Richard Sutton's 26-word Bitter Lesson gets a scaling-first rewrite
samsja19 · x · 2026-10-29
RL pioneer Richard Sutton restated the Bitter Lesson in 26 words: don't be distracted by human knowledge, as AI has been historically; instead focus on methods for creating knowledge that scale with computation, like search and learning.
Republisher samsja19 offers his own variant: don't be distracted by human knowledge or learning priors—focus on methods for creating intelligence that scale and generalize with computation and real-world data, citing gradient descent and test-time scaling. The exchange reopens the debate on scaling versus priors.
More from AGI Musings
- Will personal AI agents fail in China? Closed super-app ecosystems, not model smarts, are the wall — sujingshen · 2026-10-29
- Ant IPO at $2T would hand 2023-era employees $110M-$400M each in stock — TheZachMueller · 2026-10-29
- Thompson rips Zitron's two-year "peak AI" streak: like covering Intel's 1971 IPO with a Moore's Law denier — binarybits · 2026-10-29
- Gary Marcus mocks AI boom: "that's what they said in the tulip craze!" — GaryMarcus · 2026-10-29
- Inside AI's brutal race to solve a famous math problem — miniapeur · 2026-10-29
- Experts sell a lifetime of skill once as AI squeezes knowledge workers, no equity — curious_vii · 2026-10-29