Table tennis robot defeats some of world’s best players – why this has major implications for robotics

Sony AI's table tennis robot Ace defeated elite human players in competitive matches, winning three out of five against athletes with over a decade of experience. Ace uses event-based vision sensors and high-speed cameras to track ball spin at up to 9,000 rpm, deep reinforcement learning trained in simulation for real-time decision-making, and a high-performance robotic arm capable of returning balls at 20 m/s. A key achievement is narrowing the sim-to-real gap — Ace adapted mid-rally when a ball clipped the net, demonstrating reliable performance under real-world uncertainty. The implications extend beyond table tennis to manufacturing, healthcare, and human-robot collaboration in unstructured environments.