in reply to
Genetic Programming or breeding Perls
There's a rich literature on
genetic programming. For example, check out John Koza's stuff at www.cs.bham.ac.uk/~wbl/biblio/gp-html/JohnKoza.html.
Another interesting thread is Tom Ray's
Tierra. Realizing that computer programs are just too brittle to evolve well (as one bad character in a program can render it all useless, vs. the laxness in say DNA coding and codons), Ray proposed an interesting computer architecture loosely based on biology that supports evolving programs.
Another interesting idea: Danny Hillis (of Connection Machine fame ) suggested in a 1990 paper that programs evolve better when confronted with "parasites". Hillis claimed he could evolve better sorting programs when they were pitted against an evolving landscape of "hard" sort sequences (which in turn were generated by different program evolving hard sequences with the goal of stumping the sorters... see
A good site for such topics is SFI
This field hasn't generated much of practical significance yet, but it is cool.