Tinkering With Art And Locking The Gains

In generative art, accidents are sometimes celebrated. One misplaced variable, one typo, accidentally randomizing a variable: these accidents can sometimes produce new aesthetic.

For example: here’s a piece of art that I’ve been experimenting with recently:

generative art 1

The basic idea: a sphere in the middle, thousands of tiny lines surround it. The wavy lines are made with Perlin noise. By accident, I changed the noise value to a random value—noise value is wavy, random value is “jagged.” Here’s the result:

generative art 2

A small difference in the code, but the aesthetic is wildly different.

The goal of generative art is to create good visual aesthetic, and it’s achieved mostly by meandering, exploring, and tinkering. This is the game of tinkering and “locking the gains.” Out of ten experiments, nine might fail, but the one that succeeds might be really, really good.

To be more precise: The goal of generative art is to create good visual aesthetic with the most variability.

Suppose I have a system that makes a black-colored square in the middle of the screen, with a certain fixed size. This system has no variability: if I make it output a hundred JPEGs, it will output the same squares a hundred times. Not cool. I can inject variability to the extreme: random shape, sized randomly, random amount, placed randomly, with random colors. If I make this highly-random system output a hundred JPEGs, it will output random garbage—high variability, zero aesthetic.

Gain is when both aesthetic and variability increases. It’s mostly found by accident. Locking those gains is as easy as running git commit.

After lots of iterations and happy accidents, here’s one of the most recent results of the sphere-in-the-middle system:

generative art 3

Perhaps generative art should be called accidental art instead.