PERSON

Frank Rosenblatt

The Cornell psychologist who invented the perceptron in 1958—the first trainable artificial neural network and the computational ancestor of every AI system running today—was buried by the backlash his own press conference helped create, and died in 1971 without seeing the half-century vindication his foundational bet would receive.

Frank Rosenblatt is the man who placed the central bet of modern artificial intelligence: that the road to machine intelligence ran through learning from examples rather than through the hand-coding of rules. In 1958, working from the Cornell Aeronautical Laboratory in Buffalo, he introduced the perceptron—a single layer of artificial neurons that adjusted its weights in response to error until it could classify patterns it had never been explicitly programmed to recognize. The learning rule he proved, the perceptron convergence theorem, guaranteed that if a solution existed within the machine's reach, the iterative nudging procedure would find it. Every neural network running today descends from that unit: the weighted sum, the threshold, the weight update from error. Rosenblatt did not live to see it. He drowned in Chesapeake Bay on his forty-third birthday in 1971, inside the AI winter that a devastating 1969 critique had helped bring on,

Keep reading with YOU ON AI