On March 5, 2025, the Association for Computing Machinery named Andrew G. Barto and Richard S. Sutton the recipients of the 2024 ACM A.M. Turing Award “for developing the conceptual and algorithmic foundations of reinforcement learning.” The Turing Award is often called the Nobel Prize of Computing and carries a one-million-dollar prize funded by Google. At the time of the award, Barto was Professor Emeritus at the University of Massachusetts Amherst, and Sutton was a professor at the University of Alberta in Canada.
Barto and Sutton built their partnership in the late 1970s and 1980s, when Barto was Sutton’s PhD advisor at UMass Amherst. In a series of papers starting in that period they introduced the main ideas of reinforcement learning, gave it a mathematical foundation rooted in Markov decision processes and dynamic programming, and developed core algorithms. Their 1981 work showed how temporal-difference learning could explain animal-learning behaviors that earlier models could not, and they later developed actor-critic methods that separate the policy from the value estimate.
Their 1998 textbook “Reinforcement Learning: An Introduction” became the standard reference for the field and helped a generation of researchers learn the subject. The committee credited their work as the groundwork for the breakthroughs that followed once reinforcement learning was combined with deep learning, including DeepMind’s Atari-playing Deep Q-Network and AlphaGo.
The US National Science Foundation noted that it had sustained Barto’s research over four decades through programs including the National Robotics Initiative and Robust Intelligence, a reminder that much of the foundational work predated the commercial AI boom by many years.