AMI Labs Co-founder Xie Saining: LLMs Are 'Anti-Bitter Lesson' to Some Extent, World Models Are the Future

Xie Saining, co-founder of AMI Labs, recently publicly criticized the current LLM route, arguing that it goes against Richard Sutton's 'Bitter Lesson' to some extent. He pointed out that language is a refined shortcut of human civilization, and over-reliance on LLMs will limit AI's learning of the real world.

World models differ from LLMs: LLMs predict the next token, while world models predict the next state based on actions, which can anticipate behavioral consequences and achieve safer and more controllable intelligence. He also said 'AGI is a pseudo-proposition', and building an agent that can survive in the real world is harder than solving math competitions or writing code.

Xie has rejected Ilya's invitations twice. AMI Labs is headquartered in Paris instead of Silicon Valley (as Silicon Valley is deeply stuck in the LLM route), and its short-term application scenarios will be AI smart glasses and robots.