Founding the Field

Walter Pitts

Self-taught logician who, with Warren McCulloch, co-authored the 1943 paper modeling neurons as logic gates — the mathematical starting point for every neural network.

Walter Pitts (1923–1969) was largely self-taught, having left an unhappy home as a teenager and taught himself formal logic from library books. He met Warren McCulloch in Chicago in the early 1940s, and the two spent nights working out whether the all-or-nothing firing of a biological neuron could be described with the same propositional logic Pitts had taught himself. The result, published in 1943, was the McCulloch-Pitts neuron: a unit that sums weighted inputs and fires if the sum clears a threshold.

The paper proved, mathematically, that networks of these simple units could represent any logical function — the first formal argument that a machine built from artificial neurons could in principle compute anything a brain could. It had no notion of learning from data; the weights were set by hand to encode a desired logical function. That piece came later, from Frank Rosenblatt's perceptron, but the basic unit — weighted sum, threshold, fire — is still recognizable in every neural network trained today.

See also: Warren McCulloch, Frank Rosenblatt

Learn more: Neural Networks & Backprop · Wikipedia: Walter Pitts

Mentioned in

Lessons where this comes up in context.