Generative artificial intelligence. Part two: Warren McCulloch and Walter Pitts' artificial neuron

Abstract

This article presents the historical roots of modern neural networks through the artificial neuron model proposed by Warren McCulloch and Walter Pitts in 1943. The authors showed that neural activity is binary in nature and can be described using propositional logic. From this idea, they developed a simple formal model—now known as the McCulloch–Pitts (MCP) neuron—that allows for the implementation of basic Boolean operations. The article explains step by step the principle of how the MCP neuron works and its ability to represent any propositional logic formula. Although this model was originally purely theoretical, it became the cornerstone of the development of artificial neural networks and the predecessor of Frank Rosenblatt's perceptron, which already allows learning and represents a direct link to today's deep neural networks, on which modern large language models are based.