Roger Alsing has written a genetic algorithm – a computer simulation of evolution via random mutation and “natural” selection – that recreates the Mona Lisa. It achieved a pretty good replica layering only 50 semi-transparent polygons of various colors, in just shy of a million generations. And it got pretty close in the first hundred thousand generations; a neat example of R. A. Fisher’s “geometric model” of evolution toward an optimum, in which evolutionary change slows as the distance to the optimum decreases.
Via kottke.org and BoingBoing.
(Considerable debate on the BoingBoing thread about whether this is “really” evolution, since there’s a preordained optimum – I’m going to to say that it is, in fact, evolution. Specifically, a single bout of adaptive, directional evolution towards “Mona Lisa”-ness. The equivalent of which happens all the time in nature, except that usually the selective optimum shifts from “Mona Lisa” to “Les Demoiselles D’Avignon” after a million years or so.)