AI pioneer Rich Sutton warns synthetic data is a big mistake

In a Sequoia podcast episode, Turing Award recipient Rich Sutton criticized the AI sector’s growing reliance on synthetic training data, calling it a major error. He explained that artificially generated information, like computer-created images or fake records, cannot capture the complexity of real human behavior or physical environments, citing examples such as drone dynamics and motor wear. While major firms such as OpenAI seek large, proprietary datasets and Google recently paid $10 million for Spirit Airlines’ internal data, Sutton maintains that genuine experiential data—gathered by an agent interacting with the real world—is essential for true progress.

He suggested that the industry’s current trajectory overlooks this need. To put his ideas into practice, Sutton and his former student Khurram Javed launched Oak Lab, a startup focused on building agents that learn continuously from their own experiences rather than relying on pre-curated data. Oak Lab has not disclosed funding details.