Early AI Pictures: Evolution of Generative Art

Pixels of the Past: The Wild, Weird, and Wonderful Origins of Early AI Pictures

Today, typing a few words into a prompt box can instantly generate a hyper-realistic masterpiece. But long before high-definition diffusion models and viral text-to-image generators dominated our feeds, early computer artists and researchers were laying the groundwork with glowing green code, clunky plotters, and some seriously trippy neural network experiments.

The story of early AI pictures is a fascinating journey from rule-bound logic to machine hallucinations.

1. The Rule-Bound Pioneers (1970s–1980s)

Before “deep learning” or massive datasets existed, creating art with a computer required writing rigorous mathematical rules.

• Meet AARON (1970): Developed by British artist and programmer Harold Cohen, AARON is widely considered the grandfather of AI art. AARON wasn’t a neural network; it was a physical machine hooked up to a robotic drawing arm. Cohen wrote thousands of lines of code telling the machine how to draw shapes, botanical figures, and lines on paper.
• The Vibe: Black-and-white, precise line drawings. While AARON didn’t “think” or learn from data, it proved a revolutionary concept: a machine could autonomously execute creative intent.

2. The Psychedelic Era: Google DeepDream (2015)

If AARON was the quiet academic beginning, DeepDream was the loud, bizarre explosion that introduced AI art to the mainstream internet.

Originally built by Google engineer Alexander Mordvintsev to help scientists see how computer vision networks “think,” DeepDream took an image and ran it backward. Instead of asking the network to identify a dog in a photo, engineers told it to amplify whatever dog patterns or shapes it vaguely recognized.

• The Vibe: Absolute digital fever dreams. Ordinary photos of clouds, buildings, or landscapes morphed into swirling, hallucinatory nightmares packed with endless repeating dog eyes, multi-legged creatures, and pagoda-like architecture. It was weird, mesmerizing, and the first time millions of people realized AI had an imagination of its own.

3. The Rivalry That Changed Everything: GANs (2014)

The real leap toward photorealism kicked off with the invention of GANs (Generative Adversarial Networks) by researcher Ian Goodfellow in 2014.

GANs introduced a brilliant psychological mind game between two different AI models:

1. The Generator: An artist trying to forge realistic fake images.
2. The Discriminator: An art critic trying to spot the fakes.

Every time the generator failed, it learned from its mistakes. Through this constant, high-speed competition, the generator got frighteningly good at creating entirely original faces, objects, and landscapes that looked completely real—even though the people and places depicted never existed.

Why Early AI Art Still Matters

Looking back at early AI pictures reminds us that technology doesn’t evolve in a vacuum. From Harold Cohen’s robotic drawing arm to Google’s puppy-eyed hallucinations, every strange, glitchy, and pixelated step forward was a building block.