Sutton Sets the Record Straight on the Bitter Lesson

Episode 1 · Bitter Lesson Author Sutton: Brute-Force Compute Alone Won't Produce Intelligence (2026-11-08, 25 posts)

On November 8, reinforcement learning pioneer and Turing Award winner Richard Sutton stated that brute-force computation alone is insufficient for intelligence, that clever new methods are still needed, and that the science of AI is far from done. As the author of The Bitter Lesson, his remarks were widely read as a correction to the misreading of his classic essay as a defense of pure compute scaling, and they promptly triggered an open debate among prominent researchers.

Confirmed

• Sutton explicitly said brute-force compute cannot produce intelligence by itself; AI science remains unfinished. Researchers and executives had frequently cited the Bitter Lesson to argue that more data and compute alone approach AGI.
• He also reaffirmed, in 26 words, the essay's original meaning: don't be distracted by human knowledge; focus on knowledge-creating methods that scale with compute, such as search and learning.
• When AI Street Talk countered that "the bitter lesson wins," Sutton quote-posted and held his position; he also added the contrarian view when someone reshared the original essay.
• Researcher Jayden Teoh wrote a long post arguing the Bitter Lesson is stifling academic research, with promising ideas dismissed as "not bitter enough."

Responses

• François Chollet: data scaling and compute scaling are only multipliers of a method's generalization, not intelligence itself.
• Omar Sanseviero (dair-ai) strongly agreed, citing over a decade of pre-LLM experience; reshares framed it as a signal the Transformer paradigm may end. Lukasz Kaiser agreed, adding that even with RLVR the current Transformer architecture is insufficient.
• John Langford noted the scale-invent-repeat cycle predates the debate, recalling 2011 claims that n-grams had solved language modeling.
• David Klindt clarified that the Bitter Lesson never meant ideas don't matter; good ideas beat bad ones at equal resources and improve scaling.
• Tarun Kathuria backed Sutton, noting that before o1-style reasoning models the field believed pre-training scale alone sufficed.
• Dileep George quote-posted with a chart rebutting Sutton; Arohan said young scientists who believed scaling alone suffices would be dismayed; Graham Neubig quipped that many dogma-holders never read the original essay; giffmana mocked; Rohit Krishnan's "Bitter bitter lesson" reshare implied skepticism; Pedro Domingos, Ben Goertzel, and Tkaraletsos also engaged.
• @sensho offered a counter-reading: most people mistakenly think Sutton means you can invent a new architecture beating LLMs, when the real bottleneck is the vast engineering grunt work behind existing scaling.

Why it matters

• Sutton is the author of the Bitter Lesson and a Turing Award winner; his "apostasy" directly collides with the mainstream scaling narrative. Chollet's multiplier framing gives the debate a sharper technical vocabulary, while Kaiser's judgment extends it to the sustainability of the Transformer paradigm. Teoh's essay and Arohan's remarks show the debate's real-world toll on academic research and young scientists, reflecting a genuine split in confidence over the scaling路线 across academia and investment.

• Richard Sutton: Brute force compute isn't enough, the science of AI is far from finished — RichardSSutton · 2026-11-08
• Richard Sutton pushes back on Bitter Lesson: brute force isn't enough for intelligence — davidmanheim · 2026-11-08
• Richard Sutton says brute-force compute isn't enough for intelligence, drawing pushback from scaling loyalists — PMinervini · 2026-11-08
• Sutton says brute-force compute won't yield intelligence; researcher laments scaling's toll on young scientists — eigenron · 2026-11-08
• Sutton: brute force isn't intelligence; Langford cites 2011 n-gram era to frame scale-invent cycle — weijie444 · 2026-11-08
• Rich Sutton walks back the Bitter Lesson: brute force isn't enough for intelligence — eigenron · 2026-11-08
• Sutton restates the Bitter Lesson; Dileep George fires back: 'don't fall for it' — anshulkundaje · 2026-11-08
• Chollet: Scaling Is Only a Multiplier, Smarter Methods Scale Better — fchollet · 2026-11-08
• Richard Sutton pushes back on the Bitter Lesson: brute force isn't enough for intelligence — dbreunig · 2026-11-08
• Richard Sutton: brute-force compute isn't enough, AI science far from finished — dair_ai · 2026-11-08
• Richard Sutton: Brute-Force Compute Isn't Enough, and the Bitter Lesson Says So — PMinervini · 2026-11-08
• Sutton: Brute force isn't intelligence — and Kaiser says we must move beyond Transformers — Sam_Witteveen · 2026-11-08
• Sutton fires back: brute force isn't enough, the science of AI is far from finished — chaitjo · 2026-11-08
• Richard Sutton says brute force compute is not enough, clever new AI methods are needed — npinto · 2026-11-08
• Richard Sutton: brute-force compute alone won't yield intelligence, AI science far from done — sloppenheimer · 2026-11-08
• Sutton pushes back on Bitter Lesson misuse: brute force alone won't yield intelligence — anshulkundaje · 2026-11-08
• Richard Sutton: Brute Force Compute Isn't Enough for Intelligence, Igniting Debate — josephdviviano · 2026-11-08
• Counter to Sutton: AI progress is bottlenecked by tedious scaling work, not clever new architectures — burny_tech · 2026-11-08
• Sutton pushes back on Bitter Lesson: brute force alone won't yield intelligence — SchmidhuberAI · 2026-11-08
• Sutton himself adds nuance to Bitter Lesson: brute force alone isn't enough — CatAstro_Piyush · 2026-11-09

5 more related posts →

Episode 2 · 'Bitter Lesson' Debate Reignites: Scaling Alone Isn't Enough (2026-11-09, 3 posts)

Richard Sutton's renewed comments on the 'Bitter Lesson' argue the essay opposes hard-coding human intuitions, not algorithmic ingenuity, and that scaling compute alone is insufficient—reigniting debate over whether current methods generalize well enough.

• Bitter lesson redux: methods that generalize beat compute brute force, Sutton weighs in — ryanallred · 2026-11-09
• Richard Sutton: brute force isn't enough — AI still needs clever new methods — RichardSSutton · 2026-11-10
• Richard Sutton's 'brute force isn't enough' gets the meme treatment — burny_tech · 2026-11-10