How does the text generated by LLM make “our” understanding deeper compared to text written in the books?