A robot developed by Sony AI has defeated elite table tennis players in a landmark achievement for artificial intelligence in real-world sport. Named Ace, the system won three out of five matches against top-level players, though it lost both matches against professional competitors, securing only one game in those seven contests.
The matches were played under official competition rules, with Ace demonstrating advanced spin control, handling of difficult shots such as net cords, and executing a rapid backspin shot that a former Olympic player considered impossible. The research detailing Ace's performance was published in the journal Nature on Wednesday.
Peter Dürr, director of Sony AI in Zurich and project lead, stated that Ace has improved since the paper was submitted, defeating progressively stronger opponents. The robot uses an eight-jointed arm on a movable base and relies on multiple cameras to track the ball's position and spin, rather than human-like vision.
Ace's skills were refined through 3,000 hours of simulated gameplay. Early difficulties with slow, low-spin balls were overcome, though such serves proved effective against it. Elite player Rui Takenaka noted that Ace returned complex spin with similar spin, but simple knuckle serves led to simpler returns, allowing easier attacks.
Despite its prowess, Ace lacks human elements like eye contact or body language, and does not feel pressure during close games. Professor Jan Peters of the Technical University of Darmstadt called the project impressive but cautioned that table tennis robotics does not solve broader challenges like object manipulation. He predicted a transformative moment in robotics within the next decade.
