The Brush That Never Slept: Inside 1973 and the Birth of AARON, the World’s First AI Artist

Long before text prompts could summon hyper-realistic oil paintings in seconds, and decades before neural networks sparked a global debate over machine creativity, a British painter packed his bags for California with a radical idea.

His name was Harold Cohen, an internationally acclaimed artist who had represented Great Britain at the prestigious Venice Biennale. By the late 1960s, Cohen had grown fascinated by a provocative question: What makes a drawing work? More specifically, could the cognitive rules of human mark-making be translated into computer code so that a machine could create original art entirely on its own?

The answer arrived in 1973 with the public unveiling of AARON—the world’s first artificial intelligence image generator.

From the Canvas to the Mainframe

In an era when computers occupied entire rooms, whirred with magnetic tape, and were primarily used to solve logic puzzles or calculate ballistic trajectories, treating a computer as an artist was blasphemous. Most early computer graphics experiments relied on pure randomness—varying mathematical parameters and seeing what spilled out of a printer. The results were typically cold, geometric, and largely accidental.

Cohen took a completely different path. As a master painter, he wasn’t interested in random computer noise; he wanted to codify cognition.

While working at the University of California, San Diego (and previously at Stanford’s Artificial Intelligence Laboratory), Cohen spent months breaking down the foundational grammar of art. He asked himself fundamental questions:

• How does the human eye perceive boundaries?
• When is a closed shape structurally sound?
• How do overlapping lines create a sense of depth and composition?

Rather than teaching a machine to copy an existing picture—which is how modern machine learning models operate today—Cohen manually wrote a massive rule-based system. He programmed AARON with cognitive rules about space, occlusion, and proportion.

The Birth of an Autonomous Creator

When AARON generated its breakthrough outputs in 1973, it did something extraordinary: it drew autonomously.

Once Cohen hit “start,” he stepped away. The computer program made independent decisions about composition, deciding where to place lines and how to construct intricate, abstract forms on paper. Because AARON lacked a screen display capable of rendering complex fine art, the program’s digital calculations were translated into physical movement via a pen plotter—a mechanical arm holding a pen that meticulously dragged across paper under the computer’s command.

The resulting 1973 drawings were mesmerizing: intricate, abstract structures of closed forms and delicate lines that looked purposeful, balanced, and distinctly human-composed. Yet, no human hand guided the pen.

A Forty-Year Partnership

Most technological proofs-of-concept from the 1970s were short-lived, fading into obscurity once grant money or interest dried up. But AARON was different. It marked the beginning of a four-decade collaboration between Cohen and his software.