Artificial Intelligence: Making Up Our Minds

MACHINES AND INTELLIGENCE A Critique of Arguments Against the Possibility of Artificial Intelligence. Stuart Goldkind. Greenwood Press. Westport, CT, 1987. 138 pp. $29.95. MAN-MADE MINDS The Promise of Artificial Intelligence. M. Mitchell Waidrop. Walker and Company, New York, 1987. 288 pp. $22.95 HB, $14.95 PB. After 30 years, investigators and critics of artificial intelligence (AT) continue to differ in their definitions of the domain and the fundamental claim that AI is or is not possible. Academics have a stake in the outcome. So do scientists and engineers who leverage their efforts with computers. More and more, software products are likely to claim additional function and benefit due to AI techniques. Should these products be required to display the message: “Caution: this software is based on a scientific domain that may not be possible”? Two recent publications offer contemporary summaries of this continuing debate. Stuart Goldkind’s monograph Machines and Intelligence examines the arguments surrounding the possibilities of AI. He reviews the classic philosophical questions: Can machines have purpose? Can they make errors? Can machines interpret or exhibit behavior appropriate to different contexts? Goldkind examines the key combatants, provides overviews of their arguments and adds a measure of his own thinking. M. Mitchell Waldrop’s Man-Made Minds takes a broader approach, blending basic ideas with the scientists behind them, and examining Al’s historical context, theoretical framework and contemporary research. Machines and Intelligence con- sists of four substantive chapters focusing on the work of long-standing Al critic Hubert Dreyfus, Alan Turning, Richard Taylor and Norman Malcolm. Goldkind, an artificial intelligence researcher, begins with a critique of Dreyfus’ 1979 book What Computers Can’t Do. Goldkind emerges as an AI partisan and the chapter emerges as another obligatory Dreyfus-bashing treatise. A chapter entitled “Machines and Mistakes” contains an elaboration of ‘Thring’s ideas about the place of error in a mechanical model of mind. Taylor’s 1966 book, Action and Purpose, also is discussed. Finally, the topics of explanation and causality are explored in the context of Malcolm’s 1968 paper ‘The Conceivability of Mechanism.” The book ends abruptly with no summary chapter. The score, according to Goldkind, weighs heavily in favor of the possibility of AI. Man-Made Minds consists of 11 chapters organized into sections providing historical context, descriptions of the current state of affairs in academic and industrial settings, and speculations on the future, particularly the ethical and social implications of AI. Waldrop, a physicist and science writer for Science magazine, provides an excellent journalist’s treatment of the issues. As I read his accounts of AI progress in Science in the past, I wondered how he was able to cut to the center of controversies, set context effectively and inform both the insider and observer with clarity. Man-Made Minds answers that question— Waldrop did his homework. The book is a science writer’s journal, a compendium of the detail necessary for distifling an issue to its essence. Waldrop defines the boundaries of AI in an enthusiastic, descriptive manner. The result is an optimistic account of AI theory and practice. By citing achievements along with open problems, Waldrop leaves the reader intrigued about the next 10 years. Engineers will find Waldrop’s account rich in implications for current and future applications of Al. Consumers of AI technology are advised to be discriminating. Man-Made Minds provides the necessary education. Readers will be pleased to discover that their intuitions about hyperbole are correct and, at the same time, that more effective uses of computing have emerged from the AI laboratories. Together the two books exemplify the tension between academic and applied approaches to technology. I recommend both, but for different audiences and purposes. For a review of the philosophical underpinning, read Goldkind, for an extraordinarily well-written “biography” of the past, present and future of AI, look to Waldrop. King is director of training operations for Teknowledge, Inc., P.0. Box 10119, Palo Alto, CA 94303, a company that specializes in providing high performance inference engines for mainstream applications of knowledge engineering. He has a background in cognitive science and education and is co-author of Expert Systems: Artificial Intelligence in Business (John Wiley 1984).

Artificial Intelligence: Making Up Our Minds

MACHINES AND INTELLIGENCE

MAN-MADE MINDS

Interested in reading more?

Become a Member of