The Tierra project was initiated by Tom Ray, a biologist, as a study into the mechanics of evolution. Originally it started out as an experiment in genetic algorithms on a computer but it evolved a couple of years ago into Network Tierra where the computer programs actually migrate from host to host in search of an evolutionary advantage. I think one of the great achievments, was the set of instructions chosen to create the programs. By sheer luck they turned out to be one of the least brittle sets of instructions possible, meaning that large mutations in the programs would still have a good chance of being valid code and executing. By contrast every high level programming language is extremely brittle - change 1% of the instructions at random and the program ceases to execute.
Other posts tagged as ai