Sutton calls synthetic data a 'big mistake' for scaling

The Decoder · rss · 2026-08-20

RL pioneer Richard Sutton labels synthetic data a 'big mistake' for scaling LLMs. He argues the world is infinitely complex and simulations are 'microscopic,' with human expertise acting as a bottleneck. He advocates for agents that learn continually from their own experience instead of relying on frozen models.

Original post →

More from AGI Musings

AGI Musings channel →