The problem with reading Hofstadter is that, because he's so brilliant and so lucid, you'll often find yourself readily him uncritically: thoroughly convinced by anything he's saying, and simply absorbing it all.

Some of his books, the aforementioned in particular, are works of insight and genius. Others, such as his recent book on metaphor-as-thought, are speculative steps onto a bit thinner ice.

(Note: I realize Hofstadter and his colleagues have spent decades exploring metaphor-based AI, and that there might be something quite real to it -- but he doesn't make a convincing case for the sufficiency of this hypothesis, though he insists it's there).