Particle.news
Download on the App Store

Sony AI’s ‘Ace’ Robot Beats Elite Table Tennis Players in Peer-Reviewed Tests

A Nature paper signals progress toward robot decisions at human speed in real play.

Overview

  • Sony AI’s study, published Wednesday in Nature, documents Ace winning three of five matches against elite players under official rules while losing two matches to Japanese professionals and taking a set from one.
  • Ace improved after the manuscript was filed, with December 2025 matches showing wins over two elite players and one professional as the team raised shot speed, placement near the table edge, and rally pace.
  • The system fuses nine high-speed image sensors with event-based “gaze” cameras and tilt mirrors to track the 40 mm ball, read spin in flight, and steer an eight‑joint arm trained by deep reinforcement learning.
  • Measured performance includes about 10.2 milliseconds to process ball trajectory and a consistent return rate above 75% for shots within defined spin thresholds, indicating control rather than brute force.
  • Tests ran on an Olympic-size table at Sony’s Tokyo site with accredited referees and named pros Minami Ando and Kakeru Sone, and Sony cites potential transfer to manufacturing and assistance along with dual-use risks.