What Was the First AI-Generated Image in History?

The Genesis of Synthetic Canvas: What Was the First AI-Generated Image?

When we think of AI-generated images today, our minds instantly jump to the frantic pace of modern diffusion models. We picture typing a quick text prompt into a browser—“a cybernetic cat sipping coffee on Mars in the style of oil painting”—and watching a hyper-realistic masterpiece materialize in seconds.

However, the lineage of artificial intelligence art stretches back much further than neural networks and GPUs. Long before algorithms scraped the internet for pixels, the very first AI-generated image was born out of an entirely different kind of ambition: an artist trying to codify the human act of drawing itself.

Meet AARON: The World’s First AI Artist

The title of the first true AI-generated image creator belongs to a software program called AARON, developed beginning in the late 1960s and early 1970s by Harold Cohen, a prominent British painter and professor at the University of California, San Diego.

In 1968, Cohen left behind a successful traditional painting career to explore a radical question: What is the minimum set of cognitive rules required to make a drawing? He wasn’t interested in simply having a computer shuffle random pixels or plot geometric shapes (which early computer programmers were doing with basic plotters). Instead, Cohen wanted to build a machine that understood the conceptual rules of art.

By the early 1970s—with a crucial developmental stint at the Stanford Artificial Intelligence Laboratory between 1973 and 1975—Cohen successfully brought AARON to life. The resulting black-and-white line drawings produced by the system are widely recognized as the earliest instances of true artificial intelligence art.

How Did AARON Create Images Without Deep Learning?

Unlike today’s AI, which learns statistically from billions of existing images using massive neural networks, AARON operated entirely on symbolic, rule-based AI (often referred to in computer science history as GOFAI—”Good Old-Fashioned AI”).

• Hand-Coded Logic: Cohen did not feed AARON a dataset of other artists’ paintings. Instead, he manually translated his own understanding of cognitive psychology, spatial reasoning, and visual art into computer code.
• Autonomous Decision-Making: AARON was given rules on how to handle composition, occlusion (how objects overlap), perspective, and proportions.
• The Robotic Arm: Once the program made a decision on what to draw, it sent commands to a physical drawing machine—a turtle-like robotic device holding a pen—which physically rendered the lines onto paper on the floor.

In its earliest iterations, AARON generated abstract forms and simple line art. As Cohen continuously updated the program over the decades, AARON evolved to understand organic shapes, complex figures, and eventually botanical and jungle-like scenes. Later versions were even programmed to apply color autonomously, though Cohen still mixed the physical paints according to the program’s color choices.

From Code-Driven Lines to Modern Diffusion

Looking back, AARON’s primitive black-and-white drawings stand in stark contrast to the cinematic outputs of modern systems like Midjourney or DALL-E. Yet, the core philosophical breakthrough remains identical: the machine was the author of the composition. AARON didn’t just execute a pre-written vector file; it made spontaneous choices on every single sheet of paper it drew on, ensuring no two pieces were ever identical.

Harold Cohen continued developing and collaborating with AARON for over four decades until his passing in 2016. Today, his pioneering work is celebrated in major institutions like the Whitney Museum of American Art, reminding us that the journey toward machine creativity began not with a high-tech corporate lab, but with a painter asking profound questions about the nature of human expression.