Mark I Perceptron: The Room-Size Machine That Mistook George Harrison for a Woman

In 1958, psychologist Frank Rosenblatt unveiled the Mark I Perceptron at Cornell University, a room-size grid of 400 light-registering sensors designed to perceive and sort images. The New York Times heralded it as an "Electronic 'Brain' Teaches Itself." Rosenblatt claimed the machine could "grow wiser as it gains experience." The Perceptron mimicked a single neuron, using yes/no feedback to learn—for instance, distinguishing men from women. In a mid-1960s training session, it misidentified George Harrison of the Beatles as a woman due to his long hair. Despite its limitations, the Perceptron was the first step toward modern AI. Rosenblatt died in a 1971 boating accident at age 43. His 1962 book, Principles of Neurodynamics, is considered the first textbook on artificial neural networks. The term "artificial intelligence" originated at a 1956 Dartmouth conference. Neural networks advanced in the late 1980s when Yann LeCun's Bell Labs team trained machines to recognize handwriting for ZIP codes. Crucial developments came from GPUs (graphics processing units) from the computer games industry and the internet, enabling massive data training. The original Mark I machine is now in the Smithsonian's collections.