In this work, Fogel makes a major contribution to the interface between evolutionary computation and artificial intelligence (AI). One of Fogel’s main points is that intelligence, whether obtained by human or machine, can be best understood within the same simulated framework involving physics and biological simulations. To this end, the phrase “artificial intelligence” is a misnomer. There is “nothing artificial about it.”

This text is the third edition of the book, published ten years after the appearance of the first edition [1]. Although much has happened, both in world history and in the advance of science, since the appearance of the original edition, one event provides a most significant motivation for the ongoing endeavor of this work. In May 1997, the IBM supercomputer program Deep Blue defeated Garry Kasparov, the world chess champion of the time. Admittedly, Deep Blue accomplished its task by an untiring exploration of all possibilities in a record amount of time. The author correctly questions to what extent this can be considered “intelligence.” This issue is the heart of the first chapter, defining artificial intelligence, and introducing a number of new technologies that attempt to derive intelligent rules present in the environment of the system, such as expert systems, fuzzy logic, and neural networks.

The second chapter provides the scientific fundamentals for a formidable alternative to previous statistically based methods. It introduces a neo-Darwinian approach to speciation, and formulates survival of the fittest in terms of optimization objectives. The third chapter offers a very nice introduction to computer simulations of evolutionary processes. The book then provides a solid foundation in the mathematical theory behind the methodologies and techniques involved. As a numerical analyst, I was particularly pleased with the straightforward discussion of convergence rates. This is a topic that many authors do not include in any comprehensive way.

Because the focus of the book is intelligence, the applications explored in chapter 5 involve classic problems studied in the literature, such as the prisoner’s dilemma, and strategies for tic-tac-toe and chess playing. To solve the prisoner’s dilemma, finite state automata are bred. This is a particularly interesting implementation, because of the coevolutionary adaptation of the machines. The author has made prior efforts in developing strategies for games. In 2002, the author wrote an engaging text [2] on a project called Blondie24, which describes efforts by machines to teach themselves to win at checkers. The approach utilized was a neural network that achieved an expert-level rating of 2045. In 2004, Fogel coauthored an Institute of Electrical and Electronics Engineers (IEEE) proceedings paper [3] on a hybrid approach using evolutionary computations and neural networks for chess playing. The program learned to play chess by playing games against itself. The book concludes, in chapter 6, with a perspective on a unifying principle of intelligence, and, based on the material presented in the text, suggests a “new philosophy” of machine intelligence.

This book can be viewed on many levels, and, as such, has practical interest for a broad-based readership. As a textbook for both undergraduate and graduate students, the book contains numerous detailed graphs and charts, to help students and researchers elucidate the results or concepts. This treatise makes a major contribution to the evolutionary computation literature, from historical, scientific, and educational perspectives. It is recommended reading for experienced researchers, as well as novice students interested in a broad perspective on machine intelligence.