Long before the era of neural networks, text prompts, and instantaneous web-browser generators, the birth of artificial intelligence art took place in a quiet university lab. Decades before anyone could type a few words into a modern engine to conjure a digital painting, a British-born painter and computer scientist named Harold Cohen set out to answer a radical question: What is the minimum set of cognitive rules required to make a drawing?
The result of that inquiry was AARON—a software program whose early outputs in 1973 mark the true dawn of AI-generated imagery.
The Man Behind the Machine
Harold Cohen was not a programmer who dabbled in art; he was an accomplished, internationally recognized abstract painter who represented Britain at the prestigious Venice Biennale in 1966. At the height of his traditional career, Cohen accepted a visiting professorship at the University of California, San Diego, where he encountered early mainframe computers.
Intrigued rather than intimidated by the technology, Cohen wanted to understand the cognitive mechanics of human representation. He asked himself: How do we recognize a collection of lines on a flat page as a human figure or a plant, even when those lines bear very little physical resemblance to reality?
To crack the code, Cohen decided to step away from the canvas and teach a computer how to draw.
What 1973 Looked Like: The Genesis of AARON
Beginning his deep dive into programming in the late 1960s, Cohen moved to the Artificial Intelligence Laboratory at Stanford University in 1973. It was during this residency that AARON truly came into its own as an image-generating entity.
Unlike modern AI models that ingest billions of images scraped from the internet to train statistical weights, AARON was built entirely on symbolic, rule-based logic (known in computer science as GOFAI—Good Old-Fashioned AI).
- No Neural Networks: AARON didn’t “learn” by looking at thousands of photos. Instead, Cohen manually codified human cognitive understanding.
- Understanding Space and Form: Cohen wrote rules that taught the software how to structure a composition, balance positive and negative space, handle occlusion (how one object overlaps another), and map out proportions.
- The First Output: Operating on massive mainframe computers, AARON began producing distinct, primitive black-and-white line drawings. The program made autonomous decisions about composition, rendering figures, rocks, and foliage that had never been drawn before.
Because computer screens of the era were not yet optimized for rich graphic display, AARON’s digital calculations were often translated into physical reality via custom-built mechanical drawing machines—turtle-like robots connected to the computer that clutched pens and physical markers to drag across giant sheets of paper on the floor.
A Philosophical Milestone
The images produced in 1973 challenged the traditional art world’s definition of creativity. When people looked at an AARON drawing, they weren’t looking at a pre-programmed stamp or a random computer glitch; they were looking at a machine making autonomous choices within boundaries set by a human mind.
Cohen famously viewed his relationship with AARON not as a master and tool, but akin to a Renaissance master working with a studio assistant. Cohen provided the intellect and the rule-set, but AARON executed the creative labor.
As AARON evolved over the subsequent decades, it moved from black-and-white line work into complex color painting, eventually finding a permanent place in history with exhibitions at major institutions like the Tate Modern and the Whitney Museum of American Art.
Today, as AI image generators produce hyper-realistic art in fractions of a second, looking back at Harold Cohen’s 1973 breakthrough reminds us that the dream of machine creativity is more than half a century old—born not from corporate tech labs, but from the curious mind of a painter trying to decode his own hand.
Leave a Reply