I've come across a use case LLMs seem inexplicably bad at: reading and understanding board game rules.
It seems like it should be something an LLM would be excellent at. After all, a rulebook is just a self contained set of instructions.
And yet, not only do LLMs fail to play a board game, I have yet to get one to successfully understand a rulebook enough to even answer basic rules questions.
This seems like a massive red flag overall for the status of AI. I'm wondering if this is an overcomeable issue, or if it speaks to the underlying limitations of AI in general.