Rich Sutton: Synthetic Data Is AI's 'Big Mistake'

Reinforcement learning pioneer Rich Sutton has declared the AI industry's growing reliance on synthetic data a "big mistake," a direct challenge to the prevailing scaling strategies of major labs. Sutton's critique, rooted in his long-held "bitter lesson" thesis, argues that computational power and real-world data, not human-curated or AI-generated facsimiles, are the only paths to general intelligence. This assertion arrives just as companies like Databricks and Scale AI are heavily investing in synthetic data generation platforms, positioning it as a key enabler for enterprise AI adoption and a solution to data scarcity, creating a fundamental schism in developmental philosophy.