萨顿:仅靠算力不足以产生智能
AI 总结
图灵奖得主理查德·萨顿(Richard Sutton)在 X 上发文称,单靠原始算力无法实现智能,AI 科学需要新的方法,该领域仍「远未完成」。《The Decoder》指出,萨顿此前就批评过单纯扩展大语言模型的局限,称合成数据是「一个巨大的错误」,并主张让带有世界模型的智能体从自身经验中持续学习。研究者 Jerry Tworek 与 Jayden Teoh 对此作出回应,而 OpenAI CEO Sam Altman 与 Anthropic CEO Dario Amodei 则认为 LLM 扩展远未触及极限。
深度见解
这则表态让「继续扩展 LLM 是否是正确路径」这一争论在 AI 社区重新升温,也为《苦涩的教训》作者本人提供了明确的反方论据。
The inventor of the bitter lesson has a bitter lesson for LLM scaling enthusiasts
Matthias Bastian View the LinkedIn Profile of Matthias Bastian
Nov 8, 2026

Nano Banana Pro prompted by THE DECODER
Key Points
- Turing Award winner Richard Sutton, author of the influential 2019 essay "The Bitter Lesson," writes in a new post that raw compute alone isn't enough for intelligence. AI science needs clever new methods and is "far from finished."
- Sutton had already criticized the limits of simply scaling large language models, called synthetic data a "big mistake," and called for agents with world models that continuously learn from their own experience.
- The real world is infinitely complex, Sutton argues, and human expertise as a bottleneck prevents true scaling.
#Richard Sutton#Oak Lab#Jerry Tworek#Core Automation#Jayden Teoh#Sam Altman#OpenAI#Dario Amodei#Anthropic