The Genesis of Machine Vision: How 1973 Gave Birth to the First AI-Generated Art
When we think of artificial intelligence and imagery today, we picture instant digital prompts, hyper-realistic photo simulations, and endless debates over copyright and creativity. It feels intensely modern—a product of the 2020s. Yet, the conceptual foundation of AI-generated art was laid over half a century ago.
The year was 1973. Long before neural networks, diffusion models, or consumer GPUs, a British painter named Harold Cohen achieved what many thought was impossible: he built a computer program that could create original artwork entirely on its own.
Meet AARON: The World’s First AI Artist
In the early 1970s, Harold Cohen was already an established, internationally acclaimed artist who had represented Britain at the prestigious Venice Biennale. However, his curiosity pulled him toward an unexpected frontier: computing. He took up a fellowship at the Stanford Artificial Intelligence Laboratory, diving headfirst into a world where scientists were just beginning to teach computers how to play chess and solve logic puzzles.
Cohen asked a radical question: If a machine can learn rules of logic, can it also learn the cognitive rules of art-making?
At the time, the mainstream art and tech worlds largely agreed that creativity was an exclusively human trait—a mystical domain driven by emotion and intuition that could never be reduced to code. Cohen set out to prove them wrong.
By 1973, his efforts culminated in a pioneering software program named AARON.
How the First “AI Image” Was Born
Unlike modern AI systems that scan billions of internet images to synthesize pixels via machine learning, AARON was born in an era of symbolic, rule-based AI.
- The Rulebook of Drawing: Cohen didn’t train AARON with a massive dataset. Instead, he manually translated his own deep understanding of visual cognition into code. He programmed rules governing how lines intersect, how shapes are bounded, how objects overlap, and how compositions achieve balance.
- True Autonomy: Once Cohen pressed start, he did not intervene. The program made its own aesthetic choices in real time—deciding where a line should go, when a shape was complete, and how the overall composition should breathe.
- The Mechanical Hand: Because computer monitors of 1973 could not display complex vector art gracefully, AARON’s decisions were translated through a mechanical pen plotter—a device that held a physical pen and moved it across a sheet of paper based on the computer’s instructions.
The resulting 1973 outputs were striking: intricate, abstract compositions of lines and closed forms that possessed a distinct internal logic. They were genuinely original creations, rendered by a machine executing autonomous artistic judgment.
A Forty-Year Partnership
What makes the story of AARON truly remarkable is that it wasn’t just a one-off science experiment. Cohen didn’t just write a program and walk away; he spent the next four decades of his life in continuous collaboration with his creation.
As computing power evolved, so did AARON. Over the years, Cohen updated the program, gradually teaching it about perspective, shading, and eventually, rich coloring and figurative shapes like plants and human figures. AARON’s evolutionary journey eventually graced the walls of world-class institutions like the Tate Gallery in London and the Whitney Museum of American Art.
Why 1973 Still Matters Today
When looking at the hyper-polished AI art landscape of the present day, it is easy to forget the humble, ink-on-paper beginnings of 1973.
AARON’s creation forced a profound philosophical shift. It challenged our definitions of authorship, inspiration, and what it means to be “creative.” Harold Cohen demonstrated that art is not merely an inexplicable stroke of magic, but a deeply structured cognitive process—one that a machine could learn to share.
Leave a Reply