What Was the First Ever AI Generated Image? History and Origin

The Pixel Pioneers: The Fascinating True Story of the First-Ever AI-Generated Image

We live in an era where typing a simple sentence into a prompt box can conjure a hyper-realistic photograph, a surreal oil painting, or a precise architectural rendering in mere seconds. Tools like Midjourney, DALL-E, and Stable Diffusion have made AI image generation a ubiquitous part of modern culture.

Yet, long before neural networks crawled the internet for training data or GPUs rendered millions of pixels, a lone artist and programmer sat in a university lab trying to teach a machine how to draw.

The story of the first-ever AI-generated image is not a tale of corporate tech labs or multi-billion-dollar models. It is a decades-long artistic experiment that began in the late 1960s—and its protagonist was a piece of software named AARON.

Meet AARON: The World’s First Robotic Artist

To find the birthplace of AI imagery, we have to travel back to 1973 and look toward Harold Cohen, a celebrated British painter and professor at the University of California, San Diego.

In the late 1960s, Cohen found himself grappling with a radical philosophical question: What is the core essence of drawing? If a human artist can translate a mental concept into physical marks on a canvas by following a set of cognitive rules, could a machine be taught to do the exact same thing?

Cohen wasn’t an engineer by training, but he taught himself programming. He began developing a computer program that would eventually be named AARON. Unlike modern AI, which learns statistically by looking at billions of existing images, AARON was built from scratch using symbolic artificial intelligence.

How the First AI Images Were Born

In its earliest iterations, AARON did not use a screen or a digital canvas. Instead, Cohen connected his computer program to a custom-built, turtle-like drawing machine that moved across a large sheet of paper on the floor, clutching a pen.

By the early 1970s (with breakthrough outputs solidified around 1973–1974), AARON successfully produced its first autonomous black-and-white line drawings.

• No Neural Networks: AARON didn’t “know” what a horse, a human, or a rock looked like based on photographs.
• Rule-Based Logic: Instead, Cohen manually coded cognitive rules of representation into the program. He taught AARON how to construct forms, manage spatial relationships, handle overlapping objects, and balance compositions.
• Autonomous Decision Making: Within the strict boundaries of those rules, AARON made its own real-time creative choices. The computer decided where to place lines, how to construct abstract figures, and how to fill a page.

In essence, Cohen provided the grammar of drawing, and AARON wrote the sentences.

From Floor-Crawling Turtles to Global Galleries

As computing power evolved over the decades, so did AARON. What started as primitive, abstract geometric line drawings eventually grew more complex. By the 1980s and 1990s, Cohen had taught AARON how to handle color, and later, to paint organic shapes, complex foliage, and stylized human figures.

AARON’s creations broke historical ground. They were exhibited in prestigious venues worldwide, including the Tate Modern in London and the Whitney Museum of American Art in New York. Harold Cohen continued to collaborate with and refine AARON until his passing in 2016, making it one of the longest-running artificial intelligence art projects in history.

Why the First AI Image Matters Today

Looking at AARON’s modest black-and-white lines next to today’s hyper-detailed, photorealistic AI generations can feel like comparing a bicycle to a rocket ship. Yet, AARON laid the fundamental philosophical and technical groundwork for everything that followed.