Cells that fire together wire together

McCulloch and Pitts had shown what a network of neurons could compute. They had said nothing about how it could learn. In 1949 the Canadian psychologist Donald Hebb supplied the missing piece in a single sentence: when one cell repeatedly helps to fire another, the connection between them grows stronger. The slogan came later, but the idea was his, and it is the first rule anyone wrote down for changing the weights of a network in response to experience.

Hebb was writing about brains, not machines. He wanted to explain how perception and memory could emerge from cells that individually knew nothing. His answer was the cell assembly: a group of neurons that, having fired together often enough, become a unit that can be triggered as a whole. A thought was a pattern of strengthened connections.

Marvin Minsky's SNARC of 1951 was an attempt to build a Hebbian learner out of vacuum tubes. Rosenblatt's perceptron replaced Hebb's rule with an error-driven one, and backpropagation later replaced that. But the frame Hebb set has held: learning is a change in weights, and knowledge lives in the connections rather than in any single cell.