jevai.art

Learn

A short history of generative art

Generative art is older than you might think. Long before neural networks, artists were writing rules and letting a process, a plotter or a program carry them out.

Generative art is art made by a system that the artist designs but doesn't fully control. The artist writes the rules; the rules, plus some chance, make the work. That definition doesn't mention computers at all, and for good reason: the idea came first.

Rules before machines

Composers played with chance and procedure for centuries, and twentieth-century artists made it explicit. Conceptual artists in the 1960s began issuing artworks as instructions: a wall drawing, for example, could be a short text describing lines to draw, executed by other people, slightly differently each time. The art was the rule; every execution was one instance of it.

1965: the first computer art

When computers and pen plotters became available to researchers, a few people pointed them at art. In 1965, the first exhibitions of computer-generated drawings took place, by mathematicians and engineers including Georg Nees, Frieder Nake and A. Michael Noll. Their pieces were grids of shapes nudged by random numbers, lines that wandered by rule, and careful imitations of famous paintings made to ask whether a machine could be an artist.

Around the same time, Vera Molnár, a painter, started using computers to explore systematic variations of squares and lines that she had been drawing by hand for years. Her work is one of the clearest examples of the generative idea: one rule, many disciplined deviations.

The program as collaborator

In the 1970s, the painter Harold Cohen began building AARON, a program that made drawings autonomously, and developed it for decades. AARON pushed the question further: if the program makes choices, whose art is it? Cohen's answer was that the art was in building a system capable of making interesting choices.

Code for everyone

For a long time, generative art required programming skills and access to unusual hardware. That changed in 2001 with Processing, a language and environment designed for artists and designers, created by Casey Reas and Ben Fry. A generation of people learned to code by drawing with it. Today, its descendants run in every browser.

Shaders: art at sixty frames a second

Graphics cards made a different kind of generative art possible: shaders, tiny programs that compute the color of every pixel in parallel, many times a second. Communities like Shadertoy, launched in 2013, turned shader writing into an art form of its own, with entire landscapes, oceans and creatures written in a few hundred lines of code and rendered live. jevai.art's avatars are shaders. If you're curious how they work, read shader art for beginners.

Where AI fits

Recent image models changed the conversation, but they work differently from classic generative systems: they produce a finished picture directly from a prompt, and the rules are learned rather than written. jevai.art goes back to the older tradition with one new ingredient. The system is a hand-written shader with explicit rules, and an AI model doesn't draw anything. It plays a part that used to belong to the artist or to chance: deciding, with measurable confidence, which way each rule should go for a particular person. Here's how that works.

What makes generative art good

Across sixty years, the same three ingredients keep appearing:

Tell Jev something trueMake your avatar →