KI-Pioneer Sutton calls synthetic data a "big mistake" in the face of an infinitely complex world
Turing Award winner Sutton: synthetic data is a dead end. The future is agents learning live.
Why it matters
A foundational AI researcher challenges the dominant scaling approach (synthetic data for LLMs) and argues for a competing paradigm (continual agent learning). This is lab-race strategy and capability philosophy — directly relevant to how practitioners think about model architecture and data strategy.
The key facts
5 to knowRichard Sutton (Turing Award winner) argues synthetic data is 'microscopic' relative to world complexity
Critique: human expertise becomes bottleneck for synthetic data scaling
Alternative thesis: agents learning continually from live experience, not frozen models
Frames as infinitely complex world vs. simulation problem
Published August 2026
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Turing Award winner Richard Sutton calls synthetic data a "big mistake" for scaling large language models. The world is infinitely complex, and any simulation of it is "microscopic," with human expertise acting as a bottleneck that blocks real scaling. Sutton's alternative is agents that learn…