Sony AI Table Tennis Robot Beats Pros, Honor Wins Beijing Race
A table tennis robot created by Sony AI has taken on and beaten strong human competitors in formal contests. Reuters reports the machine belongs to the field known as physical AI, which involves AI controlling devices in actual settings.
Ace Robot Masters Table Tennis Challenges
Sony AI built the robot Ace to handle the demands of competitive table tennis. This sport calls for quick choices and exact movements. The team paired fast-sensing tech with AI controls to make shots during games.
Ace played matches following International Table Tennis Federation guidelines. Licensed referees oversaw the events. Records from April 2025 show it claimed victory in three of five bouts against top players and dropped two to pros. Sony AI noted further successes against professionals in December 2025 and early 2026.
Table tennis machines date back to the 1980s, yet none reached the level of expert humans before. Peter Dürr, director at Sony AI Zurich and project head, pointed out that while AI excels in digital games like chess, real-time physical sports such as table tennis pose tough hurdles.
AI thrives in simulated worlds like video games, Dürr added. His group aimed to test robot reactions in fast-changing scenes. They outlined the effort in a Nature journal paper.
The game tests limits with the ball's high speed, spins, and shifting paths. Players must sense and move fast in short windows. Ace relies on nine synced cameras and three vision setups to follow the ball's path and rotation. It handles visuals quick enough to catch details blurry to people. Dürr noted this speed grabs motion invisible to human sight.
The robot's frame has eight joints for the paddle. Three adjust position, two set angle, and three set power and pace. This setup covers basics for contest-level action.
Most AI learns by copying people, but Ace trained solely in virtual setups. This let it craft unique tactics unlike human styles. Dürr explained it picks up skills through solo practice in sims, not by observing players.
Pro Mayuka Taira, beaten by Ace, called it hard to foresee since it lacks tells. Elite Rui Takenaka, who split games with it, praised its spin handling but noted patterns on basic serves. Taira added the absence of feelings makes it tough to spot weaknesses. She said you cannot gauge shots it mishandles without reading its face.
Dürr highlighted Ace's skill at spin detection and swift replies. Current efforts boost match flexibility. The team sees the sensing and control methods working in factories and service bots.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Sony, founded in 1946 as a electronics giant, launched Sony AI in recent years to push boundaries in machine learning and robotics. Zurich lab focuses on advanced AI for real-world tasks.
Honor Humanoid Tops Beijing Half Marathon
Over 100 humanoid robots joined about 12,000 humans in the 2026 Beijing E-Town Humanoid Robot Half Marathon. Robots ran a 21-kilometer path separate from people.
Honor built Lightning, which crossed the line in 50 minutes and 26 seconds. That beat Olympic runner Jacob Kiplimo's 57 minutes and 20 seconds from the Lisbon Half Marathon in March. Lightning hit a barrier yet pressed on to win. Other Honor units grabbed second and third spots.
Times advanced from last year's top robot mark of two hours, 40 minutes, and 42 seconds. Planners meant the race to check robots in big, live settings.
Associated Press said a second Honor robot finished in 48 minutes via remote operation. Rules favored self-guided runs, so Lightning got the win.
Honor staff said features like sturdy builds and cooling fluids suit factory uses.
Honor, a Chinese firm known for smartphones since splitting from Huawei in 2020, expands into robotics with durable designs for tough conditions.