What is an LLM (intuition)
A non-mathy mental model of input → output.
Goal
Build an intuition for large language models—no neural-network deep dive required.
Core idea
Think of an LLM as a function that continues text: you give it words (and maybe images), and it predicts plausible next tokens based on patterns learned from data. It is not a lookup table and not always “searching the web”; mostly it is statistical continuation.
Implications for you
- Clearer requests reduce ambiguity about what you want.
- It can be confidently wrong—so verify and iterate instead of trusting a single shot.