From an era of toy-model connectionist AI 20 years ago, a set of critiques that still hold some weight in pointing to basic symbolic structure and relationships our thinking exhibits that spicy autocomplete may categorically still be struggling with. What's lovely about this book (which probably isn't relevant overall and anymore) is the generosity and curiosity the author gives to the other side, in exploring and expanding on the models he's critiquing he's also bridging language and concepts between subfields rather than jumping precipitously to disagreement.