arflThorny engineering problem got you down? Just put Dr. Frankenstein on the case. He’s not the only one creating New Life Forms

Some programmers die and go to heaven. Others go to hell. Bruce Damer went to purgatory — voluntarily.

Until recently, the 33-year-old Canadian worked for a California software company whose claim to fame was an application that printed much of the world’s junk mail. He quit a few months ago to develop software that will sort a user’s E-mail and toss out the crap.

Egad, he’s got a lot to atone for, you might say. And you can bet it won’t be easy. He has to hatch intelligent, artificially alive objects to make it work.

Artificial whaaaat?

Life, he calls it.

Damer may be the first programmer to renounce his job in pursuit of the artificial life grail, but he’s not alone in seeing its promise.

Researchers around the world, from the Santa Fe Institute in New Mexico to the Free University in Brussels, see this bizarre branch of computer science that picks up where artificial intelligence left off as a way to solve problems that once seemed intractable.

Already companies are investing in it. Maxis, of Orinda, Calif., has developed SimCity, a multimedia computer game that uses artificial life techniques. Even British Telecommunications PLC is working on applying it to call routing.

In many ways it’s the opposite of artificial intelligence. Instead of mimicking the workings of the human brain, artificial life (also known as evolutionary or genetic programming) involves making millions of tiny program “cells” that are used to attack a problem.

The best are kept and bred with each other. As Darwinian theory predicts, they get better at solving the problem over successive generations. But unlike human evolution, which takes millions of years, this sort of evolution can be done very quickly on PCs.

“Nature and biology have discovered how to solve a lot of very complicated engineering problems,” says Christopher Langton, director of the artificial life program at the Santa Fe Institute. “[Artificial life] is a good way to find a solution when we don’t have a theory about what the right solution should even look like.”

Langton says it could be used to develop vertical applications for a lot of different industries. For instance, NASA could use it to create adaptive control software for robot explorers that would travel to other planets and adapt to situations the engineers couldn’t have anticipated in advance.

Sound a bit farfetched? Maybe so, but artificial life techniques have already been used successfully at biotech companies to develop drugs in a fraction of the time it normally takes, says Gerald Joyce, a biochemist at the Scripps Research Institute in La Jolla, Calif., which has done pioneering research in a field known as directed molecular evolution. “The idea now is to make huge libraries of potential drugs, use the ones that are best as parents, and let them produce baby molecules,” Dr. Joyce says. U. S. biotechnology firms Darwin Molecular Technologies Inc. in Seattle and Gilead Sciences Inc. in Foster City, Calif., have done just that, he adds.

Apart from vertical applications,

artificial life may also be used in the PC realm to develop the first operating system that truly takes advantage of massively parallel computer systems, Langton says. Until now, engineers have tended to treat a parallel computer system like one big serial computer. And that’s a bit like treating a human being like one big foot.

“Much of what we’re doing in parallel computation is aimed at eliminating the funny kinds of side effects and weird ways in which parallel systems can behave,” he says. “But the true power of parallelism lies exactly in these side effects.” Rather than eliminating them, using evolutionary programming computer scientists could capitalize on aspects of parallel computing we don’t fully understand, he adds.

And that’s not all. Artificial life techniques have the potential to extend existing commercial software applications to new heights, says Will Wright, founder of Maxis. His own company’s programs, SimCity and SimLife, are games based on a series of simple rules that, when combined, come up with very complex dynamics that simulate the way cities and creatures evolve.

But artificial life will probably find its way into more practical applications in the near future, Wright says, including CAD software that would produce optimal solutions rather than merely satisfactory ones. Similarly, it may be used in complex scheduling packages or forms software.

Damer hopes to use it in E-mail, and he’s working on an interface that will “learn” what sort of mail users want to receive and what is junk after it’s “taught” — with a pat on the head or some form of negative reinforcement. Although most of today’s artificial life programs are simple creatures, Damer and his cohorts have high hopes they’ll evolve into higher life forms in time. Eventually, in the most complex, evolved artificial life programs, consciousness will appear, he says. “One day it will sit up and say, ‘Hi.'”

The idea isn’t quite as wacky as it sounds, says Esther Dyson, a New York-based computer industry analyst. “I think if something shows up that’s more intelligent and has a better soul than we do, God bless it,” she says. “And that could well happen.”