The study of artificial intelligence can no longer be simply a survey of impressive artifacts. As Nils Nilsson emphasized in his choice of the title of an earlier book [6], enough artifacts have been accumulated to justify consideration of the underlying principles of artificial intelligence. With this new book Nilsson, together with Michael Genesereth, addresses the proposition that logic is a foundation upon which such principles may be erected. Thus, the book is not so much about artificial intellegence as it is about this foundational role which logic plays.
Whether or not we sympathize with this position, we should agree that the issue is important enough to deserve the attention of a full-length book. Previously, the only author to undertake this task was Robert Kowalski [3]. However, Kowalski's book was basically a camera-ready version of a set of course notes. It was not a terribly polished production; and it was heavily biased by a commitment to PROLOG, often to the exclusion of other interesting aspects of logic. This new book by Genesereth and Nilsson has attempted to be more thorough. While the quality of the production is much higher, there are still elements of sloppiness (particularly in the bibliography) which should have been caught by better editing; and, in the final analysis, these authors are no less biased than Kowalski, although their biases tend in another direction.
The foundation of the book is the predicate calculus. After an introduction Chapter 2 discusses its use in the representation of knowledge. Inference techniques for reasoning with knowledge so represented are presented in Chapter 3. Chapter 4 focuses on reasoning based on resolution, and Chapter 5 introduces strategies to make resolution computationally tractable. The remaining chapters deal with specific domains of reasoning, best summarized by their titles: nonmonotonic reasoning, induction, reasoning with uncertain beliefs, knowledge and belief, metaknowledge and metareasoning, state and