The First Ever AI Image: History and Origins

The Pixel Before Time: The Story Behind the World’s First AI-Generated Image

Long before the internet was flooded with hyper-realistic portraits generated by text prompts, and decades before neural networks learned to dream in multi-layered diffusion, a quiet revolution took place in a university laboratory.

Today, we take generative artificial intelligence for granted. With a few keystrokes, anyone can conjure up digital masterpieces. But the story of the very first AI-generated image does not start in a tech startup or a Silicon Valley data center. It starts in the late 1960s and early 1970s with a painter who dared to ask a radical question: Can a machine learn to make art on its own?

Meet Harold Cohen and AARON

To find the genesis of AI imagery, we have to look at Harold Cohen, a celebrated British painter and professor. By the late 1960s, Cohen had achieved mainstream success, representing Britain at prestigious global art exhibitions. Yet, he grew deeply fascinated by the rising computing technology of the era. He wanted to understand the cognitive mechanics of human drawing—how the human brain translates abstract thoughts into lines, shapes, and spatial awareness on a flat surface.

Instead of just using a computer as a digital canvas, Cohen wanted the computer to think like an artist.

In 1968, while working at the University of California, San Diego, Cohen began writing a software program that would eventually change art history. He named his creation AARON.

By the early 1970s—particularly during a pivotal residency at the Stanford Artificial Intelligence Laboratory between 1973 and 1975—Cohen refined AARON into the world’s first autonomous art-making AI system.

How the First AI “Drew”

If you are imagining modern machine learning—where an algorithm sifts through billions of scraped internet photos to synthesize a new image—think again. AARON was built during the era of “Good Old-Fashioned AI” (GOFAI). There was no internet, no neural networks, and no deep learning.

Instead, Cohen painstakingly wrote symbolic rules into the code:

• Rules of geometry and physics: How lines connect to form shapes, and how objects overlap in three-dimensional space.
• Cognitive autonomy: Cohen didn’t pre-draw the images. Once he set AARON running, the program made its own composition choices. It decided where lines would go, how elements would balance, and what shapes to form without human intervention mid-process.

The output of early AARON consisted of primal, mesmerizing black-and-white line drawings. Often connected to a physical turtle-like drawing machine or plotter, AARON would autonomously drag a pen across paper, physically manifesting lines born entirely from code.

Seeing these early primitive sketches materialize stroke by stroke marked a watershed moment: It was the first time a machine had autonomously executed a creative act.

From Black-and-White Lines to a Lifelong Partnership

AARON didn’t stay static. Over the next four decades, Cohen continually updated his digital collaborator.

• In the 1980s and 1990s, AARON evolved to understand color, texture, and complex shading.
• Later iterations began populating its canvases with organic figures, plants, and intricate jungle-like scenes.

Cohen viewed AARON less as a tool and more as a companion or an independent student he was mentoring. Their decades-long partnership ended only with Cohen’s passing in 2016.

Why the First AI Image Still Matters Today

When museums like the Whitney Museum of American Art showcase historical exhibitions of AARON, they highlight a profound truth about modern technology: The quest to replicate human creativity is not a 21st-century invention.