Putting Artificial Life to Work

The ability to genetically breed computer programs is a relatively new approach to problem solving. Recent advances in parallel processing have provided platforms to efficiently carry out this type of computation. In this paper we describe a genetically optimized artificial lifeform. The goal of this work is to breed swarms of artificial organisms that can move over corrupted bitmapped images and correct any errors that are found. We will show that the genetically optimized lifeforms can correct most instances of some forms of errors. In addition, we will show that ability of the organisms to communicate can improve the fitness of an organism. To efficiently carry out this task, an organism exploits two different forms of parallelism. The first form is an implementation specific form of parallelism that makes use of the massively parallel processing paradigm to evaluate an organism's performance in parallel for each generation. The second form is derived from the independent and parallel operation of each organism with the others.