Exclusive: Computer 'Life Form' Mutates in Evolution Experiment

By SANDRA BLAKESLEE

In an experiment that may shed light on basic evolutionary processes, researchers have created an artificial "life form" that dwells on a network of 150 computers worldwide and is undergoing sophisticated mutations that appear to mimic the transition from simple to more complex organisms.

"I'm very excited by the results," said Dr. Thomas Ray, creator of the experiment, known as Network Tierra. "We've been on the project for three years and to have it finally working, to watch the evolution unfold, is a thrill," he said in a telephone interview.

Network Tierra has been evolving since Nov. 3, and "we are finding new things every day," said Ray, a researcher at ATR Human Information Processing Research Laboratories in Kyoto, Japan.

Dr. Freeman Dyson, a professor emeritus at the Institute for Advanced Studies in Princeton, N.J., who is familiar with Network Tierra, said the results were "tremendously exciting," but it was always a problem to assess the importance of "something so far ahead of its time."

The relevance of any digital analogy for understanding life processes is controversial, said Dr. John Casti, editor of the journal Complexity and a researcher at the Santa Fe Institute in New Mexico.

Network Tierra explores the "structures and patterns" of information that drive evolutionary processes "on this planet or wherever in the universe you might find them," he said. "Tom is interested in what makes evolution work, in general." The new research will be published in Complexity.

Ray is a biologist and expert on rain forests. Several years ago, he designed the first version of the experiment, Tierra, a computerized artificial world in which digital organisms (written in computer code similar to a genetic code) replicated themselves and showed other signs of lifelike behavior, like exploiting sources of energy. Most important, the silicon-based organisms evolved by natural selection in ways similar to those seen in carbon-based life, like humans.

The concept of organisms evolving in cyberspace is very difficult for most people to grasp, Ray said. The creatures are invisible, two-dimensional entities represented as simple mathematical equations that are logically connected. In Tierra, the creatures mutated but showed only modest increases in complexity.

One reason may have been that the organisms needed more space in which to evolve, which led to the design of Network Tierra, Ray said. If the organisms could travel to and from many computers connected by the Internet, he said, they might make better use of a heterogeneous environment, and natural selection pressures, like natural disasters created by the researcher, might force them to differentiate.

Last year, Ray and his colleagues designed an "ancestor organism," a computer code containing two cell types. One cell, with lines of code turning on specific "genes," carries out reproduction and makes daughter cells. The second cell type, with a different set of genes turned on, handles sensory information, like knowing how fast a particular computer runs and how much "energy" it contains in its central processing unit.

Once released into a computer, the ancestor organism (a zygote of sorts) differentiates to form Tierran organisms of 10 cells each. These seed creatures are then free to roam 150 computers on the global computer network, looking for landscapes on which to evolve.

The project has two goals, Ray said. One is to show that the organisms can survive under conditions of free evolution. The second is to develop a digital model of the Cambrian explosion of life, which occurred on Earth some 530 million years ago, when the first multicellular creatures with hard parts suddenly evolved. It was as if carbon-based life reached an evolutionary tipping point for complexity. Will the silicon-based creatures do the same?

To find out, the researchers released seed organisms onto 150 computers around the world linked by the Internet. The organisms are encouraged to move freely among different machines, which are like "islands" in an archipelago, Ray said.

They tend to go to machines idling for the night, when other users are off line, and there is more energy available. They reproduce using a special computer-version of a ribosome, the "factory" in animal cells where new proteins are made.

In test runs in the last year, Network Tierra made many daughter cells but the sensory cells were always eliminated, Ray said. The creatures actually became less complex. But in recent weeks, Joseph Hart, a programmer who works with Ray, found bugs in the code for implementing the sensory system. A new test was begun on Nov. 3, and Network Tierra immediately began evolving more complex organisms.

"Our first goal has been met," Ray said. "The organisms are surviving, and each day they are different. For the first time, we are able to study the evolution of digital organisms with more than one cell type."

It is too soon to know if the second goal -- a huge increase in complexity -- will be met, but the researchers have developed an observational tool to find out. It allows the scientists to pull out organisms living on any island, or computer, on the network.

"We can see the birth, death and migration of the creatures," Ray said. "We can examine their genetic code" and study them just as an ecologist examines possible new species in a rain forest, he said.

Thus far, the creatures have shown modest increases in complexity. For example, the researchers have seen three instances of genes being duplicated -- that is, they make wholly new, additional copies of themselves. In each case, both copies of the duplicated genes are still being expressed -- one by reproductive tissue and one by sensory tissue. Moreover, as time goes on, the two copies of the gene have diverged slightly.

That is significant, Ray said, because biologists believe that gene duplication, followed by divergence of the two copies, is a pivotal feature of the evolution of complex creatures and the transformation from single cell to multicellular life.