I have read many game development books, and this is definitely one of the best ones out there. Just by flipping through the pages, you can see right away how much care Mr. Buckland has taken in writing this book. There are a plethora of diagrams and code examples, and the layout and organization is excellent. The balance between theory and implementation is just right. The writing style is concise and the book covers a lot of material, yet it is enjoyable and painless to read. Personally I also like the fact that the book is physically compact so it fits easily on my desk :)
By far though, my favorite aspect of the book is the "real-life" examples he gives. It's easy to skim over all the theory without really "absorbing" the material, but when you see the concrete examples, it jogs your brain and gets you thinking about how to actually apply it. For example, when discussing the basic idea of "states", he doesn't just make up a contrived example and leave it at that; he gives several examples from actual types of games. For example, a Quake style bot might implement FindArmor, FindHealth, SeekCover, and RunAway. Even the weapons may implement mini state machines like Move, TouchObject, and Die.
Digging deeper into the actual content of the book, it covers all of the practical topics an AI programmer should know, such as FSMs, pathing, group behavior, scripting, fuzzy logic, etc. All topics are explained in enough detail that you can incorporate them comfortably in your game, and if you want to learn more, you have a solid foundation to build on. As it says in the book, being an AI programmer isn't just about memorizing a handful of techniques, but also about how to apply them in combination. So, the book covers two concrete examples: a soccer game and a simple shooter game called Raven. To further your understanding, each section also contains some suggested practice exercises, like "Write code to update their sensory system so that a bot is able to sense when it is being shot at." Finally, as a bonus, this book also talks about UML class diagrams in the appendix, which is very useful to know if you don't already.
It doesn't talk about genetic algorithms or neural networks, but that seems like a good decision because those topics rarely come up in practical AI. If you want to learn more about those, Mr. Buckland has written another book called "AI Techniques for Game Programming".
Incidentally, notice that my review, plus all the reviews before mine, all contain the word "best". This is no accident- this book is a gem, and should be required reading for anyone looking to get into game AI.