A popular way to explain how current LLMs work is to say that “all” they do is predict the next most likely word in a sentence. From one perspective, this is correct. Trained on all human language, the LLMs distilled … Continue reading →