Next token prediciton as a form of AI alignment mechanism.
If AGI related to LLMs, they will have some bit of us reflected in them from training on next token prediction
To talk like a humans, LLMs have (to some extent) think like a human
If AGI related to LLMs, they will have some bit of us reflected in them from training on next token prediction
To talk like a humans, LLMs have (to some extent) think like a human