Recursion, self reference, emergent properties are important topics and there are many good resources for studying them, GEB just isn't one of them. Most people that embark on the journey of reading the book fail and are unable to complete it. I believe the author tried to make the book itself some kind of self referential repeating work of art, which might be appealing to some but doesn't end up giving a rigorous treatment of the topic.
GEB is not a study book. It's written as a multi-layered musing on various inter-connected subjects. It's also not hard to read and entertaining. I get the feeling from the anecdote about the AI professor rejecting students based on having read it and your own reaction that people in the field resent it for not being course material? I also sense a sub-text of resentment about the positions and criticisms of the author about various approach to AI. It's clearly an exposition of his philosophical basis for his position. You might disagree with him but that's no reason to be insulting.
It's not so much resentment inasmuch all symbolic AI is evaporating in the shadow of the new new connectionist wave (the "new" connectionist wave being the PDP stuff).
Why not, for example, see it from Hinton's philosophical point of view (besides that the papers are very boring and dry, of course)? He made a very good defense against Pylyshin and Fodor and a few critiques of Hofstadter _and_ the algorithms are shipping on Google Search and things like that.
I read it when it first came out; I was 18, and it was formative in my early career in CS. For someone in their teens it was a wonderful introduction to things that my high school education never really touched.
I don't know if I'd have the patience to read it today.
