It goes without saying that this book remains a landmark in the history of AI research: a sobering antidote to all the research hype.
But what a patchwork it has become. Every ten years, the book has a new section and a new introduction bolted on to it. But not just bolted on to the end, or on to the beginning.
No reader is given an easy path through the text: neither the reader who is new to the material (and who wants to read through the text in chronological order), nor the reader who has read the previous incarnation of the book (and who wants to pick up just the new parts).
The message of the book, too, is unfortunately weakened by its subjective stance, and its invocation of the personalities from the debate.
However, the book still remains a refreshingly different contribution to the world of AI research, an important warning, and a greatly worthwhile read.