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

Before the Prompts: The Untold Story of AARON and the First AI-Generated Images of 1973

Long before text-to-image generators allowed anyone to conjure hyper-realistic masterpieces with a casual sentence, the story of artificial intelligence art began in a quiet laboratory in the early 1970s.

While today’s algorithms rely on massive neural networks trained on billions of scraped internet photographs, the very first true AI image generator was born out of a profound philosophical question asked by a single painter: Can a machine learn to create on its own?

Enter Harold Cohen and his creation, AARON.

The Man Behind the Machine

In the late 1960s and early 1970s, Harold Cohen was a celebrated British-born artist representing Britain at prestigious global showcases like the Venice Biennale. Yet, at the height of his traditional painting career, he chose to walk away from the canvas to pursue an unexpected fascination with computing.

Relocating to the United States—eventually teaching at the University of California, San Diego—Cohen began collaborating with computers. He didn’t want to use a computer merely as a digital paintbrush or a tool to trace pre-existing designs; he wanted to encode human cognitive processes into software.

By 1973, those efforts crystallized into AARON, widely recognized as the world’s first autonomous AI art system.

How 1973 AI Art Actually Worked

AARON’s architecture bore almost no resemblance to modern machine learning. There was no deep learning, no cloud computing, and no web scraping. Instead, Cohen painstakingly handcrafted a set of symbolic rules:

• Cognitive Rules: Understanding the relationship between foreground and background.
• Structural Rules: Mathematical representations of how organic shapes, plant life, and eventually human figures grow and occupy space.
• Compositional Rules: Guidelines on how elements should balance within a framed boundary.

Once Cohen hit “start,” he stepped away. AARON would drive a physical drawing machine—a robotic turtle holding a pen over a large sheet of paper—to execute compositions entirely on its own.

The resulting 1973 outputs were intricate, abstract black-and-white line drawings. Crucially, Cohen himself could not entirely predict what AARON would draw. The software combined rules with randomized parameters, meaning the machine generated novel visual paths that even its creator had not consciously envisioned.

A Decades-Long Evolution

What makes AARON remarkable isn’t just its origin in 1973, but its longevity. Cohen did not treat AARON as a one-off experiment; he spent the next four decades expanding its capabilities:

• The 1970s: Abstract geometric compositions and primitive line work.
• The 1980s and 90s: The introduction of closed-form objects, rocks, plants, and complex human figures.
• The 2000s: AARON evolved to handle color autonomously, painting rich, vibrant scenes of figures and foliage that were eventually exhibited in world-class institutions like the Tate Modern and the Whitney Museum of American Art.

Why 1973 Still Matters Today

Today, we live in an era where AI-generated imagery is ubiquitous, contentious, and lightning-fast. It is easy to look at modern hyper-realistic renders and forget the humble, deliberate steps taken decades ago.

AARON’s debut in 1973 fundamentally shifted human imagination. It proved that creativity was not an exclusively biological magic trick, but a process that could be understood, translated, and mirrored by code. Harold Cohen and his digital collaborator laid the philosophical and technical groundwork for every pixel of AI art we see today, proving that the marriage of art and artificial intelligence has a history stretching back half a century.