A Brief History of Early AI Pictures and Generative Art

Beyond the Prompt: A Look Back at the Fascinating Era of Early AI Pictures

Today, we take hyper-realistic, text-to-image wonders for granted. With a few typed words, modern neural networks can conjure cinematic landscapes, photorealistic portraits, and intricate digital paintings in seconds.

However, the intersection of artificial intelligence and visual art didn’t begin with modern diffusion models or deep learning giants. Long before algorithms could paint like Rembrandt or dream up surrealist creatures, a small group of visionary artists and computer scientists were laying the groundwork.

The story of early AI pictures is a tale of clunky mainframes, mechanical turtles, pure logic, and a radical reimagining of what creativity actually means.

The 1960s: Plotters, Punch Cards, and Mathematical Beauty

The genesis of algorithmic imagery traces back to the early 1960s, when computers were room-sized behemoths fueled by punch cards and magnetic tape. Far from being tools for entertainment, these machines belonged to universities, government labs, and corporations.

A handful of pioneers—such as mathematicians and artists like Frieder Nake, Georg Nees, and A. Michael Noll—realized that the precise calculations driving these machines could also dictate spatial coordinates. Using high-speed plotters (mechanical devices holding actual ink pens that moved back and forth across large sheets of paper based on computer code), they began generating geometric compositions.

• The Medium: These early works were stark—usually black ink on white paper.
• The Method: They relied on stochastic (randomized) processes, loops, and mathematical formulas. Computers weren’t “learning” yet; instead, humans wrote strict rule-sets, allowing the machine to introduce controlled variations.
• The Cultural Impact: Exhibitions of “computer art” in galleries in Stuttgart, New York, and Zagreb sparked fierce debates. Critics scoffed at the idea that a machine could produce “art,” while proponents argued that code was simply a new kind of paintbrush.

The 1970s and 1980s: Enter AARON, the Autonomous Artist

While 1960s computer art was mostly abstract and geometric, the true ancestor of modern AI image generation emerged in the 1970s: AARON.

Developed by British-born artist Harold Cohen starting in 1968 at the University of California, San Diego, AARON was a pioneering software program designed to autonomously create original artwork. Unlike simple pattern generators, Cohen didn’t just write code to draw a single predetermined image; he sought to encode cognition and rules of representation.

• How AARON Worked: Cohen spent decades teaching his program rules about objects, human figures, perspective, and shading. AARON “understood” how objects overlap, how plants grow, and how light falls, writing its own compositions on the fly.
• The Output: Connected to custom drawing machines (and later painting arms), AARON would physically ink or paint its own creations on paper. By the 1980s, AARON possessed a rudimentary form of artificial intelligence—making independent stylistic choices about color palettes and composition without human intervention mid-process.
• The Philosophy: Cohen famously remarked that AARON was an autonomous entity. He wasn’t using the computer as a tool; rather, he was collaborating with an artificial mind that he had birthed.

The 1990s to 2010s: Neural Networks and DeepDreaming

As computing power grew exponentially through the late 20th century, early AI art transitioned from symbolic rule-based logic (like AARON) to early machine learning models.

• The Shift to Neural Networks: Researchers began experimenting with genetic algorithms and early neural networks—systems inspired by biological brains that could “learn” patterns from large datasets rather than relying solely on hard-coded human rules.
• Google DeepDream (2015): A massive cultural milestone arrived when Google released DeepDream. Utilizing a convolutional neural network designed for image recognition, DeepDream inverted the process. Instead of asking the computer to identify what was in a picture, researchers asked it to enhance what it thought it saw. The result? Psychedelic, nightmarish, dream-like landscapes swirling with eyes, dogs, and strange architectural structures. It was the public’s first widespread glimpse into how computers “perceived” visual data.

Why Early AI Pictures Matter Today

Looking back at the plotter drawings of the 1960s and the mechanical paintings of AARON reveals a profound truth: the desire to co-create with machines is deeply rooted in human history.

Today’s generative models stand on the shoulders of these digital pioneers. While early AI art lacked the hyper-realism of modern tools, it possessed a raw, conceptual purity. Those primitive lines drawn by clicking plotters and whirring mainframe drives weren’t just technical experiments—they were the first whispers of a machine conversation that is still reshaping our world today.