Long before modern diffusion models could conjure a hyper-realistic photorealistic portrait from a simple text prompt in seconds, the genesis of machine-made art was born from an entirely different philosophy. The story of the first-ever AI-generated image is not one of neural networks scraping billions of internet images, but of a celebrated British painter who asked a radical question: Can a machine be taught to understand the cognitive rules of art?
The Artist Who Walked Away from Success
In the late 1960s, Harold Cohen was at the peak of the contemporary art world. A brilliant painter trained at London’s prestigious Slade School of Fine Art, Cohen had achieved international acclaim, even representing Great Britain at the prestigious Venice Biennale.
Yet, at the height of his traditional success, Cohen grew restless. He became obsessed with a fundamental philosophical puzzle: What actually makes an image evocative? Why do a few simple charcoal lines on a page register in the human brain as a figure, a rock, or a tree?
Convinced that representation was rooted in cognitive rules rather than mere physical copying, Cohen made a staggering choice. In 1968, he left his thriving career behind, moved to the University of California, San Diego, and taught himself how to write computer code. He wanted to test whether a machine could internalize the grammar of drawing and independently generate original artwork.
Enter AARON: The Pioneer Program
By the early 1970s, Cohen had built a computer program that would change art history forever: AARON.
Unlike modern generative AI, which relies on statistical probability, deep learning, and vast datasets scraped from the web, AARON was entirely rule-based. Cohen did not give AARON a pile of pictures to mimic. Instead, he systematically deconstructed his own understanding of human cognition and drawing. He manually programmed rules about:
- Geometry and spatial awareness: How objects relate to one another in space.
- Overlap and occlusion: Understanding that a solid object blocks what is behind it.
- Proportion and scale: How human figures and organic shapes are constructed structurally.
Using a custom-built drawing machine—often a mechanical “turtle” controlled by a computer that squeezed ink out onto paper—AARON began to output physical, autonomous drawings. The computer was given the freedom to make composition and design decisions entirely on its own, meaning no two outputs were ever identical.
The First Images Come to Life
The earliest iterations of AARON in the early 1970s produced stark, abstract black-and-white line drawings reminiscent of primitive or abstract art. However, a major milestone arrived in 1973, when AARON produced its first groundbreaking hand-colored computer-generated drawings (such as Cyc 3), where the program mapped out the composition and lines, which were then filled with color.
When these early pieces were exhibited at major venues like the San Francisco Museum of Modern Art and later the Tate Modern and the Whitney Museum of American Art, they caused a profound cognitive dissonance for viewers. People looked at complex, organic figures and rich compositions, only to realize no human hand had planned the strokes. A machine was acting as an independent creator.
A Legacy That Echoes Today
Harold Cohen continued to evolve AARON over four decades until his passing in 2016, upgrading it from simple line drawings to complex botanical jungles and full-color painting assistants.
Looking back, AARON’s primitive black-and-white lines represent the true Cambrian explosion of artificial intelligence in creativity. Long before algorithms were trained to mimic Rembrandt or generate hyper-detailed sci-fi landscapes, Harold Cohen and his digital collaborator proved that a machine could do more than just crunch numbers—it could participate in the uniquely human act of making meaning.
Leave a Reply