China’s Robots Surge Forward

This is The Stepback, a weekly newsletter diving into a key story from the tech realm. For insights on falling robots and the US-China AI rivalry, keep an eye on the latest updates. The Stepback lands in subscribers’ inboxes every Sunday at 8 AM ET.

How it started
I’ve previously shared my love for robot fail videos, and the World Humanoid Robot Games certainly delivered on that front.

The five-day event, which concluded this week in Beijing, was packed with drama. One robot, crafted by a smartphone manufacturer, lost a leg mid-sprint, while others sparked and stumbled spectacularly. Another competitor ran straight into a crash mat, bounced off, and ignited in flames. During the weightlifting event, a robot lost its balance under a modest barbell and toppled into the judges’ table, arms flailing comically. Robotic cheerleaders tumbled through their routines, and one bot fell off a platform in a manner that, if it were human, might have looked suspiciously like it was distracted by the performance.

This marked only the second World Humanoid Robot Games, yet it was already a significant leap from the inaugural event last year. The previous year’s records were underwhelming, with the winning 100-meter sprint clocking in at a sluggish 21.50 seconds and the highest standing high jump barely reaching 0.95 meters. True to China’s flair for turning robotics into public spectacles — from intricate dances at festivals to demonstrations at major tech events — both this year’s and last year’s games were held at Beijing’s National Speed Skating Oval, a former Olympic venue.

And what a spectacle it was! In most sports, competitors losing limbs, crashing into walls, catching fire, and being carried off on stretchers would likely lead to a halt in the action. But at the Humanoid Robot Games, that was just part of the show.

Beyond the fiery mishaps, the games featured some unusual events. In addition to the typical sprinting, football, boxing, and weightlifting, there were quirky competitions like bed-making, parcel delivery, shelf stacking, clothes folding, and hammering nails. Notably, there was even a firefighting contest.

These unconventional events highlighted the true aim of the games: to transition robots from the lab to the unpredictable real-world environments their creators envision them working in. And China is clearly eager to take the lead in this arena.

How it’s going
Despite the spectacle, the games showcased significant advancements in robotics over recent years. Records were shattered, with one humanoid completing the 100-meter sprint in an astonishing 9.39 seconds, surpassing Usain Bolt’s 2009 world record. Another robot achieved a standing high jump of 2.88 meters, nearly half a meter above the human record that has stood since 1993. These performances are a stark contrast to robotics competitions from a decade ago, where even the most advanced machines struggled to maintain balance.

However, the games also underscored the distance still to cover before humanoids can be genuinely useful in everyday life. While the flashy events were captivating, the more mundane tasks were less exciting — I doubt activities like picking up soybeans with tweezers or block-building will ever become Olympic staples.

Many events required robots to operate autonomously — including gymnastics, football, tug of war, table tennis, and some track events — while others permitted human assistance through teleoperation, albeit with a scoring penalty. The dexterity displayed indicates impressive hardware improvements, but it’s clear there’s still a long way to go before these machines can reliably function independently in settings like factories or homes.

What happens next
Beijing has yet to confirm a third Humanoid Robot Games, but given the government’s enthusiasm for the industry, it would be surprising if another event doesn’t materialize. The city has already announced another humanoid half-marathon for April, and plans are in place to open the country’s first robot-run hotel next year.

China emerged as the clear champion of the games, positioning itself well for future competitions and the humanoid robot industry overall. The medal table was predominantly Chinese, with AgiBot and Tien Kung each racking up more medals than nearly all other competitors combined, excluding each other.

This dominance partly reflects the lack of international competition. Although the games were marketed as a global event, 641 of the 666 teams were Chinese, representing 157 companies and 200 universities. Notably absent were major US humanoid players like Tesla, Boston Dynamics, and Figure.

Nonetheless, the variety of Chinese institutions present in Beijing showcased the rapid maturation of the country’s robotics sector, even amid concerns about circular investment and a potential humanoid bubble.

It’s uncertain whether a stronger US presence at the games would have made a significant difference. The US still leads in cutting-edge AI and advanced chips, but China is quickly closing that gap while prioritizing embodied AI as a national strategy. In contrast, Washington has historically paid less attention to robotics and now lags behind China in nearly every aspect, including manufacturing and deployment. Some reports even suggest that recent efforts to ban foreign robots, aimed primarily at China, could backfire due to the reliance of domestic firms on Chinese components.

Catching up to China will require serious and coordinated effort, but with technology evolving rapidly, there’s still time to bridge the gap. Washington is already prioritizing other areas of AI in its competition with Beijing; robotics could be next on the list. Expect plenty of stumbles along the way, and hopefully, more entertaining robot fail videos too.

By the way
* DARPA still conducts robotics challenges, but they garnered much more attention a decade ago.
* When the US government banned foreign robots, it didn’t just target humanoids and quadrupeds; it included Roombas as well.
* “The brains may come from the US, but the body could very well be Made In China,” highlights the need for the US to strengthen its hardware supply chain.
* Unlike language models, creating an AI that can interact with the real world requires generating a lot of data, which is why companies are eager to film you doing chores or send someone to your home to capture those tasks.