Before Neurons and Prompts: The True Story of the First AI-Generated Image in 1973

When we think of artificial intelligence and art today, our minds instantly jump to modern text-to-image engines—systems that ingest billions of internet parameters, crunch data through massive graphic cards, and spit out hyper-realistic paintings or photographs in mere seconds.

Yet, long before the terms “diffusion models,” “neural networks,” or “prompts” entered the public lexicon, the genesis of machine-made imagery took place in a quiet university lab in 1973.

The creator was not a computer scientist looking to disrupt the tech world, but a celebrated British painter. His name was Harold Cohen, and his creation was a software program named AARON.

Meet Harold Cohen and AARON

In the late 1960s, Harold Cohen was an established artist representing Great Britain at prestigious global showcases like the Venice Biennale. However, a profound curiosity gripped him: What is the core essence of the act of drawing? What cognitive rules allow a human hand to translate abstract thoughts into physical lines on a page?

To answer this, Cohen walked away from his traditional canvas and stepped into the unfamiliar world of computing. He took a position at the University of California, San Diego, and began teaching himself programming.

By 1973, his efforts materialized into AARON. Running on a mainframe computer, AARON was designed to generate original, autonomous visual compositions without human intervention once the code was executed.

“The resulting 1973 outputs were abstract, intricate compositions of lines and closed forms. Crucially, once Cohen pressed start, he did not intervene in the creation of the drawing. The machine conceived and executed the artwork on its own.”

How Did the 1973 AI “Think”?

Unlike today’s AI, which learns statistically by scraping billions of images from the web, AARON possessed zero data about existing photographs or paintings. It didn’t “look” at the world.

Instead, AARON operated on manually coded cognitive rules written by Cohen.

• The Logic of Space: Cohen programmed rules about how lines intersect, how shapes overlap, and how to maintain a sense of balance and composition across a 2D surface.
• Autonomous Decision-Making: Once the program booted up, the software made independent choices about where to place forms, how large to draw them, and when a composition was complete.

A mechanical drawing arm—a flatbed plotter connected to the computer—would then physically grip a pen and meticulously drag it across paper, translating the computer’s code into ink lines.

A Forty-Year Partnership

What makes the story of AARON and that pivotal milestone in 1973 so fascinating is that it was not a one-off science experiment. Cohen spent the next four decades of his life continuously updating AARON.

As computing power evolved, so did AARON:

• 1970s: Abstract, black-and-white geometric lines and closed forms.
• 1980s & 90s: The software evolved to understand closed shapes that began representing rocks, plants, and eventually human figures.
• Later Years: AARON was even given a digital palette and automated brush-washing systems to paint full-color compositions displayed in world-renowned institutions like the Tate Modern and the Whitney Museum of American Art.

Why 1973 Matters Today

Looking back from an era where AI-generated art is ubiquitous, controversial, and technologically complex, the abstract line drawings of 1973 look quaint. Yet, AARON fundamentally challenged the romantic notion of art as an exclusively human endeavor born of a mystical soul.

Harold Cohen viewed his creation not as a tool that he used, but as an entity he collaborated with—much like a Renaissance master working alongside a studio assistant.

The next time you type a prompt into an image generator and watch a digital masterpiece materialize in seconds, spare a thought for 1973. That was the year a painter and a mainframe computer first proved that a machine could hold the pen, look at a blank page, and decide to create something entirely new.