Artificial Intelligence: A New Synthesis by Nils J. Nilsson

Artificial Intelligence: A New Synthesis Reviews

• 
  Stefan Berteau

  A very solid reference guide which covers simple implementations in many different branches associated with AI. It does not deal with fuzzy logic, but has good coverage of the propositional and first order predicate calculuses, feedforward neural networks (but not feedback architectures), bayesian probability inference and bayes nets, and genetic algorithms, among other things. I often use it as a reference when I have to dabble in an area that's new to me.
• 
  Franck Chauvel

  This—not really—new synthesis on Artificial Intelligence, remains a good summary. It proceeds gradually, from building simple agents that apply rules to more complicated ones that deduce knowledge, prove, learn, plan and could eventually self-organise.
  
  In my view, it lacks a chapter about control theory, which is also a solution to the "what to do" question and techniques such as Kalman Filters have also their place. I also miss a complete case study that would illustrate how we integrate these techniques (also the last chapter briefly touches upon that).
• 
  Gregor Erbach

  Nilsson's book presents all major areas of artificial intelligence from the unifying perspective of the problem of constructing an intelligent agent. Many of the important subfields of AI are introduced, including machine learning. The book is clearly written and can be understood if you have a good knowledge of computer science.