The First AI-Generated Image: The Story of AARON in 1973

The Genesis of Machine Dreams: How 1973 Sparked the Era of AI Art

When we think of artificial intelligence and art today, our minds instantly drift to modern text-to-image engines—sophisticated neural networks that spin hyper-realistic masterpieces out of a single typed sentence in seconds. It feels like a futuristic phenomenon born entirely of the 2020s.

Yet, the conceptual lineage of AI-generated imagery stretches back over half a century. Long before GPUs, deep learning, or massive internet image scrapers, the very first AI-generated image was born in 1973, crafted not by a Silicon Valley corporation, but by a bewildered, brilliant British painter seeking to answer a radical philosophical question.

The Artist Who Wanted to Teach a Machine to Think

In the late 1960s, Harold Cohen was at the absolute height of the contemporary art world. Representing Great Britain at prestigious international showcases like the Venice Biennale, he was an established master of abstract painting. Yet, success left him restless. Cohen became consumed by a fundamental question: What is the absolute minimum requirement for an image to be recognized as art?

He wasn’t interested in using a computer merely as a digital paintbrush or a high-tech calculator. Instead, Cohen wanted to externalize the cognitive process of drawing. He envisioned an autonomous entity—a software program that possessed its own understanding of the rules of art (composition, perspective, closure, and proportion) and could make independent decisions.

Moving to the University of California, San Diego, and spending a pivotal residency at the Stanford Artificial Intelligence Laboratory in 1973, Cohen began writing the foundational code for a software program that would change cultural history. He named his digital collaborator AARON.

What Did the First AI Images Look Like?

If you picture a 1973 computer, you likely imagine massive refrigerator-sized mainframes humming in dark rooms. AARON lived on one such machine.

The resulting output bore zero resemblance to modern digital art:

• The Medium: Simple, elegant black-and-white line drawings.
• The Hardware: A mainframe computer communicating with a specialized drawing machine—often a flatbed plotter or a tethered mechanical “turtle” that crawled across a giant sheet of paper spread out on the floor, holding a pen.
• The Style: Abstract, organic shapes, primitive human-like figures, rocks, and foliage.

Crucially, AARON did not copy pre-existing images. Unlike modern machine learning models that ingest billions of internet pictures, AARON had no training dataset. Instead, Cohen manually coded a complex rulebook of human cognition and artistic composition.

AARON understood how to construct a figure, how to overlap objects so one appeared in front of another, and how to balance visual weight across a canvas. Because the software incorporated internal randomization within its strict rules, every single image AARON produced was completely unique. The machine was making creative choices of its own within boundaries set by its human creator.

From Floor Turtles to the Tate Modern

Over the next four decades, Harold Cohen continued to refine AARON, mirroring the rapid evolution of computing hardware itself:

• The 1980s and 90s: Cohen taught AARON how to handle color. The software began generating its own color palettes and painting instructions, which Cohen would manually mix and apply to canvas.
• The 2000s: Later iterations, such as AARON KCAT, evolved to generate lush, complex botanical and jungle-like scenes projected onto gallery walls or printed directly.

AARON’s work grew so sophisticated and culturally significant that it transcended the computer science lab, landing in some of the world’s most prestigious cultural institutions, including the Tate Modern in London and the Whitney Museum of American Art in New York.

Why the 1973 Milestone Still Matters

Harold Cohen passed away in 2016, having spent more than 40 years in a profound, uninterrupted creative partnership with his digital creation. He never viewed AARON as a tool meant to replace the artist; rather, he viewed the software as a separate entity—an electronic collaborator.