First Robot Achieves Expert Level in Competitive Physical Sport

Sony AI has developed a robotic table tennis player called Ace that has become the first robot to achieve expert-level performance in a competitive physical sport, according to a study published in the journal Nature. The autonomous system won three out of five matches against elite human players with at least 10 years of experience and 20 hours of weekly training, and it defeated professional athletes Minami Ando and Kakeru Sone in follow-up competitions conducted after the initial study. The research was conducted in April 2025 at a custom-built Olympic-sized court at Sony's Tokyo headquarters, with official matches judged by licensed umpires from the Japanese Table Tennis Association following International Table Tennis Federation rules. Ace uses nine synchronized cameras and three vision systems to track spinning balls with exceptional accuracy, processing information fast enough to capture motion that would be imperceptible to the human eye. The robot was trained using reinforcement learning rather than manual programming, which researchers say allowed it to develop unexpected shot techniques. In matches against professional players who provided commentary, Mayuka Taira noted that Ace's inability to display emotional reactions made it difficult to anticipate its shots, while Olympic veteran Kinjiro Nakamura observed a return shot that he said no human player could have executed. The robot's architecture features eight joints enabling precise racket positioning and rapid response, and it improved its performance over subsequent months, defeating top-25 World Table Tennis-ranked player Miyuu Kihara in March 2026. The research represents what Sony AI president Michael Spranger called a "ChatGPT moment for robotics," demonstrating that robots can be trained to adapt in real-time physical environments rather than merely executing pre-programmed trajectories. Lead author Peter Dürr said the project's goal extended beyond table tennis to develop insights applicable to "manufacturing and service robotics, as well as applications across sports, entertainment and safety-critical physical domains." While researchers intentionally limited Ace's speed and reach to humanComparable levels to ensure fair competition, the technology's underlying capabilities in high-speed perception and real-time decision-making could have broader implications across multiple industries.