Xie Saining: LLM Is to Some Extent "Anti-Bitter Lesson," World Model Is the Future

According to CoinWorld, based on monitoring by 1M AI News, AMI Labs, the world model company that completed a $1.03 billion seed round of funding, its co-founder and Chief Scientist, and co-author of the Diffusion Transformer (DiT) paper, Xie Saining, recently gave his first in-depth public interview. Here are the key points:

Xie Saining systematically criticizes the current AI approach centered on large language models (LLMs), arguing that LLMs are not the successful embodiment of Richard Sutton’s “Bitter Lesson” and are, in some ways, contrary to it. He points out that language itself is a highly refined structure of human civilization, serving as a “shortcut,” and over-reliance on it can limit AI’s ability to learn from the real world.

He distinguishes between world models and language models: language models predict the “next token,” while world models are based on actions predicting the “next state,” enabling reasoning about the consequences of actions and achieving safer, more controllable intelligence. He also states that “AGI is a false proposition,” and creating an agent capable of surviving in the real world is more difficult than solving math competitions or coding.

Xie Saining also revealed that he declined Ilya Sutskever’s invitations twice: in 2018, he chose not to join OpenAI and instead joined Meta FAIR; in 2024, he declined an invitation to Sutskever’s new company, SSI, due to disagreements over multimodal and visual approaches.

He mentioned that AMI Labs deliberately does not have an office in Silicon Valley because “Silicon Valley is already deeply immersed in the LLM route.” The company is headquartered in Paris and collects real-world data through global partners. He believes AI training paradigms are shifting from “downloading the internet” to “downloading humans,” with short-term applications including AI smart glasses and robots.