Messenger - Vol. 1, No. 2, Page 12 Winter 1992 Digital Darwinism; Synthetic organisms evolve spontaneously into cheaters, freeloaders, good neighbors Using a computer as his petri dish, evolutionary biologist Tom Ray has created an artificial world in which synthetic creatures reproduce and spontaneously evolve, and, in the process, sometimes behave like freeloaders, cooperative neighbors or sticky-fingered cheaters. The inhabitants of Ray's world, known as Tierra, are self-replicating digital organisms that compete for computer time (energy) and memory (space). The primordial soup in which the creatures evolve is a memory block of 60,000 instructions, and Tierran life seeks to replicate itself to fill all available memory while resisting other denizens that grow and struggle for the same space. Using an IBM supercomputer and his own machine language, Ray has created a silicon world whose residents appear to demonstrate evolutionary behavior "not designed in or preconceived by the creator." Besides introducing technological innovations based on the biology of natural organisms to the creation of artificial life, he has developed a novel laboratory for the study of ecology and evolution by biologists. In April, Ray won second place in a national IBM Supercomputing Competition and a prize of $15,000 for a paper detailing his experiments in synthesizing artificial life. Since then, his findings have been reported in Science News, The New York Times, Nature, the Chronicle of Higher Education, the New Scientist and numerous popular European publications. His technological accomplishments are all the more remarkable because Ray, who studies the evolution and ecology of rain forest organisms, has never had any formal computer science training. Tierra (Spanish for Earth) and its inhabitants provide the first method for experimental studies of large-scale evolution. Ray can make multiple universes in minutes and observe emergent behavior in hours or days, whereas until now, scientists could only speculate about the process because natural systems evolved over millions of years. And he can also "tweak" the model to focus on special areas of ecological interest or pull out particularly interesting inhabitants to populate yet another universe. "We have never been able to study evolutionary systems in action in the natural world," Ray says. "Researchers have bred generations of fruit flies for selected characteristics and conducted other animal and plant breeding programs over a few generations, but we never have been able to observe change at the species level or keep track of the changing genome of any species. Evolution in my world is so much faster that it allows us to study the process on a macroscale while keeping track of the genetic code of individuals." An added advantage, he says, is that the experiment can be duplicated exactly to retrieve any information missed in the first try. To prevent his digital organisms from running rampant in the real world as computer viruses or worms, Ray sets up his artificial-life scenarios within a software program that acts as a virtual computer. "A virtual computer is a computer that could exist but doesn't," Ray says. Outside the program, the computer creatures and their environment are simply innocuous streams of data that are no more risky than a text file on a word processor. No other attempts to create artificial life have been this successful, according to Ray. "A number of people have simulated evolution but none really created it," he says. "In simulations, everything is predefined. You can define a gene for offense and one for defense and you end up with a little evolutionary arms race. But, if you set up preconditions ahead of time, you only get back what you set up. Evolution is more freewheeling. Nature starts out with molecules that stumble onto innovations, and they get ahead when they make these innovations." By synthesizing artificial life rather than simulating it, Ray says he "gave up control of the system to evolving entities so they could invent their own games and strategies." Ray's only requirements were that the creatures could reproduce and were capable of open-ended evolution. "Nothing else was predefined except for the physics and chemistry of the universe they live in, the architecture of the computer itself," he says. Several necessary mechanisms were built into the program. A "time slicer" doles out computer time uniformly and chunks of memory are always available. A "reaper" removes either the oldest creatures or the ones that were most defective, that is, those who generated the most errors in executing their procedures. "Without the reaper, the creatures would live forever and there wouldn't be any room for anyone else to be born," Ray says. A "gene banker" was also created to save to disk interesting creatures of varying sizes as they evolved. These digital organisms could be inserted in subsequent experiments. Spontaneous diversification occurs rapidly, Ray says. The ancestor of Tierra was made up of 80 instructions, but creatures soon emerged that were capable of reproducing in less time-that is, with fewer instructions. The smallest creature to arise required only 22 instructions and could reproduce almost six times faster than the ancestor. Some larger sizes also appeared and went extinct quickly. The largest was over 23,000 instructions in size. Evolution occurred in Tierra in a completely natural way. "In the Tierran world, creatures behave in ways that aren't programmed because they interact and influence each other," Ray says. "Innovation was the surprising thing," Ray says. "I had no idea that Tierra would get all these different types-parasites that cause their hosts no harm; hosts that are immune to parasites; hyper-parasites that steal energy; and social interaction where a cluster of hyper-parasite neighbors pass an energy source (the pointer) back and forth to help each other reproduce." Parasites and hyper-parasites, Ray says, behave like a virus that invades a cell and uses the metabolic machinery in the cell to replicate its own genome. "A parasite uses information from the host, but doesn't take anything away and doesn't include the information in its genome. It's like a person reading a newspaper from over your shoulder," Ray says. "But a hyper-parasite steals energy, taking computer time from another creature, and that is harmful. It's the equivalent of a neighbor stealing your electric lines and rewiring them to the house next door. You can't turn on the lights or run your television set, but you are still paying for the electricity." An even faster, sneakier competitor- known as the hyper-hyper parasite-ultimately appeared in Tierra and took advantage of the social hyper-parasites by cheating. Once this creature was passed the pointer by the cooperating group, it refused to relinquish the energy source. "They kept coming up with ways to exploit each other, and this is just the way it works in the biological world," Ray says. Ray, who holds master's and doctoral degrees from Harvard University, joined the University's School of Life and Health Sciences in 1981. He says he believes his biological training helped him successfully create "evolution in a bottle." Rays says he succeeded because of two innovations in computer language. First he made his instruction set small, based on the size of the human genetic code. The other feature borrowed from molecular biology is the addressing mode. The creatures search for a matching numerical pattern, just as a biological molecule would. For example, a code with the order 0001 will search for the complimentary pattern 1110. Tierra illustrates the evolutionary theory that adaptation to other organisms is the primary force driving the diversification of organisms, Ray says. The artificial world also makes it possible to study other ecological processes, including competitive exclusion and coexistence, punctuated equilibrium (a period of stasis changing to rapid evolution), the keystone predator effect (a predator increases the diversity of plants and animals in a community) and population cycling. The first creatures he designed were not very clever and his first instructions not very powerful, Ray says, yet many different types of artificial life formed and evolved. "It appears that it's rather easy to create life," he says. "Virtual life is out there waiting for us to provide environments for it so it can evolve."