Sutskever and LeCun: Scaling LLMs Won’t Yield More Useful Results