The History of the First AI-Generated Image

The Ghost in the Machine: The Untold Story of the First AI-Generated Image

Long before the internet was flooded with hyper-realistic portraits, surreal dreamscapes, and prompt-engineered masterpieces born from modern diffusion models, a quiet revolution took place.

We tend to think of artificial intelligence art as a recent phenomenon—a technological sprint that seemingly materialized overnight. But the journey of machines creating visual art actually began more than half a century ago. The story of the very first AI-generated image is not one of neural networks or massive data scraping, but of a lone artist, a mainframe computer, and a radical question: Can a machine learn to draw?

Meet AARON: The World’s First AI Artist

The pioneer behind the world’s first AI-generated art was Harold Cohen, a successful British abstract painter who represented Great Britain at the prestigious Venice Biennale in 1966. At the height of his traditional career, Cohen grew deeply fascinated by the emerging world of computing. Moving to the United States, he took a position at the University of California, San Diego, and set out to decode the human act of drawing.

In the late 1960s and early 1970s, Cohen began developing a computer program that would eventually be known simply as AARON.

Unlike modern AI systems that scan billions of images to mimic styles using deep learning, AARON had no access to a dataset, a GPU, or the internet. Instead, AARON was built on symbolic, rule-based logic (often referred to as Good Old-Fashioned AI, or GOFAI). Cohen painstakingly translated his own understanding of cognitive and artistic rules into code. He taught the program how to handle composition, how objects overlap in space, how to balance proportions, and how lines intersect.

The First Breakthrough: 1973

By 1973, AARON was capable of producing its first autonomous line drawings.

Using a custom-built mechanical drawing machine—a robotic arm connected to a computer—AARON would physically hold a pen and plot lines onto paper based on real-time computational decisions. The human artist didn’t dictate what the final composition would look like; Cohen wrote the rules, but AARON made the creative choices regarding where to place shapes, how to balance the canvas, and what abstract figures to manifest.

The earliest outputs were modest: stark, black-and-white geometric and organic line drawings. Yet, their historical weight cannot be overstated. When these pieces were exhibited at major institutions like the Los Angeles County Museum of Art, the Tate Modern, and the Whitney Museum of American Art, they marked a staggering milestone. For the first time in human history, an autonomous non-human intelligence had created original visual art.

From Rules to Pixels: The Evolution

For decades, Cohen continued to evolve AARON. As computing power grew, so did the program’s capabilities:

• The 1980s and 90s: Cohen updated AARON to understand color theory, allowing the program to autonomously select paints and fill in the shapes it drew.
• The 2000s: AARON evolved from abstract shapes into creating rich, figurative compositions featuring stylized human figures, plants, and complex jungle-like environments.

Cohen famously viewed AARON not as a tool (like a brush or a camera), but as an collaborator—an independent entity possessing its own specialized mind. When Cohen passed away in 2016, AARON’s active development ceased, but its legacy was permanently cemented into the history of technology and art.

Why the First AI Image Still Matters Today

When we look at a modern AI image generator outputting a cinematic masterpiece in seconds, it’s easy to view AARON’s primitive black-and-white lines as a quaint historical footnote. But AARON proved a profound philosophical concept decades before society was ready for it: Creativity is not exclusively a biological trait.

Harold Cohen’s creation bridged the chasm between computer science and fine art, proving that a machine could conceptualize, decide, and execute visual expression independently. AARON walked so that modern generative models—from GANs to state-of-the-art diffusion transformers—could run.