In a milestone for robotics, an artificial intelligence-powered robot has beaten elite table tennis players under official competition rules. Named Ace, the system developed by Sony AI won three out of five matches against top-tier players, though it lost both contests against professionals, clawing back only one game across seven sets.
The feat, published in the journal Nature on Wednesday, showcases Ace's ability to master spin and handle challenging shots, such as balls grazing the net. One rapid backspin shot was deemed impossible by a former Olympic player, Kinjiro Nakamura, who now believes humans could learn the technique.
Ace operates with an eight-jointed arm on a movable base, avoiding the need to stand on two legs. Instead of human-like vision, it uses multiple cameras around the court to track the ball's position and spin, even estimating rotation by zooming in on the logo during the milliseconds it takes the ball to arrive. Its skills were honed through 3,000 hours of simulated gameplay, with serves adapted from expert players.
Despite its prowess, Ace struggled early on with slow, low-spin balls, returning them weakly. Elite player Rui Takenaka noted that while Ace countered complex spin with complex returns, simple serves led to simpler returns, allowing her to attack effectively. Professor Jan Peters of the Technical University of Darmstadt called the project impressive but emphasised that table tennis research alone won't solve broader robotics challenges such as object manipulation.
