Overview
- Sony AI detailed Ace in a peer-reviewed Nature paper this week, presenting a rail-mounted robot arm that competed under official table tennis rules with licensed referees.
- In the published trials, the system beat three of five elite players yet lost most games to two professionals, recording one win in seven against them.
- The company says newer versions beat professionals in December 2025 and again last month, signaling gains beyond the results locked into the paper.
- The robot tracks a spinning ball with nine synchronized high-speed cameras and event-based sensors, then applies learned physics of gravity, friction and rotation to place consistent, high-spin returns.
- Independent reviewers credit Ace’s varied spin and reliability over brute speed and point out an advantage from external cameras around the table that human opponents do not have.