AI Game Programming Wisdom 2

Rating:
Summary: Another useful addition to the game developers shelf
Review: All of the books in the Game Programming Gems and AI Wisdom series have been quite useful to me as a game programmer. AI Game Programming Wisdom 2 is a continuation of the Charles River Media series devoted to gaming issues. As with the rest of the series the articles are of varying lengths and detail. Many of the articles simply reference other articles in the series rather than basic or classic works in AI. So, this series is less useful as a bibliographic source. The topics covered in the articles are usually short and written at a high-level, but nonetheless very interesting and worth further research. And therein lies some of the frustration. It is often difficult to find more information about a particular subject since most of the references simply refer to articles in other books in the series, which are themselves are rather vague.

AI Game Programming Wisdom 2 is structured very much the same as the first book in the series. The articles are broken into similar sections, with the addition of a new section on finite-state-machines and splitting the section on learning into two. Having already covered the A* algorithm, its various optimizations, and navigation meshes in the previous books those topics are virtually absent here (though they may be buried as a secondary topic in some of the path-finding articles). I thought that the series had exhausted discussion of finite-state machines, but the editors managed to include a couple of interesting articles. The demand for more realistic AI behavior in character-based games in growing. And while a full simulation of emotions and human characteristics is unteneable, at the moment, there are some interesting ideas in the Learning section of the book e.g. "Motivational Graphs: A New Architecture for Complex Behavior Simulation.".

I can't say that this book is indispensible from an implementation perspective. Most of the articles are simply too vague - though there are notable exceptions throughout the entire series of books. What I found most useful in these articles was that they stimulated thought and helped to generate discussion among the programmers and designers at work.

Rating:
Summary: Downplay AI and just extract what you need
Review: Has Artificial Intelligence just become a catchall phrase for any advanced algorithmic programming? Because if you peruse this book, you may consider some methods it describes as just being sophisticated algorithms, and not necessarily anything to do with AI per se.

Ok, let us not try to get bogged down into semantics. The book does have many nice, state of the art (2004) methods for handling diverse problems arising in game coding. For example, the section on speech recognition is quite well done. The main thing to take away from that section is that you need to restrict the problem so that the range of possible user responses is limited. Which may not necessarily be what that author intended. But from your point of view of easier and more robust coding, this is the pragmatic way to use speech recognition.

Of course, an AI purist might argue that what I have just suggested is not really AI. I agree. I use the speech recognition as just one example. You may derive the greatest benefit from this book if you deprecate the AI aspect and just see what methods you can usefully cull. Always remember that true AI is really hard. You are writing games. NOT research. Leave that to others.

Rating:
Summary: "Ok, let us not try to get bogged down into semantics?"
Review: Honestly, I just had to submit a review to comment on this line from a reviewer of this book. "Ok, let us not try to get bogged down into semantics." What the heck does this mean? I think he/she meant, "Okay, let us not get bogged down in semantics."
What else is wrong with this reviewer's reading/writing skills that contributed to their negative review of this book? If you can't write clearly...