PERSON
Walter Pitts
The homeless teenager who reduced the brain to a logic circuit and gave every artificial neuron in every neural network its founding abstraction—then helped run the experiment that showed his founding dream was wrong, and could not survive the discovery.
Walter Pitts is the most consequential mind most people who use artificial intelligence have never heard of. In 1943, before he was twenty, with no degree and no fixed address, he co-wrote “A Logical Calculus of the Ideas Immanent in Nervous Activity” with the psychiatrist Warren McCulloch—the first mathematical model of a neuron, the founding document of both artificial intelligence and theoretical neuroscience, and the direct ancestor of every
neural network in existence today. The unit they described—a threshold switch that sums weighted inputs and fires or does not—is the atom of the technology that now writes, translates, and unsettles us, softened from a hard step to a smooth curve but recognizably the thing Pitts wrote down in a Chicago suburb at nineteen. He had read *Principia Mathematica* at twelve, run away from home at fifteen to be near logic, and built his identity on the proposition that the mind was a theorem-proving machine and the universe