Overview
- Ace’s peer-reviewed results, published Wednesday in Nature, document formal matches under international rules with licensed umpires and report continued gains through March 2026.
- Ace tracks the ball with nine high-speed cameras and event-based sensors that read spin, then uses an eight‑joint robotic arm and a simulation‑trained reinforcement‑learning policy to strike with about 20 milliseconds of end‑to‑end latency.
- In April 2025 tests, the robot won three of five matches against elite players but lost two matches to professionals, then in December 2025 and March 2026 it beat multiple pros, including top‑ranked Miyuu Kihara.
- Players and observers said the system pulled off shots they thought were impossible, yet some athletes adapted by exploiting simpler serves, highlighting both Ace’s superhuman spin reading and areas where humans still find openings.
- The team frames the work as a step for physical AI with uses in manufacturing, service robots and safety‑critical tasks, though today’s setup relies on a sensorized arena and external cameras and has prompted questions about security and military applications.