Watch Sony’s elite ping-pong robot beat top-ranked players

The harder story here isn't the win rate; it's the latency and perception problem Sony apparently solved. Competitive table tennis happens in under 200 milliseconds per exchange, which means the robot's vision, prediction, and actuation pipeline has to operate faster than most robotic systems can even register a stimulus.

The Physical Intelligence story from mid-April covered π0.7, a robot brain capable of generalizing to untrained tasks, and that work sits in the same broader push: closing the gap between narrow robotic demos and machines that handle real-world variability. Sony's Ace is a narrower system — purpose-built for one sport — but it represents the same underlying pressure on the field. Where π0.7 bets on generalization, Ace bets on depth within a constrained domain, and both approaches are producing results that would have seemed implausible five years ago. The two stories together suggest the robotics field is advancing on multiple fronts simultaneously rather than converging on a single architecture. Recent coverage here has otherwise been dominated by software AI, chips, and agentic coding tools, so this is a useful reminder that physical AI is moving on its own timeline.

Watch whether Sony publishes the match data and methodology in a peer-reviewed venue. If the win-rate claims hold up under independent testing against a defined ELO bracket, this becomes a reproducible benchmark; if Sony keeps the evaluation proprietary, the headline number stays marketing.