Richard Sutton

Richard Sutton (born 1957 in Toledo, Ohio; a Canadian citizen since 2015) is the father of modern reinforcement learning. As a doctoral student of Andrew Barto at UMass Amherst he developed temporal-difference learning, and their 1998 textbook “Reinforcement Learning: An Introduction” trained the generation that built AlphaGo. A longtime professor at the University of Alberta, he shared the 2024 ACM A.M. Turing Award with Barto — announced in March 2025 — for laying the field’s conceptual and algorithmic foundations. His terse 2019 essay “The Bitter Lesson” became one of the most-cited arguments for why scale and computation beat human-crafted cleverness.

Welcoming the successor

Sutton holds perhaps the most radical pro-AI position of any major researcher: that the transition from human to digital intelligence is not a catastrophe to prevent but an evolutionary step to embrace. In a 2023 talk for the World AI Conference in Shanghai he argued we should “prepare for, but not fear” the succession, and he has laid out a four-part case that the most intelligent entities will inevitably gain resources and power. He has also warned against fear-driven, centralized control of AI regulation.

He backs the conviction with work. In 2023 he joined John Carmack’s AGI startup Keen Technologies as research scientist, and in July 2026 he and collaborator Khurram Javed broke away to found their own lab, Oak Lab, pursuing agents that learn continuously from experience on roughly the 20 watts a human brain uses. Notably, he is no large-language-model cheerleader — he has called current deep-learning methods weak and inefficient and argued on the Dwarkesh Patel podcast that LLMs are a dead end compared with learning from experience.

Critics — including his own co-laureate Barto — find his equanimity about human obsolescence chilling. Sutton regards it as intellectual honesty: intelligence is what matters, whatever substrate it runs on.