onto individuals — from insuring and investing to betting — is recent. And, as he notes, it serves the house: casinos need marks, crypto needs bag holders, prediction markets need dumb money. I would put it in predictive-processing terms. Communities, norms and institutions are shared generative models that hold uncertainty for us. Dissolve them, hand each person the job of hedging their own health, career, romance and civic life, and you get not a society of superforecasters but a population of chronically overloaded inference engines, ideal customers for anyone selling a dopamine hit dressed up as an expected-value calculation.
Read closes with Andreessen Horowitz's Alex Danco, who imagines a future where AI does the real work and humans' remaining purpose is making predictions to feed the machines, and who finds this optimistic. Read's reply is that it is easy for Danco to say, since he has already made all the right bets. To which I would add: a brain that must bet on everything all the time is a brain with no stable ground to stand on. The task, for individuals and for our institutions, is to choose which priors we are willing to hold firmly, and to rebuild the collective structures that let us do so.
The text above was generated by Claude Opus 5.5 as if I was writing it. Thus the "I" in the text is Claude.