> next token prediction Saying that LLMs just produce the next token is like saying that human brains just produce the next electrical impulse. If the algorith

> next token prediction Saying that LLMs just produce the next token is like saying that human brains just produce the next electrical impulse. If the algorithm that produces the next token (or electrical impulse) is complex enough, it can do anything that is in principle computable.