Teaching a machine beauty

"Teaching a machine beauty" is the title of my master thesis for my master in Human Media Interaction. The thesis is about the design of a system that performs intelligent interactive evolution of abstract animations using intelligent search algorithms and machine learning.

It explores the possibilities of a system that generates its own animations, while minimizing the number of unpleasant animations it generates. A system like this could be used to have ever-changing art in an art installation, inspire artists with unique forms and combinations of animations, or as a tool for artists to find interesting animations for their work (for VJ's, for example). It raises the question if a machine can be creative, and where creativity comes from.

The designed system could create a vast diversity of animations, recognize unpleasant animations based on user feedback, and reject those animations before showing them to users. It uses nature-inspired concepts such as a genetic algorithm to generate the animations, and neural networks to recognize the unpleasant animations. The animations are rendered using OpenGL shaders. Some examples of animations generated by this system can be seen in the video below.