We've seen games that can teach regular humans how to program their own video game titles, but why not avoid the middle man? That's the idea behind Angelina, an AI designed to intelligently build playable games without human intervention. Designed by Michael Cook, a PhD student at Imperial College London, Angelina designs different aspects of games — level layout, enemy behavior, bonuses, and the like — and assembles them randomly. The system then simulates playing the level, determining what the most effective variants are, and repeats the process around 400 times, pushing the most successful elements forward in a sort of game-design Darwinism.

The games are simple 8-bit throwbacks at the moment, and while Angelina can design levels and internal logic, Cook still needs to provide the graphics and audio assets. Still, it's hard to not get excited about the possibilities of a system like Angelina, and with Cook envisioning it as a tool that can aid human designers in fine-tuning their own creations, we wouldn't be surprised to see some fascinating collaborations come out of the project. To try out Angelina's games yourself, you can head on over to the Games By Angelina website.