A Turing Award winner says the industry’s fix for running out of data is ‘a big mistake’

Richard Sutton, a 2024 Turing Award laureate, argues that relying on synthetic data to compensate for dwindling human-generated training material is fundamentally flawed because simulations cannot capture the full complexity of human minds or the physical world. He cites research showing model collapse when synthetic data replaces human data, and stresses that genuine experiential data collected by agents interacting with their environments is the only viable alternative.