Training a human takes much less money and data than training an LLM. Even if I spent my entire life doing it, I would not be able to read the entire corpus of any LLM, even counting just the English subset. LLMs need so much more because they start out as blank slates, while we don't.