Long before modern diffusion models, prompt engineering, and sprawling neural networks dominated headlines, a quiet revolution in digital creativity took root in the early 1970s.
The true origin of AI-generated imagery traces back to 1973, when British-born painter Harold Cohen unveiled a software program that would fundamentally challenge our definitions of art, authorship, and machine intelligence: AARON.
From the Canvas to the Mainframe
In the late 1960s and early 1970s, artificial intelligence was largely confined to university laboratories, where computers were being taught to solve symbolic logic puzzles, prove mathematical theorems, and play chess. Creativity was widely viewed as an exclusively human domain—an untamable mixture of intuition, emotion, and “soul” that a machine could never replicate.
Harold Cohen, however, looked at the nascent computing power of the era and saw a different horizon. Crucially, Cohen was not a computer scientist who dabbled in art; he was an elite, internationally established painter who had represented Great Britain at the prestigious Venice Biennale.
Frustrated by the rigid boundaries of the traditional art world, Cohen took a sabbatical to the Stanford Artificial Intelligence Laboratory (and later the University of California, San Diego) to answer a radical question: Can the cognitive acts of drawing and composition be broken down into explicit rules?
Meet AARON: The Pioneer of Autonomous Art
By 1973, Cohen’s research culminated in AARON, an autonomous computer program designed to create original visual art.
Unlike modern AI image generators that rely on statistical machine learning, vast training datasets scraped from the internet, or prompt-based diffusion, AARON was built entirely on symbolic reasoning and cognitive rules.
- The Rule-Based Mind: Cohen painstakingly translated his own understanding of mark-making into code. He codified rules for spatial awareness, how lines intersect, how shapes enclose space, and how objects relate to a background.
- Autonomous Execution: Once Cohen hit “start,” he did not intervene in the drawing process. The computer made independent compositional choices about what to draw next, ensuring that no two outputs were ever identical.
- The Robotic Hand: To bridge the gap between software and physical reality, AARON was hooked up to a mechanical pen plotter. The program sent digital instructions that physically moved a pen across a sheet of paper, bringing code into the physical world.
The resulting 1973 outputs were abstract, intricate compositions of lines and closed forms. They did not attempt to mimic photographs or copy existing paintings; instead, they represented a machine executing its own internal logic to generate entirely novel visual structures.
A Forty-Year Partnership
Most technological proofs-of-concept in the 1970s were short-lived, but Cohen’s work with AARON evolved into a four-decade collaboration.
As computing power grew, so did AARON’s capabilities. Over the decades, Cohen updated the program’s rule sets, allowing it to transition from abstract geometry into complex, recognizable representations of plants, rocks, and eventually human figures. AARON’s artwork eventually found its way into major institutions worldwide, including the Tate Modern and the Whitney Museum of American Art.
Why 1973 Matters Today
Looking back from an era where stunning, hyper-realistic images can be generated in seconds via text prompts, it is easy to overlook the audacity of AARON’s 1973 debut.
Harold Cohen and AARON proved that creativity is not a magical, impenetrable black box. By demonstrating that a machine could autonomously generate aesthetic compositions governed by structured logic, Cohen laid the foundational philosophical and technical groundwork for every digital brushstroke, neural network, and generative algorithm that followed.
The first AI-generated image didn’t come from a server farm or a tech startup—it crawled slowly across a piece of paper in 1973, drawn by a mechanical pen responding to the mind of a painter who dared to ask what a computer could dream.
Leave a Reply