Walter Harry Pitts Jr. (1923-1969) was an American logician and one of the original architects of neural-network theory. Largely self-taught, he had no high school diploma; as a boy in Detroit he taught himself logic, mathematics, Greek, and Latin, and at twelve he reportedly read Russell and Whitehead’s “Principia Mathematica” and wrote to Bertrand Russell pointing out errors. He left home as a teenager and found his way to the University of Chicago, where he met the neuroscientist Warren McCulloch.
With McCulloch he co-authored the 1943 paper “A Logical Calculus of the Ideas Immanent in Nervous Activity,” which treated each neuron as an all-or-none logical switch and proved that networks of such units could compute any logical proposition. The IEEE Computer Society’s pioneers profile credits Pitts as co-inventor of the binary neuron “that bears their names, which has strongly influenced basic concepts in AI.” He later worked alongside Norbert Wiener at MIT, where Wiener called him an extraordinary talent.
Pitts is also one of the field’s first casualties. After a personal break with Wiener around 1952 cut him off from his mentor, and after experiments suggested the brain was messier than his clean logical models allowed, he withdrew, drank heavily, refused offered degrees and positions, and reportedly destroyed his own unfinished work. He died alone in a Cambridge boarding house in 1969 of bleeding esophageal varices, a condition tied to cirrhosis. The artificial neuron he helped invent is the direct ancestor of the units in every modern neural network.