The Pixel Before Time: The Story of AARON and the First AI-Generated Images
We live in an era where generating a breathtaking, hyper-realistic digital painting takes nothing more than a casual sentence typed into a chat box. Within seconds, modern text-to-image models churn out masterpieces built on billions of parameters and vast oceans of scraped internet data.
Yet, long before neural networks, diffusion models, and GPUs, a lone artist sat in front of a computer in the late 1960s, trying to answer a radical question: Can a machine learn to make art?
The answer came in the form of a pioneering program named AARON, and it gave birth to what many consider the world’s first true AI-generated imagery.
Meet Harold Cohen and AARON
AARON was not conceived in a tech startup or a high-tech corporate lab. It was born in the mind of Harold Cohen, a successful British-born painter who represented Great Britain at the prestigious Venice Biennale in 1966. At the height of his traditional art career, Cohen grew fascinated by the underlying cognitive processes of painting. What actually happens when a human hand translates an internal concept onto a physical canvas?
In 1968, Cohen moved to the United States to teach at the University of California, San Diego. Crucially, he taught himself computer programming. He didn’t want to use the computer merely as a digital paintbrush; he wanted the computer to make the creative decisions.
By the early 1970s—specifically crystallizing around 1973—Cohen had written the foundational code for AARON.
How the First AI “Thought”
Unlike today’s deep learning systems, which ingest millions of images to mimic styles statistically, AARON operated entirely on symbolic, rule-based logic (what was then called Good Old-Fashioned AI, or GOFAI).
Because computers at the time lacked the memory and processing power to handle pixels or photographs, AARON was built to think in terms of topology and geometry. Cohen manually coded rules of art into the system:
- How objects overlap in space.
- How proportions balance out.
- How to draw basic organic shapes, figures, and foliage.
AARON would run its code, make autonomous choices about composition and line weight, and then send instructions to a custom-built, floor-standing mechanical drawing machine (a “turtle” plotter). Using a physical pen clamped into a robotic arm, the machine would meticulously crawl across a sheet of paper, dragging ink lines to manifest an image that no human had pre-designed or sketched out beforehand.
The machine was executing a program, but the composition was genuinely emergent—decided entirely by the software’s internal logic.
From Crude Lines to Museum Walls
The earliest iterations of AARON generated stark, abstract, black-and-white line drawings. To modern eyes, they might look like primitive vector graphics or children’s coloring book outlines. But in the context of the 1970s, they were revolutionary. They proved that a machine could possess an “understanding” of spatial relationships and artistic primitives.
As decades progressed, Cohen continuously updated AARON. He taught the system how to handle color separation, later giving it brushes and paint to fill in its own creations. Over time, AARON evolved from drawing abstract patterns to rendering complex organic figures, plants, and human silhouettes, eventually being exhibited at world-class venues like the Tate Modern and the Whitney Museum of American Art.
The Legacy of the First AI Image
When we marvel at the lightning-fast capabilities of modern generative AI, it is easy to forget the philosophical leap taken by Harold Cohen over fifty years ago.
AARON didn’t rely on stealing styles or processing petabytes of data; it was an exercise in codifying human creativity itself. The early ink drawings scratched out by Cohen’s robotic plotter were the first faint footsteps of a long journey—reminding us that the desire to merge machine logic with artistic imagination is far older than the internet age.
Leave a Reply