This enormous book is a hymn to the "strange loop", a term coined by the author. Loosely, a strange loop occurs when, after moving up a level in a conceptual hierachy, one is brought strangely back to where one started. It's closely related to those paradoxes of self-reference which can occur when form and content become intertwined.
An example is the old joke about the park keeper angry that his park has been littered with leaflets entitled "Keep Britain Tidy". Another is building one computer system to test another computer system, and then needing a third system to test the one you've just built. Yet another is the Wikipedia entry of Douglas Hofstadter which, at the time of writing, contains a quote from Hofstadter stating that his Wikipedia entry is full of inaccuracies. (So, do you trust the entry enough to believe this quote claiming it's unreliable?) You get the idea.
Hofstadter sees these strange loops everywhere: in the music of Bach, the art of Escher and, most significantly, Gödel's incompleteness theorem, in which an algebraic system is used to prove a result about itself (rather than about numbers). After he's presented the various variations on these ideas, he then moves on to Artificial Intelligence, examining the "state of the art" as he sees it and discussing the implications of the earlier material for this subject.
Along the way he delves into various other diverse subjects such as the structure of the human brain or the challenges of translating a novel into different (human) languages. Much of this is fascinating stuff and if you are mathematically inclined, there is plenty to love about this book.
Given all the above, why not give the man 5 stars - what more could one possibly ask for?! Well, personally I have a number of objections to this work which I'll mention briefly before the crowd throws rotten fruit at me. Firstly, I am not sure that *all* Hofstadter's examples are on the ball. For example, the loop in Bach's "endlessly rising" canon is simply a consequence of there being 12 semitones in an octave, rather than any subtle paradox of self-reference. Similarly, the main theme from Bach's Musical Offering is not "Babbage" backwards, however you push it! In short, I suspect the author's obsessions can cause him to see patterns in the world around him which aren't really there.
Secondly, his would-be humorous writing style, quirky and lively though it is, will not be to everyone's taste ("Why, you don't say, Mr T!"). Thirdly, some readers will wish he had been more honest up-front about the book really being about AI (and something of a polemic, as evidenced by his almost mean-spirited attack on the philosopher John Lucas in several places): personally, it's not a subject close to my heart and I would have been rather more interested in delving into, say, what makes Bach's music beautiful and spiritual, as the cover suggests we will be doing. And fourthly, and most seriously, I am not convinced that Hofstadter is that great a pedagogue: the facetious style and inordinate length of the book can serve to obscure, rather than illuminate, his meaning.
These niggles notwithstanding, this book really is a fine achievement and, if you have the time and inclination (you'll need both in spades), likely to be a very rewarding read.