This is one of the most insightful books about the nature of thought ever written. It's particularly interesting now, of course, as AI models are starting to s

This is one of the most insightful books about the nature of thought ever written. It's particularly interesting now, of course, as AI models are starting to solve some of the problems Hofstadter discusses but completely failing at others. There's an example problem in the book... "If abc were changed to abd, how would you change ijk in the same way?" A good answer is "ijl". We guess that the rule is "advance the last letter by one". I just asked ChatGPT and it said "ijd". This is a valid but boring and stupid answer. Another is "If abc were changed to abd, how would you change xyz in the same way?" Do you try to move z past the end of the alphabet and say "xya", or do you recognize the edge as significant and say "wyz"? ChatGPT again just says "xyd". You can make it generate the other answers by cuing it to "consider letter position" or "consider symmetry", but it also then explains its reasoning incorrectly. As we're trying to figure out what the difference is between actual creativity and mash-ups of earlier concepts, this book is basically required reading.