The Ghost in the Machine: The Story of AARON and the First AI Image

Long before the internet was flooded with hyper-realistic portraits generated by text prompts, and decades before neural networks learned to mimic Van Gogh or cyberpunk aesthetics, a British painter packed his bags, moved to California, and asked a radical question: Can a machine learn to draw?

His name was Harold Cohen, and in the early 1970s, he created AARON—the computer program widely recognized as the world’s first artificial intelligence art system.

A Painter Who Traded His Brushes for Code

To understand the birth of the first AI-generated image, you have to look at its creator. In the 1960s, Harold Cohen was a celebrated traditional artist. He had represented Great Britain at the prestigious Venice Biennale and taught at the Slade School of Fine Arts in London. He was firmly entrenched in the international art establishment.

Yet, Cohen grew deeply curious about the cognitive mechanics of art. What is it that makes a collection of lines on a flat surface evoke a human figure or a plant in the mind of a viewer? Why do we perceive intention in certain marks and not others?

Seeking answers, Cohen left his successful career in London in 1968 to take a post at the University of California, San Diego. There, surrounded by early computer scientists, he decided to teach himself how to code. He didn’t want to use the computer as a mere digital paintbrush or a fancy calculator; he wanted to encode the rules of human cognition and visual representation into software, allowing a machine to make autonomous creative decisions.

Meet AARON: The Pioneer of Autonomous Art

By 1973, Cohen’s experiment bore fruit. He unveiled AARON, a primitive rule-based software system designed to generate original line drawings.

Unlike modern generative AI, which relies on statistical deep learning trained on billions of scraped photographs and paintings, AARON operated entirely on symbolic logic. Cohen manually programmed the software with rules about geometry, perspective, occlusion, and human anatomy. He didn’t tell AARON what to draw line-for-line; instead, he gave it a foundational understanding of the physical world and cognitive visual rules, allowing the program to compose its own images on the fly.

Connected to a custom-built drawing machine—a mechanical arm holding a pen that hovered over a flat sheet of paper—AARON began to move.

The first images produced by AARON were stark, black-and-white abstract forms and rudimentary figures that resembled ancient petroglyphs or tribal art. To the modern eye, they look simple. But for 1973, they represented a seismic philosophical shift: a machine was autonomously executing a creative act without a human guiding the pen in real-time.

An Ongoing Dialogue

As computing power evolved over the next four decades, so did AARON. In the 1980s and 1990s, Cohen expanded the program’s capabilities, teaching it how to handle color selection and complex compositions. Later iterations of AARON graduated from pen-and-paper plotters to painting massive, vibrant jungle scenes populated by stylized human figures and lush botanical forms. These later works were exhibited in elite institutions like the Tate Modern and the Whitney Museum of American Art.

What makes AARON’s legacy so compelling is the unique partnership at its core. Cohen spent the rest of his life working alongside his software. He viewed AARON not as a tool, but as an intellectual collaborator—an independent entity that surprised him, challenged his assumptions, and forced him to rethink the boundaries of human creativity until Cohen’s passing in 2016.

Today, as we marvel at instant digital masterpieces rendered by cloud servers in seconds, it is worth remembering those quiet rooms in the 1970s. Long before the era of algorithms scrolling through billions of images, a lone painter sat down with a mainframe computer and watched a mechanical arm draw the very first ghost in the machine.