When artist and tool designer Joel Simon created the website GanBreeder in November 2018, he programmed a network of computers to blend and “breed” together images over and over using users’ preferences as its guide. Although thousands of users, breeding millions of bizarre and beautiful images, Joel’s goal was more conceptual: He wanted to see if the system could evolve art and  what types of forms might emerge from the process.

Like nearly all of his other projects, GanBreeder was an experimental blending of computer programming, artistry, and the scientific concept of emergence—where a biological form is created not from a direct instructions, but emerges from a set of rules that can give rise to an endless myriad of novel forms. Whether he’s blooming digital sculptures of coral using genetic algorithms or producing floor plans for buildings using the same rules observed in ant colony or blood vessel growth, Joel’s goal is to leverage emergence for all its creative power.

But Joel has taken this one step further. His new and improved site, ArtBeeder, uses machine learning to create genetic markers in each image. Users can then “breed” new images based on the “genes” they want to highlight or combine two images together to produce a bizarre mixture. By harnessing the aesthetic choices of crowds, Joel has been delighted to witness a new form of artistic emergence.