The first paper on neural networks was written by a homeless teenager

Walter Pitts ran away from home in Detroit at age 15. He had read Russell and Whitehead's Principia Mathematica at 12 and decided to find Bertrand Russell. He found instead the University of Chicago, where he attended seminars without enrollment, did odd jobs, and slept on park benches.

In 1942, Pitts met Warren McCulloch, a neurophysiologist working on whether nervous systems could compute in the way Turing had described for machines. McCulloch invited the homeless eighteen-year-old to live with his family. They worked together at night. In 1943, they published "A Logical Calculus of the Ideas Immanent in Nervous Activity" in the Bulletin of Mathematical Biophysics. The paper proposed the first mathematical model of a neural network. Pitts was twenty. He had no academic credentials. He would never earn more than an Associate of Arts.

The 1943 paper did three things at once. It introduced the artificial neuron as a logical threshold device. It showed networks of such neurons could compute any Boolean function. It used computation to address the mind-body problem, which philosophy had been arguing about for three centuries.

Norbert Wiener, MIT's resident genius, recognized the paper's importance and brought Pitts into his research circle. By the early 1950s, Wiener cut all contact with Pitts and McCulloch for reasons that remain disputed. Pitts was devastated. He drank heavily, burned much of his unpublished work, and died in 1969 at forty-six.

The mythology of AI traces the field back to Turing, McCarthy, Minsky. The 1943 McCulloch-Pitts paper is the actual origin point for neural networks. Every modern neural network architecture descends from it. Almost no AI textbook tells the reader that one of its authors had no diploma.