The Genesis of Synthetic Sight: Inside the Creation of the World’s First AI Image
When we think of artificial intelligence and imagery today, our minds instantly drift to hyper-realistic, hyper-fast text-to-image engines. With a simple sentence typed into a browser, we conjure floating castles, photorealistic portraits, or surrealist landscapes in mere seconds.
Yet, the story of artificial intelligence making its own visual art didn’t begin in a Silicon Valley tech startup or a modern cloud cluster. It began more than half a century ago, born from the mind of an artist who asked a radical question: Can a machine be taught to understand the act of drawing?
Long before neural networks and diffusion models, the world’s first true AI image was sketched out by a pioneering software program called AARON.
Meet Harold Cohen and AARON
In the late 1960s, British-born artist Harold Cohen was enjoying an established career. A respected painter who had represented his country at the prestigious Venice Biennale, Cohen made a jarring pivot in 1968. He moved to the United States—eventually landing at the University of California, San Diego—and threw himself into the alien world of computer science.
Cohen wasn’t interested in simply using a computer as a digital paintbrush, nor did he want to punch coordinates into a machine to plot rigid, geometric computer art (which was popular at the time). Instead, he wanted to explore whether a machine could possess autonomous creativity.
He began writing a computer program that would later be named AARON.
Unlike modern AI that relies on scraping billions of images from the internet, AARON possessed zero database of human-made pictures. It didn’t “learn” by looking at millions of photographs. Instead, Cohen imbued AARON with a set of symbolic rules about cognitive processes, spatial reasoning, and physical boundaries. He codified human intuitions about how lines relate to shapes, how objects overlap, and how figures are structured.
The First “AI Images” Come to Life
Throughout the 1970s—with a monumental stint at the Stanford Artificial Intelligence Laboratory—AARON evolved.
The earliest iterations of AARON’s output weren’t displayed on glowing computer monitors. The code executed commands that drove custom-built, turtle-like drawing machines resting on large sheets of paper spread across the floor. Holding a pen, the mechanical plotter would trace lines autonomously, guided entirely by the program’s real-time decisions.
The resulting images—featuring abstract organic shapes, primitive human figures, and botanical motifs—were groundbreaking. When audiences viewed these works at venues like the Los Angeles County Museum of Art and later at the Whitney Museum, they were witnessing something unprecedented: visual art generated not by human hand-eye coordination, but by a software system making independent decisions.
While Cohen remained the programmer, the aesthetic choices on the page belonged entirely to the logic of the machine.
From Mechanical Plotters to Modern Marvels
Looking back at AARON’s primitive, black-and-white line drawings provides a humbling contrast to today’s multi-billion-parameter models. AARON did not understand “cats” or “space travel” via semantic text prompts; it understood abstract rules of construction, letting organic compositions emerge line by line. Cohen continued developing and expanding AARON for over 40 years, right up until his passing in 2016, watching the software transition from rudimentary line work into rich, color-painted outputs.
Leave a Reply