Artificial Intelligence: A Knowledge-Based Approach

To be considered exceptional, a textbook must satisfy three basic requirements. First, it must be authoritative, written by one with a broad range of experience in, and knowledge of, a subject. Second, it must effectively communicate to the reader, in the same manner in which a course instructor must be capable of imparting knowledge to students in a classroom. Third, it must stimulate the reader into thinking more deeply about the subject and into viewing it from fresh perspectives. In Artificial Intelligence: A Knowledge-Based Approach (Boyd & Fraser, Boston, 740 pp., $48.95), author Morris W. Firebaugh has succeeded in meeting each of these requirements. To his considerable credit, he has created a well-written and comprehensive introduction to AI. It covers all the basics and then some, leaving the reader with a well-rounded understanding of the field. In addition, the book makes for enjoyable reading, which is indeed a rare and commendable feat for a scientific text. As a graduate student, my preference was for textbooks that began by acclimating me to the subject matter. The first three chapters of this book accomplish precisely this goal. Although these chapters are perhaps a bit overambitious in terms of the volume of material covered, the reader is nonetheless left with an overall feel for what AI is, where it came from, and where it might be headed. Through the introduction of the major features of Lisp, Prolog, Smalltalk, and Pop-11, one gains an appreciation for why these special programming languages are needed for AI applications. The presentation of search techniques is handled particularly well, beginning with descriptions of several problem situations requiring a search strategy and followed by a discussion of the search techniques available to solve them. Game playing, the subject of chapter 4, is an interesting topic in its own right and a good choice to illustrate the concept of search heuristics. Having already been convinced of the necessity of limiting the search space, you are clearly shown how the application of appropriate heuristics can effectively limit the search. You begin to understand the role of AI as a separate discipline of computer science by understanding the distinction between problem domains where strictly algorithmic solutions suffice and those where AI approaches are required. Firebaugh’s attempt to describe automated reasoning in Chapter 5 falls victim to an abundance of terminology, laws, and syntax rules with inadequate definition and explanation. The treatment assumes a prior knowledge of logic and propositional and predicate calculus. The chapter on problem solving, however, is Firebaugh at his best. The reading is interesting, and the text flows easily. All the topics in the previous two chapters are neatly tied together,and are well illustrated with good examples and figures. During a fascinating discussion of the general problem solver, the importance of separating the general problem-solving capability of a program from the specific domain knowledge becomes apparent. A brief survey of other problem solvers nicely rounds out the chapter. The chapter on natural language processing starts with an engaging discussion of communications between humans and computers using the English language. The separate problems of syntax and semantics are discussed, with demonstrations of how the rules of grammar help to define syntax and how the inclusion of knowledge is essential in dealing with semantics. Excellent use is made of Prolog and Lisp code segments as examples, and new concepts are introduced at a perfect pace and are fully defined. Chapter 8, however, is afflicted with a malaise similar to that of Chapter 5 and, as such, provides little further insight. An example is the discussion of transforming the surface representation of a communication to a deep representation of its meaning. Here, reference is made to a meaningless figure consisting of two boxes, an arrow connecting them, and an uninformative caption, along with sample listings of nothing more than input and output. I was left with no idea of how the transformation actually occurs. Chapters 9 through 13 constitute the heart of the book and are arguably the best chapters in the text. They concern the power of knowledge, how it is best represented, and how it is most effectively applied. The order of presentation is superb, with each subsequent chapter building on the previous one. We are taken on a journey through the classic knowledge representation techniques, illustrated by several benchmark expert system development efforts. In a move from specifics to generalities, the properties of knowledge-based systems, features of human problem solvers, and architectural principles are explored. A most impressive feature—the construction of a simple rule-based system using a commercially available development tool—comes next. The presentation is in an interactive mode, with the printed page reflecting exactly what you would see if seated at a terminal. This approach is as close as one can get to hands-on learning from a textbook! Supplemented with clear diagrams and examples, the material in each chapter flows smoothly into the next. Although introductory in nature, these chapters nonetheless manage to provide a thorough understanding without overwhelming the reader. The book also includes chapters on pattern recognition, computer vision, and robotics. Although these areas are unquestionably treated thoroughBook Reviews