ABSTRACT
It might strike you as a bit odd to see this chapter on machine learning pop-up in a book concerned with the practical application of AI. Well, your first impression would be appropriate if this chapter reviewed the theories and mechanisms that are generally taken to constitute the AI sub-field of machine learning, but it will not. My goal in this chapter is to convince you that machine learning mechanisms will be a necessary part of many AI systems, and that this necessity brings with it additional aggravation for those of us that wish to engineer AI software. I do this not because I think that engineering artificially-intelligent software is a well understood and boring procedure which is in need of some injection of excitement—far from it. I do this because the potential power of machine learning mechanisms is something that we should be well aware of even though it introduces further problems on top of those that we already face. It may be wise, if not essential, to have the full picture in mind even though work on just one part of it gives us quite enough problems to be going on with.
