KI-Pioneer Sutton calls synthetic data a "big mistake" in the face of an infinitely complex world

Turing Award winner Richard Sutton has criticized the industry's increasing reliance on synthetic data for training large language models. He argues that because the world is infinitely complex, simulations are too limited to provide the necessary scale. Instead of relying on simulated environments, Sutton advocates for autonomous agents that learn continuously through direct interaction with the real world. This perspective highlights a potential strategic divide in AI research regarding the most effective methods for achieving future advancements in model capability.