"Intelligent Java Applications" was my first introduction to neural networks and AI in general, when I bought the book a couple of years ago. It took an undergraduate neural modeling class at an Ivy League university before I appreciated what a lousy introduction to neural networks this book actually is. Watson glosses over all of the theoretical basis of neural networks, and does an inadequate job of explaining just what a neural network is and how it works. The author could have easily explained (in a paragraph) what a linear associator is, before launching into the more convoluted, multiple-layer back-prop algorithm that is presented in the book. If you want a real book on neural networks (along with neurophysiological, theoretical, and historical developments), I highly recommend "Introduction to Neural Networks" by James Anderson.
Watson also wastes paper by baby-stepping the reader through some useless GUI and networking packages that makes his code presentable (when run) but much more difficult to decipher. A useful AI tutorial would not be so tied to a specific language. Alas, I have a feeling that the Java reference, as well as the "For the Internet and Intranets" subtitle was a clever bait-and-switch tactic to sell more copies of this book. If you're interested in the function of AI rather than the theory (if, suppose, you're progamming a game), then this book might be useful. But I have a feeling that better resources exist if this is your avenue of interest.
I have not, as of yet, done much course-based or indepenent research on expert systems, genetic algorithms, character recognition, natural language processing, or any of the other topics that are covered in this book. But I have a feeling that, if I do, I'll find the rest of Watson's material similarly lacking.