current crop of LLMs still can't reason about abstract concepts, impressive but ultimately shallow