In mid-August, Beijing’s National Speed Skating Oval – the “Ice Ribbon” built for the 2022 Winter Olympics – hosted a new kind of sporting event. For four days, more than 500 humanoid robots from 16 countries sprinted, boxed, danced, and performed industrial tasks in what organizers billed as the world’s first “Humanoid Robot Games.” Chinese teams dominated, and domestic media hailed the event as a “new starting line for a shared human–machine future.”
The spectacle was real. Robots stood in starting positions, accelerated down the track, and moved in sync with upbeat march music. Crowds cheered, investors watched closely, and China showcased its engineering prowess. Yet behind the show lay a paradox: humanoid robots were running fast but still could not “see.” Most lacked the autonomy, perception, and decision-making that would make them genuinely useful.
The competition revealed less about the future of artificial intelligence (AI) than about the pressures shaping how China – and the world – presents technological progress.
AI’s recent breakthroughs have been in algorithms. Large language models generate persuasive essays and dialogues, but they are confined to the realm of text. They cannot sense temperature, weight, or space. Without bodies, AI systems remain statistical engines: powerful at prediction, but disconnected from the real world.
This limitation has fueled interest in “embodied intelligence,” which places algorithms inside machines that can move and sense. Robotics provides AI with presence, turning code into an actor in lived reality. But embodiment comes in many forms – robotic arms on assembly lines, quadrupeds for search and rescue, or tracked rovers for exploration.
The case for a human shape is less functional than social. Research shows that people instinctively respond to machines that mimic human gestures, expressions, and movements. A humanoid robot can lower psychological barriers, build trust, and foster more natural interaction. That makes humanoid design appealing for caregiving, education, and companionship – roles where social acceptance matters as much as technical performance.
But this also clarifies what humanoid robots are not for. Their advantage is not speed or strength but sociability. A caregiving robot must recognize when a child has fallen, or carry water safely to an elderly person – not run 100 meters faster than a wheeled machine.
The Humanoid Robot Games thus highlighted two core paradoxes: the sprint paradox and the autonomy paradox.
By staging sprinting contests, the event celebrated an inherently inefficient design. Wheels, tracks, and drones can outpace humans and humanoid robots alike. Forcing bipedal robots to race makes for drama, but misses the point of why humanoids exist.
Meanwhile, many competitors were not autonomous at all. Remote operators controlled boxing matches, and human handlers ran behind sprinting robots to keep them upright. The image of independent machines competing masked a simpler reality: most were elaborate remote-controlled devices, not intelligent actors.
The result was a striking inversion – robots that can mimic human legs but cannot see the world around them. Their weakest skills, perception and decision-making, were precisely those most obscured by the spectacle.
Why organize such games? Because humanoid robots are visually compelling. They capture headlines, attract investment, and symbolize progress. In China, this aligns with a broader pattern: mobilizing public excitement and private capital around “national priority” technologies.
But spectacle is double-edged. By framing humanoids as near-human athletes, media coverage risks inflating expectations about AI’s maturity. That can distort investment, steering resources toward flashy but shallow demonstrations rather than the slow, difficult work of developing autonomy, vision, and safe human–robot interaction.
The real contest for humanoid robots lies not in stadiums but in homes, hospitals, and schools. The critical tasks are mundane but transformative: recognizing when a child needs help, navigating a cluttered apartment, or assisting an elderly person without error.
These tasks demand perception, adaptability, and contextual understanding. A robot that can walk slowly but safely through a crowded room offers more value than one that sprints under remote control. If future competitions aim to advance the field, they would do better to test domestic abilities – delivering an object across obstacles, interpreting instructions in a noisy room, or providing safe physical support.
The games did reveal real progress. China has advanced in mechanical engineering, actuator design, and balance control. Building robots that can sprint – even imperfectly – is no small feat.
Yet the imbalance is clear. China excels at producing the “bodies” of robots, drawing on its manufacturing depth. But the “minds” – algorithms for vision, cognition, and independent reasoning – remain a comparative weakness. As in semiconductors or aviation, China risks mastering mid-range production while trailing in high-value design. Without breakthroughs in cognition, it may remain a manufacturer of parts rather than a leader in high-end robotics.
The Humanoid Robot Games offered a glimpse of speed and spectacle, but they also underscored what is missing. For embodied AI, the critical milestone will not be when robots run faster than humans. It will be when they can see, interpret, and act reliably in complex environments.
A sprinting robot may entertain. A caregiving robot that safely hands a glass of water to an elderly person may transform society. For China, and for the world, the challenge is to shift attention from running fast to seeing clearly – from performance to perception, from spectacle to substance.
That will determine whether humanoid robots remain Olympic curiosities or become trusted partners in everyday life.