Review by Mark Kantrowitz: Ian Pratt "Artificial Intelligence" Macmillan, London, 1994. 280 pages. ISBN 0-333-59755-9. $35. This book provides an in-depth introduction to several advanced AI topics, using inference as a central theme. Other key themes that run throughout the book include knowledge representation, search, and heuristics. Because the book is based around the idea of inference, several important topics have been deliberately left out of the book. Of these, the most notable omissions are natural language processing, expert systems, speech processing, genetic algorithms, vision, and robotics. Yet the book does include a chapter on neural networks, which seems a bit at odds with the emphasis of the rest of the book. It also lacks more specialized topics, like case-based reasoning, qualitative physics, and theorem proving. As a result this book is a good themed introduction to a subset of AI, but is not suitable for a general purpose overview of AI. Nevertheless, the book does fulfill its stated goal of framing an introduction to AI from an inference perspective. Overall, the presentation is superb, with clear and concise descriptions of concepts. When introducing a new concept, the book first covers the essential insights behind the idea before getting into technical details. The book makes good use of diagrams to help explain difficult concepts. The technical presentation depends heavily on the use of predicate calculus formulas, but a tutorial introduction to predicate calculus is included in the appendix. The technical discussion includes pseudocode for many of the algorithms. Every chapter concludes with a set of exercises and a bibliography for further reading. This book does not attempt to be a comprehensive survey of artificial intelligence, and instead focuses on a cohesive subset, going into a little more depth than the usual introduction to AI. Thus the book is suitable for a one-semester advanced undergraduate course in AI or an introductory masters-level course, assuming that other courses are devoted to NLP and Robotics & Computer Vision. The book could also be used as a supplementary text for a survey course. The book's index is not very good. For example, it does not include an entry for A* seach, even though the book devotes three pages to A* search. Topics: Backward Chaining, Closed-World Assumption, Defeasible Inference, Dependency-Directed Backtracking, Forward Chaining, Frame Problem, Induction (ID3), Inference, Logic, Means-End Analysis, Memory Organization (Scripts and Frames), Neural Networks, Non-Monotonic Logic, Planning, Probabilistic Reasoning (Bayes Networks), Search, Truth Maintenance Ordering Information: UNITED STATES -- Scholium International Inc. PO Box 1519 Port Washington New York, NY 11050 Tel: 516-767-7171 Fax: 516-944-9824 CANADA -- McClelland & Stewart Inc. 380 Esna Park Drive, Markham Ontario L3R 1H5 Canada Tel: 416-940-8855 Fax: 416-940-8864