Bitter lesson redux: methods that generalize beat compute brute force, Sutton weighs in
ryanallred · x · 2026-11-09
alexwhedon argues the bitter lesson is not just about leveraging compute but a critique of methods: approaches that generalize and scale win, and modeling after human intuition wastes time.
He lists symptoms of today's weak methods: resources tapped out early in AI deployment, intelligence gains requiring model scaling faster than compute efficiency offsets, humans as the deployment bottleneck, extreme demand for human-curated data, and robotics stuck at task-specific intelligence. Richard Sutton, in the quoted tweet, counters that brute force alone is not enough — clever new methods are still needed and the science of AI is far from finished.
More from AGI Musings
- Dev reframes LLMs as "universal translators" translating between representations — brandon_xyzw · 2026-11-09
- The bitter lesson isn't 'naive wins' — it's whatever scales best with compute — pzakin · 2026-11-09
- Owain Evans: now is the right time to switch into AI safety work — OwainEvans_UK · 2026-11-09
- Henry Farrell: the "machine gods" metaphor shrinks AI policy debate and should be resisted — sebkrier · 2026-11-09
- Thomas Dietterich: map what LLMs can assess reliably, design UX for the rest — JessicaHullman · 2026-11-09
- Built a telco anomaly detection system in 3 days: 'the code doesn't matter, the cookbook is the product' — l4rz · 2026-11-09