Particle.news
Download on the App Store

Sony’s Ace Robot Reaches Expert-Level Table Tennis, Sometimes Beating Pros

A peer-reviewed Nature study signals a shift toward AI that performs split-second tasks in the real world.

Overview

  • Ace’s results, published Wednesday in Nature, document regulated matches on a full-size court with licensed umpires under International Table Tennis Federation rules.
  • In April 2025 trials, the robot won three of five against elite non‑pros and lost to two professionals, then recorded pro wins in December 2025 and March 2026, including a victory over top‑25 player Miyuu Kihara.
  • The robot uses nine synchronized cameras and event‑based vision plus a simulation‑trained control policy to read spin and place shots with about 20 milliseconds of latency compared with roughly 230 milliseconds for humans.
  • The hardware features an eight‑joint arm designed to control racket position, orientation, and shot speed so it can generate fast, high‑spin returns that challenge skilled opponents.
  • Outside experts note the system’s heavy external sensing gives it views humans lack, while Sony AI presents the project as groundwork for fast, safe human‑robot work in manufacturing, services, sports, and other safety‑critical settings.