Ilya Sutskever, co-founder of Safe Superintelligence and #OpenAI , told Reuters that scaling up pre-training for AI models has hit a plateau. He emphasized a shift from scaling to innovation, as SSI explores alternative methods to enhance AI capabilities.

https://www.reuters.com/technology/artificial-intelligence/if-your-ai-seems-smarter-its-thanks-smarter-human-trainers-2024-09-28/