I'm not that interested in LLMs as chat bots, especially not to "replace workers," but as mathematical models starting to be able to explain mid-level human behavior, they're really cool.