ARTIFICIAL MINDS
Stan Franklin
Science
Reviewed by A. J. Sobczak

***

A professor of mathematics gives an overview of work on artificial intelligence and discusses several debates.

***

A basic question in the field of artificial intelligence is what constitutes a mind. Stan Franklin argues that the tests for intelligence that are put to machines have become increasingly more stringent, always just beyond the limits of the current technology. He sees the mind as a continuum of ability, with very simple minds coexisting with more complex minds, or perhaps with simple minds being components of more complex minds. From this perspective, Franklin sees the question of artificial intelligence not as whether machines can think but how much mental ability a machine exhibits, and of what types.

Franklin presents his book as a tour of research in artificial intelligence. As tour guide, he presents favorite topics in more depth, but he gives a good overview of current research. The tour becomes increasingly difficult as the book proceeds, with early chapters discussing basic concepts of mind and later ones containing discourses on mathematical topics and debates on theoretical issues. Franklin's attempts to make the material accessible to readers with limited background are as successful as could be hoped.

Some of the more interesting chapters describe various forms of artificial intelligence. Franklin describes the machines and the environments in which they operate, from robots that perform simple tasks in the real world to computer programs that operate in simulated environments, performing such tasks as finding "food" and avoiding predators. The discussion of these machines relates to chapter I 1, concerning the question "What do I do now?, " which Franklin sees as one of the key questions confronting both humans and artificial forms of intelligence. This chapter borders on psychology and, like others, offers interesting insights into the workings of human minds. The more complex chapters describe physical processes of sensory recognition and how machines simulate them, hierarchies of goals, and creation of "information" out of "data." The final chapter, "Into the Future," offers speculations on the future of artificial intelligence. The book is indexed well and contains fourteen pages of references.

***

(The MIT Press, 1995, 449 pages, $30.00, ISBN 0262061783)