Sony AI has made a groundbreaking achievement with the introduction of Ace, the first robot capable of outperforming elite human players in a physically demanding sport. Instead of excelling in digital realms or board games like chess or Go, Ace has conquered the challenge of translating advanced AI systems into the realm of physical performance. This accomplishment highlights the capacity of AI to demonstrate perception, planning, and control within milliseconds on a high-speed scale.
The intense and intricate nature of table tennis serves as a real-world test for robotics, demanding rapid decision-making, precise physical execution, and continuous adaptation to an unpredictable opponent. Sony AI’s Director in Zürich, Peter Dürr, emphasized that Ace’s success in competitive table tennis showcases how an autonomous robot can match or even surpass human reaction times and decision-making in a physical setting. This breakthrough underscores the potential for physical AI agents to excel in real-time interactive tasks, paving the way for robots with wider applications in fast, precise, and real-time human interactions.
Combining event-based vision sensors, model-free reinforcement learning-based control systems, and cutting-edge high-speed robot hardware, Ace boasts three innovative components crucial to its performance. Through rigorous evaluations against both elite and professional table tennis players, under ITTF regulations, Ace secured three victories out of five matches against elite players, showcasing competitive performances in the remaining matches.
The assessments revealed impressive capabilities of Ace, including its ability to return a diverse range of spins, achieving over a 75% return rate against spins at speeds up to 450 rad/s. Furthermore, Ace’s control systems facilitated quick responses to unconventional shots, such as balls rebounding off the net, showcasing its adaptability to rare and challenging scenarios in real-time play.
Subsequent to the Nature publication, Ace continued to impress by defeating professional players in matches conducted in December 2025 and March 2026. The robot displayed enhancements in shot speeds, aggressive placements closer to the table edge, and faster-paced rallies, illustrating ongoing performance improvements in competitive environments. Sony AI’s remarkable accomplishment with Ace signifies a significant advancement in the realm of physical AI agents, promising a future where robots can excel in diverse real-time interactive tasks.
