Embodied AI makes quantum leap in China

At around the 18-kilometer mark of the Beijing E-Town half marathon and humanoid robot half marathon, where the course narrows and the incline begins to bite, a humanoid robot broke its rhythm and did something unexpected. It accelerated.

Spectators along the roadside, many of them more accustomed to watching human runners grind through fatigue, leaned forward as the humanoid machine lengthened its stride and powered ahead.

Moments later, a humanoid robot named Lightning crossed the finish line in 50 minutes and 26 seconds, faster than the human world record holder, Uganda's Jacob Kiplimo, who finished the same distance in about 57 minutes at a road race in Lisbon in March.

But the real story was not the winner. It was a proving ground, a stress test, and a public demonstration of how far embodied artificial intelligence has come in China.

In a global context increasingly defined by technological rivalry, it offered a vivid snapshot of how China is positioning itself in competition with the United States in a race no longer confined to chips and algorithms, but one that extends into smart machines that move, perceive, and act in the real world.

At first glance, the spectacle carried a certain novelty. Dozens of humanoid robots, some with humanlike gaits and others still stiff and mechanical, lined up alongside support vehicles and engineers monitoring telemetry feeds. But beneath the theatrics lay a deeper shift.

A year earlier, in the inaugural event, nearly all robots required constant remote control. Engineers jogged alongside them, adjusting balance, correcting direction, and sometimes physically intervening to prevent falls. Only a handful of teams finished the course.

This year, the story has changed dramatically. More than 100 teams entered, a fivefold increase. Nearly 40 percent of the robots navigated autonomously, interpreting the course, adjusting stride, and managing energy consumption without direct human control. Operators followed in vehicles rather than running beside them.

Yan Weixin, chief scientist at the Shanghai Artificial Intelligence Research Institute and a professor at Shanghai Jiao Tong University, said the sharp improvement in robot half marathon performance is, at its core, a combined validation of hardware reliability, algorithmic stability, and system engineering capability. "It was not a breakthrough in any single technology, but the coordinated evolution of all of them."

The course of the marathon was deliberately unforgiving, combining flat stretches with inclines of up to 8 percent, descents, tight turns, narrow lanes, and even obstacles designed to mimic real urban disruptions.