You might try searching the EDN archives, I remember seeing a pretty good series of articles on the subject during the 90's (Au: Dave Brubaker?). I'm no expert on the topic, it seemed to have it's share of skeptics. Haven't seen much in the press about it lately.
My understanding of the subject is limited. I was left with the impression that the output one might generate with fuzzy logic is quite variable, centered around a norm or average - not just a single discrete and distinct output. It is in many ways like the answers a group of people might give to a question - they might be similar but won't be the same. What comes to my mind is monte carlo analysis where the inputs are variable and the output is the result of those varying inputs. Sorry I can't be clearer - just not enough time at the moment.