Complexity and art

Complexity influences the arts in many ways. In contemporary literature, there are numerous examples. There was the cartoon caricature of a "chaos mathematician" who warned of the dangers of meddling with dinosaur DNA in Michael Crichton's Jurassic Park. Ed Lorenz's butterfly fluttered in the poetry of Paul Muldoon. And Tom Stoppard's play Arcadia was inspired at least in part by the non-linear concepts of chaos and fractals.

Digital Darwinism is also entering the picture: there is a new breed of computer-literate artisans who are using evolutionary programming techniques to create novel forms of art. The images shown were created by Karl Sims using a population of genetic programs found by "interactive evolution". The computer is used to generate random mutations in mathematical equations which are known to produce colourful images, and the artist then applies an 'aesthetic selection pressure' by choosing for survival and subsequent breeding those which he or she finds most appealing. After a large number of iterations of this procedure, remarkably complex and intricate pictures emerge.

Sims has taken his work a leap forward by evolving creatures with both form and function. Certain of their controlling "genes" determine shape, which is built up with block-shaped segments, in a similar though highly simplified version of the way our body-plan is created; other genes describe a simplified program, the creature's brain, that controls its movement or reacts to sensors that respond to light, contact, or the angle of a joint.

Simulated evolution begins with a population of 300 "creatures", each randomly made of colourful blocks. Some look boring. A few are bizarre, while a handful twitch fitfully. Sims can evolve many generations of creatures in a supercomputer, selecting those with desirable characteristics. Intriguingly, the creatures "cheated" during his first attempts. "They did what I asked them to do but not the way that I wanted," Sims remarked ruefully.

They evolved to exploit errors in the program which was intended to ensure the creatures obeyed the laws of real life physics. One glitch allowed the violation of Isaac Newton's law of conservation of momentum. After a few generations, creatures evolved that shuffled along by hitting themselves with a paddle. Others found an error in the "integrator" which solved Newton's equations of motion and this enabled them to propel themselves along with unphysical haste.

Sims has a video which shows, among other things, the result of 100 generations of sexual reproduction that achieves the evolutionary goal of producing "good swimmers". Some digital creatures evolved into snake-like creatures that wiggle through virtual water. Others acquired a corkscrew motion or protruding paddles. "One of the interesting aspects of using simulated evolution is that you can make things more complicated than you can figure out," he says. "Luckily you don't have to." Sims speculates that he may be able to breed even stranger creatures if he could endow them with an eye for beauty.

Frontiers of Complexity by Peter Coveney and Roger Highfield is published next week by Faber Pounds 18.99.

Peter Coveney is senior research scientist, Schlumberger Research Laboratory, Cambridge.