Doing so, it has more positives than negatives; but of course this depends on whether you want an overview of the field, or an Intelligent Agent (IA) in "C" programming reference manual. It is more of the former than the latter.
It does cover a lot of ground and gives one a detailed taste for what artificial intelligence(AI), and IA's (Intelligent Agents) created from AI technologies are, and more importantly, what they *could* be. That is, if all the suggestions in this book were followed, we might soon have really 'intelligent' software emerging from various quarters (like the Internet or Intranets).
It concentrates on explaining the essentials of the cognitive and computer sciences that are relevant to IA design and creation, especially the considerations that seem to have contributed to 'intelligence' in the natural world (like our brains). I liked all the various definitions of intelligence!
It goes into the details of various architectural approaches to designing systems of interacting, or collaborating components. I found the material on the necessity of common agent languages (that is, languages in which the agents speak to each other) to be particularly enlightening.
Also especially useful was all the information about autonomy, agent mobility, standards that agents can make use of (like IIOP), and what languages and environments might be particularly suited to IA implementation. There were a couple of subjects that seem, in retrospect, out of place (like OpenDoc). But given that the coverage is about certain architectural and implementation concepts of those subjects that were important to IAs, and that it was only a few pages, this was a minor issue with me.
The book explains, at a surface level, the common soft-computing technologies like ANNs, genetic computing and fuzzy logi! c, as well as more traditional approaches like expert systems. It gives examples of real software that you can buy and use to incorporate these techniques into agents. For example, the book provides a good synopsis of Cyc, which can be used by agents to incorporate 'common sense' reasoning capabilities.
I did not expect it, so I was not surprised that this was not a programming manual. There are a couple of other books on IAs that concentrate on particular kinds of relatively simple agents in particular languages. Yet, to be honest, in lieu of lots of specific code and examples of actual agents, it provides a lot of pointers to other researchers' works in companies and academic settings.
This book is much more than what has appeared on the market thus far. Its breadth is actually pretty amazing considering its length of around 400 pages. I would recommend it to anyone who wanted to either gain a good grounding in intelligent agent design and development issues, or to expand one's purview of how intelligence could be enabled within today's and tomorrow's distributed computer systems.
List price: $23.95 (that's 30% off!)
List price: $19.95 (that's 20% off!)