Richard Sutton says pure compute scaling is insufficient to achieve intelligence
Understand the story
Turing Award winner and author of The Bitter Lesson, Richard Sutton, published an article on November 8, 2026, stating that simply piling on compute is not enough to achieve intelligence, and that AI science needs clever new methods and is far from complete. Starting from his 2019 essay The Bitter Lesson, he criticized synthetic data as a "big mistake" and advocated building agents with world models that can continually learn from their own experience. The views have drawn mixed reactions in the industry: researchers such as Jerry Tworek agree that algorithms matter more as compute grows, while Sam Altman and Dario Amodei insist that LLM scaling is far from over. The event is currently at the stage of a debate of opinions, with all parties' positions clear; no further developments have been reported.
AI overview of reporting · Updated 2 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- The DecoderThe inventor of the bitter lesson has a bitter lesson for LLM scaling enthusiasts
Turing Award winner Richard Sutton has published an essay arguing that simply piling on compute is not enough to achieve intelligence, that AI science needs clever new methods, and that the field is far from finished. Starting from his 2019 essay "The Bitter Lesson," he criticized synthetic data as a "big mistake" and advocated building agents with world models that learn continuously from their own experience. Researchers such as Jerry Tworek agree that the more compute there is, the more important algorithms become, while Sam Altman and Dario Amodei insist LLM scaling is far from over.
Story heat trend
Not enough continuous observations to show a trend yet.