IDEAS & TRENDS;
Beyond Artificial Intelligence, the Search for Artificial Life
LEAD: COMPUTER viruses devised by high-technology vandals have crippled computer networks, damaged data and occasionally left personal computer users in a state of near hysteria. But some scientists are considering the possibility that these self-replicating programs might have constructive uses as well.
COMPUTER viruses devised by high-technology vandals have crippled computer networks, damaged data and occasionally left personal computer users in a state of near hysteria. But some scientists are considering the possibility that these self-replicating programs might have constructive uses as well.
In the view of these researchers, the virus programs are the most visible example of a strange new field called artificial life. To better understand the mechanisms of biological life, computer scientists, biologists, mathematicians and chemists are using computers to simulate living organisms and to fabricate computerized environments in which these creatures can fight for survival in an electronic version of the Darwinian struggle.
At the Second Artificial Life Conference, held here earlier this month, several scientists insisted that the term virus is more than a nice metaphor.
In a recent paper, J. Doyne Farmer, an artificial-life researcher at Los Alamos National Laboratory, wrote, ''Although computer viruses are not fully alive, they embody many of the characteristics of life, and it is not hard to imagine computer viruses of the future that will be just as alive as biological viruses.''
By studying these and other artificial creatures, scientists hope to gain a better understanding of many of the deepest mysteries of biology, like how a single cell can divide and develop into a complex organism. But they also warn that breakthroughs in the field could be used illicitly to create more virulent computer viruses, perhaps with their own immune systems and other techniques of eluding capture.
Viruses are secret programs, written by technologically adept mischief makers and unleashed to spread through computer networks by attaching themselves to other programs. Once they have infected a new machine they can lie dormant for weeks, suddenly springing forth to surprise the user by doing something as benign as displaying a funny message or as vindictive as destroying valuable data.
But these viruses are primitive compared with what the future might bring. As they speculate about a time when the line between the real and the artificial has grown vanishingly thin, some scientists envision horrors reminiscent of the worst nightmares about genetic engineering.
Perhaps, they say, scientists using supercomputers to develop complex new artificial life forms could accidentally release an inextinguishable data-eating monster into a network, causing the country's scaffolding of data highways to collapse.
Russell Brand, a computer scientist at Reasoning Inc., a software company in Palo Alto, Calif., said that it was important to consider the consequences of artificial-life research early. ''Once things get out they are very hard to control,'' he said. ''This is the only before we get.''
Mr. Brand, who has worked as a computer security expert for the Government, said several agencies are fighting over which one should manage a center for disease control for computer viruses. In two to three years, he said, it is possible that computer users whose machines are infected will have to report the incident to a Government agency.
But the future of artificial life does not have to be drawn from the pages of ''Frankenstein'' or ''The Andromeda Strain.'' Conceivably, researchers say, the same kind of programming concepts that led to software invaders like the one Robert Tappan Morris set loose in a national computer network in 1988 could be used to make programs that learn and develop on their own.
So far, artificial organisms exist only as relatively simple programs that are in little danger of being confused with the real thing. But a number of scientists have begun to experiment with lifelike programs they call genetic algorithms.
For example, researchers at the University of California at Los Angeles have designed simulated ''ants'' - small programs that are selected according to their ability to follow trails on a computerized grid. Those that are most adept at this task survive and reproduce, passing on the skill to succeeding generations in the form of digitally inherited DNA strands - a simple 32-bit binary number that acts like a gene. Rapidly evolving through thousands of generations on a supercomputer, the ants' navigation skills steadily improve.
Other scientists believe that the best defenses against computer viruses may also come from biological ideas. Some researchers have suggested that personal computers are particularly vulnerable to viruses because software standards are becoming so homogeneous. Making software that can run on any number of computers has obvious economic advantages, they say, but only at a price. More diverse software, they argue, would make it more difficult for computer viruses to spread widely, just as the diversity of biological life keeps a single invader from wiping out entire populations.
Computer software might ultimately be more secure if it was developed using a high-speed version of natural evolution, said Danny Hillis, a computer scientist and co-founder of Thinking Machines Inc., a supercomputer maker in Cambridge, Mass. Competing programs could be exposed to swarms of viruses in the laboratory; only the fittest would survive.
As computer scientists continue to close the gap between simulation and reality, artificial-life researchers are likely to share ethical problems with more advanced sciences like molecular engineering.
''We should have a big concern over how we experiment with artificial life,'' said Eugene Spafford, a Purdue University scientist involved in the study of computer viruses. He said that the trial of Mr. Morris, who was convicted last month of a felony violation of the 1986 Computer Fraud and Abuse Act, had raised ethical questions that will become more pressing in the future, both for computer scientists and biologists.
Mr. Spafford posed the question of whether it would be ethical to release a virus (the real kind) that would provide immunity against a disease like leukemia but without public knowledge and consent. In a way, this is already done with the Sabin vaccine, a weakened polio virus that can confer ''contact immunity'' on unimmunized people who catch the generally harmless infection from someone else.
''When you conduct experiments with live subjects you must have control over the spread of your 'program' and have informed consent,'' Mr. Spafford said. And so it should be with computers, he says. ''There is too much experimentation being done just to throw things together and see what happens,'' he warns. ''That's not good science.''