Artificial Life: A Report from the Frontier Where Computers Meet Biology, by Steven Levy.

New York: Pantheon, 1992. 390 pp. w/index.

Artificial life, as opposed to artificial intelligence, tries to model aspects of living creatures in general rather than of human-level intelligence, or even to create entire living creatures. It proceeds in general through bottom-up construction of artificial creatures in simulated environments, looking for complex emergent behavior from the application of simple rules. Levy provides a good overview of the field. He is strong on history and biography of the mathematicians, computer scientists, roboticists, and biologists involved in various aspects of research. He is less so on the technical details -- I was not expecting it to be a how-to book, but I could still wish for a slightly more detailed description of some of the simulated ecologies and cellular automata. However, he offers plenty of references in the Notes section, so this is not a serious defect.

Levy also considers to some degree the philosophical questions involved -- can the creatures in an artifical ecology in a computer's memory ever be examples of life, rather than merely simulations of life? What about computer viruses? What about self-replicating robots? Is it ethical to deliberately create self-replicating organisms, when the example of computer viruses clearly shows that they have an inherent potential for getting out of control? He gives a variety of views on these questions from a number of researchers working in the field. His own position seems to be, on the first, that it is superstitious vitalism to deny the possibility of life in the form of pure information, but that the artificial systems so far constructed are probably not quite complex to qualify as being alive. On the second, he is more ambivalent. Creature-programs written in pseudocode and implemented on a virtual computer within an actual one are in no danger of spreading beyond their controlled environment and doing harm, as viruses have done and self-replicating robots might do. On the other hand, they can be used for experiments in the nature of evolution, but can't do any useful work, as it has been proposed to do with benign viruses and robots. The potential for harm as artificial life gets out of control -- by evolving from its intially built-in symbiosis with and dependency on man into parasitism on or competition with man -- seem to be directly proportional to their usefulness for doing work in the "real world."

Search on

To the Book Reviews page.
To my home page.
Email me: Jim Henry.

Get a GoStats hit counter