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 McCullochWarren McCullochNeurophysiologist who, with Walter Pitts, proposed the first mathematical model of a neuron in 1943 — the ancestor of every neural network built since. 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 RosenblattFrank RosenblattPsychologist who built the perceptron in 1958, the first neural network that learned its own weights from data rather than having them set by hand.'s perceptronPerceptronThe Perceptron (1958) was the first learning system built from an artificial neuron, and the direct ancestor of the neural network training loop., but the basic unit — weighted sum, threshold, fire — is still recognizable in every neural networkNeural NetworkA neural network is layers of simple weighted-sum-plus-nonlinearity units (neurons) chained together, trained by gradient descent and backpropagation. trained today.
See also: Warren McCullochWarren McCullochNeurophysiologist who, with Walter Pitts, proposed the first mathematical model of a neuron in 1943 — the ancestor of every neural network built since., Frank RosenblattFrank RosenblattPsychologist who built the perceptron in 1958, the first neural network that learned its own weights from data rather than having them set by hand.
Learn more: Neural Networks & Backprop · Wikipedia: Walter Pitts
Mentioned in
Lessons where this comes up in context.
Claude Shannon
Engineer who founded information theory, giving the field the mathematical vocabulary — bits, entropy, channel capacity — that loss functions and compression are still built on.
John McCarthy
Coined the term "artificial intelligence" in 1955, organized the field's founding 1956 Dartmouth workshop, and invented Lisp.