The First Ever AI Image: History, Creators, and Evolution

The Pixel Pioneers: The Fascinating Story Behind the World’s First AI-Generated Images

Today, generating a breathtaking digital painting, a hyper-realistic portrait, or a surreal landscape takes mere seconds. You type a prompt into an app, and a sophisticated artificial intelligence model conjures a custom visual out of thin air.

However, the journey of AI-generated art didn’t begin with powerful graphics cards, massive training datasets, or neural networks. It started in a quiet university lab in the late 1960s, born from the mind of a British painter who wanted to understand the very nature of human creativity.

Meet Harold Cohen and AARON

In 1968, Harold Cohen was a successful, internationally recognized artist who represented Great Britain at the prestigious Venice Biennale. Yet, at the height of his traditional career, Cohen found himself consumed by a radical question: What makes a drawing a drawing?

He wondered if the fundamental rules of visual creation could be broken down into code—not just to make a computer execute a predetermined pattern, but to let a machine make its own creative decisions.

To answer this, Cohen packed up his brushes, moved to the United States, and taught himself computer science. By 1973, he had birthed AARON, widely recognized as the world’s first AI-powered art system.

How AARON Created Art Without “Learning”

If you compare AARON to modern text-to-image generators, the differences are staggering. Today’s AI models look at billions of images scraped from the internet, using deep learning and statistical probabilities to guess what a “cat in space” looks like.

AARON had none of that. There was no internet, no massive dataset, and no neural network. Instead, AARON operated entirely on symbolic logic and handcrafted rules.

• The Rules of Seeing: Cohen manually programmed rules of human cognition, geometry, and composition into the software. He taught AARON how objects overlap, how light creates depth, and how to balance proportions.
• Autonomous Choices: Once the program ran, Cohen didn’t touch the steering wheel. AARON decided on its own where lines would go, how shapes would interact, and what the final composition would look like.
• Physical Output: In its early days, AARON didn’t output to a computer screen. It was hooked up to a custom mechanical drawing machine—a robotic turtle with a pen strapped to its belly—that slowly crawled across giant sheets of paper on the floor, meticulously plotting out its lines.

From Simple Lines to Complex Ecosystems

AARON’s earliest outputs in the 1970s were stark, beautiful black-and-white line drawings featuring abstract shapes, rocks, and rudimentary human figures.

As computing power grew and Cohen’s understanding deepened over decades, AARON evolved:

1. The Color Era: Cohen eventually taught AARON how to use color, though the computer still relied on Cohen to physically mix the paints and fill in the shapes according to the program’s color-selection logic.
2. Organic Forms: In later decades, AARON shifted from geometric figures to lush, jungle-like botanical scenes filled with stylized plants and human silhouettes, eventually being exhibited at world-class venues like the Tate Modern and the Whitney Museum of American Art.

The Legacy of the First AI Artist

Harold Cohen continued working on and refining AARON until his passing in 2016, spending nearly 40 years side-by-side with his digital creation.

AARON proved a profound point long before the artificial intelligence boom: machines could participate in the creative process. While AARON didn’t “learn” from data the way modern models do, it demonstrated true procedural creativity—generating original visual works that had never existed before.

The next time you prompt an AI to create a digital masterpiece, spare a thought for AARON and the robotic turtle whirring away on a studio floor in the 1970s, proving that the intersection of art and code was possible all along.