Artificial Intelligence: A New Synthesis

The field of artificial intelligence has an interesting history, both in terms of its content and the philosophical debate it has provoked. The field could also be loosely described as divided into two camps, those who view it as a collection of highly sophisticated algorithms, and those who view it as an attempt to create machines that exhibit human-level intelligence. Ironically, in the latter camp, it is difficult to assess the progress that has been made, since criteria for measuring machine intelligence are never explicitly given. Instead, dependence has been made on the "Turing test" for intelligence, a test that is difficult to apply, and in fact can be said to be too vague for a practical, objective assessment of machine intelligence. This book is written more in the context of the latter camp, than in the former. However, in-depth discussion of the Turing test is not given, and this actually is one of the main virtues of the book, although the author clearly believes that the purpose of doing research in artificial intelligence is to achieve human-level intelligence. As he remarks in the last paragraph in the book, it was written to overview the techniques that he believes are required to achieve human-level intelligence. Although he does not explicitly give the reader tests for machine intelligence that will allow progress to be measured, he devotes a small portion of the book to various ideas on just what constitutes intelligence. The book also gives a general (and sometimes very brief) overview of the algorithms used in artificial intelligence. Search heuristics, neural networks, and genetic programming are some of the topics that are covered. The influence of the "intelligent agent" paradigm, that is now taking the AI community by storm, is very apparent throughout the book. The author though does not neglect some of the topics in "good-ole-fashioned" artificial intelligence that arose decades ago and is still applicable today, especially in the field of logic programming. These topics include resolution in both the propositional and predicate calculus, and in expert systems. By far the best discussion in the book is on knowledge-based systems and evolving knowledge bases. This topic has taken on considerable importance in recent years due to the importance of data mining and business intelligence. Readers who are considering artificial intelligence as a career choice will find good motivation by reading this book. The field also is quite different than most others in that it respects a high degree of individual creativity and ingenuity, and has a high bandwidth for new ideas. Beginning with its origins in the 1950s, the field has grown by leaps and bounds, but its applications have exploded in the last five years, fueled mainly by business and financial applications. Concerned not only with achieving human-level capabilities, but also with other forms of intelligence and how they can be useful, artificial intelligence has become

A nice book. Especially the order in which the topics are covered is a good idea. However, you will not find the following errors reported in the book's webpage:Page 52: The "high-degree function" is not a function!Page 92: In Figure 6.6, the topmost pixels that get deleted as a result of the averaging operation should actually remain there, since both their sums are 4, which is greater than the threshold, which is 3.Page 100: In Fig. 6.13, the last row of the last image contains a spurious image boundary.Page 151: In Fig. 9.8, there are two nodes with name n; the one which is higher in the figure should have the subscript 1.Page 152, item 3 in the list: There is an implicit assumption that h-hat always returns 0 for goal states. I don't think that this assumption is stated earlier in the text.Page 165: In Figure 10.1, all arrows are supposed to be pointing away from the current state.Page 246: The last paragraph mentions ".. the two interpretations for Clear and On suggested by Fig. 15.2", but aren't actually THREE interpretations suggested for On?And in the current errata list in the book's website, something is clearly wrong with item 6, since it says n_i should be replaced by n_i.All in all, a good book.

AI is the future and I believe Nilsson is one of the best experts of AI.. thank you....

According to my former AI prof, Nilsson suffers from "Physics envy." Given that AI is a fairly new, fairly splintered facet of Computer Science research, there is a relative absence of quantitative analyses of the subject to rival such other fields as chemistry, biology, or physics. As such, Nilsson resorts to quantifying most every piece of data or concept in the book. In some cases, his formulas can more lucidly be explained in words or simple algebra, rather than polynomic summations and calculus.Nevertheless, for the non-beginning student of computer science that has an interest in the subject, this book covers the gamut of AI subjects. Topics include neural networks, genetic programming, multi-agent programming, fuzzy logic, and machine vision. While no topic is covered in-depth, the broad scope of the book allows one entering the field to decide what areas, if any, are of paramount interest.I recommend this book for a 2nd or 3rd year CS undergraduate with a background in calculus and with a serious interest in artificial intelligence.

Although I do not recommend this book as a first one (because the material is too much dense for a begginer), it undoubtedly is a "must" in any set of books if you intend seriously working inside this area. Nilsson has a clear writing and (if you have some background) it's a pleasure to follow him even through hard topics.