Particle.news
Download on the App Store

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

Experts see a leap for real-time robotics, with high cost plus narrow design as limits.

Overview

  • Ace, detailed in a Nature paper Wednesday, won several lab matches under official rules against top Japanese players.
  • The system uses twelve high‑speed cameras to read ball speed and spin and drives an eight‑joint robot arm to strike and return shots.
  • Its controller runs a reinforcement‑learning policy trained in simulation that updates decisions about every 32 milliseconds.
  • Measured performance includes successful returns on most heavy‑spin shots up to about 72 rotations per second and smashes near 59 km/h.
  • Independent researchers praise the feat yet call it task‑specific and resource‑intensive, noting a 49‑person team and warning that the control approach may waste energy and pose safety risks.