Before the Prompts: The Untold Story of 1973 and the Birth of AI Art
Long before the era of text-to-image prompts, server farms, and heated debates over neural network training data, a quiet revolution took place in a university laboratory. The year was 1973, and the art world was comfortably analog. Abstract expressionism, minimalism, and conceptual art reigned supreme.
It was into this world that Harold Cohen, a celebrated British painter who had represented his country at the prestigious Venice Biennale, walked away from traditional canvases to stare into the humming, room-sized depths of a computer.
Cohen’s radical endeavor birthed AARON, widely recognized as the world’s first artificial intelligence art-generating system. Far from being a mere party trick, the creation of AARON in 1973 marks the true genesis of a technological and artistic shift that continues to shape our world today.
The Man Who Taught a Machine to Draw
By the late 1960s, Cohen had relocated to the University of California, San Diego, gripped by a profound, almost philosophical question: What are the absolute minimum conditions under which a set of marks on a surface functions as an image?
To answer it, Cohen realized he couldn’t just use a computer as a digital paintbrush. He had to figure out if a machine could independently make creative decisions.
The result of this obsession was AARON. Unlike modern generative systems that digest billions of scraped internet images via deep learning, AARON was built on symbolic, rules-based AI.
Because computers in 1973 lacked graphical user interfaces, deep neural networks, and massive parallel processing, Cohen had to act as AARON’s cognitive architect. He painstakingly encoded foundational human understandings of the visual world into code, including:
- Cognitive Rules: Understanding what constitutes a foreground versus a background.
- Structural Rules: The basic anatomy of human figures and the geometry of plant growth.
- Compositional Rules: How objects balance, overlap, and occupy a framed space.
When AARON generated its early black-and-white line drawings in 1973, it wasn’t copying a human sketch or stitching together a collage. It was computing a composition in real-time, executing code to decide where a line should travel next.
A Partnership Between Creator and Code
What made AARON a true milestone of artificial intelligence—rather than just a tool for automated printing—was its autonomy.
Cohen frequently noted that he could never entirely predict what AARON’s output would look like. While he provided the rules, the program made its own structural choices within those boundaries. In essence, Cohen discovered forms and compositions that he had never imagined, brought to life by a machine running his own logic framework.
“The machine was showing him possibilities within his own artistic system that he hadn’t seen.”
Over the decades, AARON evolved dramatically. What started as primitive 1973 line drawings eventually grew into complex, richly colored depictions of plants, rooms, and human figures. AARON’s work eventually graduated from university labs to major global institutions, including the Tate Modern and the Whitney Museum of American Art.
Echoes of 1973 in the Modern Era
Looking back from today’s landscape of instant, hyper-realistic, text-to-image generation, Harold Cohen’s 1973 experiment might look quaint. There were no GPUs, no diffusion models, and no internet prompts. AARON drew with a physical pen clamped into a mechanical drawing machine commanded by code.
Leave a Reply