What Was the First AI-Generated Image in History?

The Pixelated Ancestor: Unraveling the True Origin of the First AI-Generated Image

Long before a single prompt could summon a photorealistic portrait or a whimsical digital landscape in seconds, a quiet revolution was taking place in a university laboratory. Today, we live in an era where generative artificial intelligence dominates our screens, turning text boxes into digital masterpieces. But where did it all begin?

To find the very first AI-generated image, we have to travel back not to a modern tech giant’s server farm, but to the early 1970s—long before deep learning, neural networks, or the internet as we know it existed.

Meet AARON: The Artist in the Machine

When art historians and computer scientists look for the genesis of AI imagery, the consensus universally points to a program called AARON.

Developed starting in the late 1960s by Harold Cohen—a celebrated British painter who represented his home country at the prestigious Venice Biennale before pivoting to computer science—AARON was an unprecedented experiment. In 1973, while working as a visiting scholar at Stanford University’s Artificial Intelligence Laboratory, Cohen achieved a major milestone: his software autonomously produced original, abstract line drawings that no human hand had explicitly mapped out.

“AARON did not use machine learning or crawl through massive datasets of existing art. Instead, it was built on symbolic artificial intelligence—manually hand-coded with human rules of composition, perspective, and geometry.”

Cohen essentially sought to externalize his own cognitive process as a painter. He taught AARON the fundamental rules of drawing: how to distinguish figure from ground, how to overlap shapes, how to judge proportions, and how to compose lines so they felt balanced. Connected to a physical drawing machine (a robotic plotter holding a pen), AARON would sit and autonomously draft intricate, geometric, and abstract compositions on paper without human intervention.

How AARON Changed the Definition of “Art”

AARON’s early outputs were primitive by today’s standards—mostly striking, black-and-white abstract line art. Yet, they crossed a monumental psychological and technological boundary.

Before AARON, computers were used as blunt tools to execute rigid, predetermined geometric grids (as seen in early computer graphics experiments by pioneers like Georg Nees or Frieder Nake). AARON, however, possessed a degree of autonomy. While it followed Cohen’s encoded rules, it made its own real-time compositional choices about where to place shapes and how to balance the canvas.

As the decades progressed, Cohen continuously upgraded AARON. By the 1980s and 1990s, the software evolved to handle color, and eventually, it began generating complex organic shapes, including stylized human figures, rocks, and lush jungle-like botanical environments. AARON’s works eventually found their way into elite institutions worldwide, including the Tate Modern and the Whitney Museum of American Art.

The Evolution: From 1973 to the Modern Era

AARON laid the foundational philosophical groundwork, but the technology behind AI imagery went through massive evolutionary phases before arriving at modern text-to-image generators:

• The Deep Learning Breakthrough (2015): Google’s DeepDream introduced convolutional neural networks to the art world, creating psychedelic, dream-like imagery by turning image-recognition software upside down to amplify patterns it “thought” it saw.
• The GAN Revolution (2018): Generative Adversarial Networks (GANs) allowed machines to learn directly from massive datasets. This era birthed works like the Portrait of Edmond de Belamy, which made history by selling at Christie’s auction house for over $432,000.
• The Modern Diffusion Era: Today’s state-of-the-art models discard rigid rule-books and adversarial nets entirely, utilizing diffusion and transformer architectures to translate natural language prompts into hyper-realistic photos, vector art, and 3D assets in seconds.

Looking Back to Move Forward

When we marvel at the lightning-fast capabilities of contemporary creative tools, it is easy to forget that the journey began with a mechanical pen, a reel of tape, and an artist daring to ask whether a machine could learn to think like a painter. Harold Cohen and AARON proved that code could carry a creative spark—transforming the computer from a simple calculator into a canvas.