artificial life

Chapter 7: The Genetic Images Application

7.1 Introduction

Beauty is an attribute of art and other objects that we have a hard time defining and quantifying. Certainly, it is possible to say that certain combinations of lines or colors will produce a work that is more aesthetically pleasing than another combination, or that adding symmetry to a piece will balance it and make it a more interesting piece. In general, however, it very difficult to say determine whether or not a piece will be received as "beautiful" or "interesting" without having a human actually look at the piece and give their opinion. Due to this lack of a concrete definition of beauty it is very difficult to program computers to generate art that is pleasing to the eye.

This is not to say that computers are incapable of being programmed to create art. Everyday we encounter representation of art on computers, on graphically rich web sites or in the special effects of the latest action movie extravaganza. However, these are instances of human artists controlling the definition of this art to create beauty.

Genetic Algorithms and Genetic Programming show that it is possible to use techniques which emulate the natural process of evolution to search a very small portion of very large search spaces and still find satisfactory solutions for the stated problem. In Genetic Programming the space of possible solutions is actually infinite, since there is no limit on how deep a function tree can become. Despite this incredible size Genetic Programs are able to find individuals in the space that are adequate for the given task in a relatively short time.

The Genetic Images Application implemented for this project combines the power of Genetic Programming with the human ability to recognize beauty to create computer-generated images that are beautiful and interesting. The application's concept is based upon the work of Karl Sims [Sims, 1991].