The first artificial neuron

The mathematical neuron model of McCulloch and Pitts

In 1943, neuroscientist Warren McCulloch and mathematician Walter Pitts published their groundbreaking paper A Logical Calculus of the Ideas Immanent in Nervous Activity in the journal Bulletin of Mathematical Biophysics. This paper is considered the starting point of what would later become computational neuroscience and artificial intelligence. It was the first formal attempt to describe the workings of the human brain in mathematical terms.

The unlikely collaboration

McCulloch was a physician and neuroscientist who had spent years trying to understand how the brain works as an information processor. Walter Pitts was an exceptionally gifted but socially vulnerable young man who as a teenager scoured the libraries of Chicago and at age twelve studied Bertrand Russell's three-volume Principia Mathematica. When Pitts was wandering homeless in Chicago, McCulloch offered him shelter and an intellectual home. Their collaboration was a rare cross-pollination of biology and mathematical logic that seemed nearly impossible at the time.

The model: logic as biology

A McCulloch-Pitts neuron receives one or more input signals, sums them, and produces an output signal if the sum exceeds a certain threshold. This binary principle — fire or don't fire — made it possible to model logical operations such as AND, OR, and NOT as networks of neurons. The central thesis of their paper was provocative: any logical computation that can be performed by a digital computer can in principle also be performed by a network of such neurons. Conversely, this seemed to imply that the human brain, which is also composed of neurons, essentially performs computations like a calculator.

Philosophical significance

This reasoning had far-reaching consequences for the way scientists viewed the brain. It broke the idea that consciousness and intelligence were mysterious and inexplicable. If neurons are logic gates and networks perform computations, then the brain is — at least theoretically — a computing machine. This was a radically materialist position at a time when brain science was still largely descriptive and anatomical. It laid the philosophical foundation for the computational theory of mind, which remains central to cognitive science and AI research to this day.

Direct influence on pioneers

John von Neumann, the founder of modern computer architecture, explicitly referenced the work of McCulloch and Pitts in his First Draft of a Report on the EDVAC (1945). Norbert Wiener, the father of cybernetics, was inspired by McCulloch to investigate the parallels between brains and machines. Donald Hebb formulated his Hebbian learning rule in 1949 partly inspired by this work. Frank Rosenblatt developed the Perceptron in 1957 as a direct successor. Marvin Minsky built SNARC in 1951, the first hardware neural network, on similar principles.

Legacy in modern deep learning

Decades later, when deep learning dominated the technology world, the fundamental building blocks were still direct descendants of the McCulloch-Pitts neuron. Nodes with input weights, a summation, and an activation function are the basic atoms of every modern neural network. The threshold had been replaced by the sigmoid or ReLU activation function, but the core idea from 1943 was recognizably present in AlexNet, GPT-4, and every other model. Without McCulloch and Pitts, no Transformer, no ChatGPT, no Claude.

Warren McCulloch died in 1969. Walter Pitts died in the same year at just 46 years of age, after a period of withdrawal and mental health problems. Their collaboration lasted barely a few years, but the intellectual legacy it left carries the digital world to this day.

Sources

• McCulloch, W.S. & Pitts, W. (1943). A Logical Calculus of the Ideas Immanent in Nervous Activity. Bulletin of Mathematical Biophysics, 5(4), 115–133.
• Piccinini, G. (2004). The First Computational Theory of Mind and Brain. Synthese, 141(2), 175–215.
• Churchland, P.S. & Sejnowski, T.J. (1992). The Computational Brain. MIT Press.
• Wikipedia — Walter Pitts
• Wikipedia — Warren McCulloch