In 2003, Lenski et. al. published an intriguing paper about the evolution of complex features in digital organisms (The Evolutionary Origin of Complex Features Lenski RE, Ofria C, Pennock RT, and Adami C Nature, 423:139-144). In this paper, organisms had digital (e.g., Von Neumann) genomes, and evolution arose via mutation. Mutations could either be point mutations, whereby an instruction in the genome was randomly replaced by a different instruction, be insertion mutations, whereby instructions drawn at random are inserted into the genome, or deletetion mutations, where an instruction is deleted from the genome. (For now we ignore another souce of genomic mutation, which is the result of a genome not being able to copy itself accurately due to prior mutations to it.)
In the experiments performed by the authors, populations of fixed size of these organisms competed to out-reproduce each other. Mutations that aided this goal were beneficial, and those that did not where deleterious. Organims that evolved the ability to perform specific logical functions had their metabolism speeded up, which gave them a reproductive advantage.
The experiments we are conducting are similar to those in the Science paper, but with one big difference: for insertion mutations we do not pick an instruction uniformly from the instruction set. Instead, we pick an instruction out of the genome being mutated. Our approach corresponds to gene duplication events in natural genomes. We expect our experiement to show that such gene duplication events in aritifical organisms speeds up evolution by favorably multiplying those instructions that have already proven to be of value to a genome.
These experiments are computationally costly. A single run takes 10 to 14 hours on a PC, and we need 50 to 100 runs for each point in the parameter space that that we want to investigate. So that's why we seem to be taking over a lot of resources during the summer when most people are gone and computer cycles get otherwise wasted.