If you're interested in AI and haven't read this (super influential) paper, I highly recommend it; at the very least, the first and last sections are very concise and interesting.

My primary goal here is to hear some expert's thoughts on these ideas in light of LLM's tremendously impressive capacity for generalization.

I'm _quite_ convinced by his points here, so would love to hear from some DL true believers on why he's mistaken! I haven't looked for modern responses in the literature yet, figuring HN is a bit more approachable ;)