Apparently, the end result can be altruistic behaviour among robots. A Swiss team (Ecole Polytechnique Fédérale de Lausanne and University of Lausanne) has applied Evolutionary theory to the development of robots, wherein the neural networks controlling the little buggers are combined using genetic algorithms over several generations. The results are rather surprising, with maze movement without collisions being solved in a hundred generations or so, and even altruistic behaviour coming to the fore for ‘genetically’ similar robots. It is a very insightful look at the applicability of mathematics to understanding human behaviour and evolution in general. More details here.