"Unpacking the hype around OpenAI’s rumored new Q* model"
I'll save everyone a click: The only actual information in this article is "Reuters says an anonymous source says Q* is capable of doing grade-school level math"
Does somebody have the links to the papers about replacing auto-regressive token prediction with planning that Yann LeCun talks about in the referenced tweet?
I would guess that is generally about DNN vs DRL (Deep Reinforcement Learning), but I'm not sure what the tweet you reference said. They are implemented with different algos & models ...hand wavy...