evolvers

for those, who would like to see a demonstration of the power of artificial evolution, here are several interactive evolutionary systems on the net:

a-biomorph, imitates the seminal "biomorph" system of richard dawkins, which was an inspiration for most of the interactive systems generated afterwards. but following systems are also influenced by karl sims' work based on mathematical expressions. image breeder seems to be evolving patterns that are formed by simple algorithms, so that the jumps between generations are rather abrupt (but i liked its images). there is this living image project, where you can vote on evolutions for pictures. picbreeder evolves pictures, in a way similar to the living image, but these look like 'something'. e-volver is an interactive-experiment-exhibition (so we can't get involved via internet i suppose), the images produced are rather complex and varied. there's also breed, it evolves three dimensional shapes and produces them using a 3d printer. there's more: evolve (what a surprising name), this one distributes source code.. [and also there's a whole domain concerning 'artificial life', which has a lot to do with other seminal work of sims, but i won't go into that one now.]

most of these systems utilize algorithms or mathematical expressions that yield a color value for a given canvas coordinate, thus define a picture pixel by pixel. these mathematical expressions [like: "grayness = y*sin(x)"] have the peculiar property of producing those regularities like curves, lines and subtle gradients (think about sinusoidal waves). starting with these expressions, systems continue by combining or modifying those expressions, so that children images carry some similarity to the initial patterns, but are different. it's easy to say that, you should rather have worked with a parametric approach in a more-straightforwardly-global-sense; if you wanted more control over the process (remember bitmap-vectoral pair here), like the animators do. that way you could go on manipulating an object and return to the original if you liked.

after all, we don't need these examples to be convinced about the power of artificial evolution. we eat products of artificial evolution everyday, and we feed them, and we read them and we think through them. evolutionary computation approach, therefore, is an attempt to have more control on this ongoing process and use the mechanism to obtain specifically-oriented (task-specific) and relevant (or useful) products, and much faster. indeed we need to know the amount of time needed to carry out each task.