I’m currently watching Melanie Mitchells introductory online course on Complex Systems on what seems to be a site dedicated to this kind of subjects. This is a really fascinating topic, as it touches on everything, that might not be easily describable by means of one-liner laws, but that can be quantitatively grasped by modern computing power. Also, on such a level of abstraction, there is the possibility to reach universal understanding of things, as complex systems, may they be weather, economies, fractals, life or even intelligence, tend to have similar structures: they’re composed of
– simple agents
– that interact in ever changing ways (nonlinearly)
– without a central control instance but
– developing certain system-wide behaviours like hierarchies or information processing habits.
Personally, I have been fascinated by fractals very early on, and several years ago I even tempered with some programs (check out the “more on deviantART” section!), that easily yield artistic results, without having understood much of the underlying math. Now I’m revisiting these things from a somewhat more scientific viewpoint as I developed a fad for rational creation, that is such, that you can later justify in a clear way, because you know what you’re doing.
Okay, but all this is really just intro. One of Mrs. Mitchells lessons introduces the programmer Karl Sims. During the 1990s he developed programs of artificial intelligence, that used what we nowadays think to know about evolution to optimize problem solving techniques. Originally, this branch of programing was introduced by John Koza, but Sims used it for decisively artistic purposes. One of those programs calculated certain color values and spatial distribution of pixels and asked a user about the aesthetic value of the product. If it was somehow pleasing, the next generation of images would be derived from the first, introducing minor mutations at random places. Bad images would be deleted, the chosen ones optimized. These “Genetic Images” were shown in Linz, Austria to museum visitors, who could actively decide on the images to be kept by stepping on sensorplates before the according video screen.
In another project, Sims devised algorithms for creatures consisting of three dimensional primitive bodies in simulated physical environments, that would adapt problem solving strategies and the physical appearence according to certain needs. Here are some screenshots of the genetical artwork and a youtube clip by Sims himself showing his virtual animals.
There is also a video about simulated evolution of plants. Things like this are used in modern movies like avatar – but think of the possibilities, when things like this are not only used for theatrical effect. You can find the movie here:
Or see here two stills:
Keep in mind that all of this has been done before the dawn of the new millenium!
For those of you interested, I list Karl Sims’ website, a java application based on this work by Tatsue Unemi, a general database on virtual art based in Austria and an extensive bibliography on related work.
Update: found out about Bill Sellers’ work, which continues on the genetic algorithm idea to simulate all kinds of vertrebrate animals. Most of their source code is freely available and a good starting point for your own endeavours, if you’re interested: