The Second Bitter Lesson: Sutton's thesis extends beyond models to the application layer
alexvoica · x · 2026-09-04
- Recaps Richard Sutton's 2019 The Bitter Lesson: hand-encoded human expertise repeatedly lost to general methods leveraging compute and data — from chess engines to computer vision and speech recognition.
- Sutton's conclusion: "general methods that leverage computation are ultimately the most effective, and by a large margin."
- The article argues generative AI already taught the industry this lesson at the model layer (search and learning scale with data/compute), and proposes a second bitter lesson now playing out at the application layer.
More from AGI Musings
- AI autoresearchers help crack 30-year-old coding theory problem, soundness up to 68.02 bits — BenBlaiszik · 2026-09-04
- Study of 1,131 Chatbot Users: Companionship Use Tied to Lower Well-Being — steverathje2 · 2026-09-04
- Yacine Proposes Monitoring Idle Compute Power Draw as an AI Escape Detector — yacineMTB · 2026-09-04
- From 6-Fingered Hand Memes to Bash Scripts: Reflecting on AI's Breathtaking Two-Year Leap — AIandDesign · 2026-09-04
- WIRED: Who cares if AI is conscious—it's basically alive, writes Steven Levy — nordicinst · 2026-09-04
- Engineers' value is judgment, not lines of code, in the AI era — iamKierraD · 2026-09-04