> And no users which are facing a LLM today have been trained on next token prediction when they were babies. That's conjecture actually, see predictive coding

> And no users which are facing a LLM today have been trained on next token prediction when they were babies. That's conjecture actually, see predictive coding. Note that "tokens" don't have to be language tokens.