“LLMs are a mathematical model of language tokens. You give an LLM text, and it will give you a mathematically plausible response to that text.”