Sony’s AI-powered table tennis robot Ace defeated top-ranked human players in matches following official ITTF rules during tests in April 2025, according to a study published in Nature.
The robot, developed by Sony’s AI division, won three out of five matches against athletes with over a decade of training and later beat professional league competitors in December 2025 and March 2026.
Ace uses eight mechanical joints to control paddle position, orientation, and shot power, supported by nine cameras tracking the ball in 3D and three systems measuring its spin, and velocity.
How Sony’s robot overcomes the physical challenges of table tennis
Unlike board games where AI has long surpassed humans, table tennis demands split-second physical reactions to a fast-moving, spinning ball, requiring mechanical precision that mirrors human reflexes.
Sony’s system processes visual data from multiple angles to calculate trajectory in real time, allowing the robot to adjust its swing mid-motion — a capability absent in earlier models like Omron’s FOREPHUS, which only played amateurs at CES 2017.
Why this marks a shift in AI’s physical capabilities
The robot’s success suggests AI is closing the gap in dynamic, real-world environments where perception, decision-making, and action must occur in milliseconds.
Last time a machine demonstrated comparable physical dexterity in a regulated sport was when DeepMind’s agents learned to play simulated football, but Ace operates in physical space with actual equipment and rules.
What makes Ace different from earlier table tennis robots?
Ace is the first robot to compete successfully against top-tier human players under official ITTF rules, whereas prior models like Omron’s FOREPHUS only faced amateur opponents.
How does the robot track the ball’s movement?
It uses nine surrounding cameras to locate the ball in 3D space and three gaze control systems to measure its angular velocity and spin for accurate trajectory prediction.
