
Unitree's founder said humans still outperform robots on most tasks. The comment came a day after his IPO surged 460%. A rival startup has shipped 100,000 units.
BEIJING – The biggest problem facing humanoid robots is that humans remain better at most jobs, Unitree founder Wang Xingxing said this week at the World Robot Conference in Beijing.
Robots need time to learn new tasks and are not as efficient as people, Wang said, speaking the day after his company's stock more than quadrupled on its IPO. Unitree shares fell 18.7% on Thursday.
The U.S. Federal Communications Commission last month added foreign-made advanced robotic devices, including humanoids, to an import restriction list. The FCC did not name a specific country and said retailers could still import previously approved models.
That limitation doesn't matter much for humanoids right now because few are deployed in the U.S., Jeff Burnstein, president of the Association for Advancing Automation, told CNBC. "Where the impact might be in the U.S. is on autonomous mobile robots, which are used in factories much more than humanoids are," he said.
Most companies looking to automate are not even focused on humanoids, Burnstein said. "What I hear from customers in the U.S. [is] 'we want solutions. We have a problem. We need a solution. We don't care if it's a humanoid. We don't care if it's a traditional robot, a collaborative robot. We don't even care if it's a robot. We need a solution,'" he said.
"So the onus is on the humanoid players to show we have a solution," Burnstein said, noting the robots must be affordable, safe and ready to deploy.
Startup Keenon builds humanoids that work alongside simpler delivery robots to handle hotel laundry services, COO Wan Bin said. The company's partners include Buffalo Wild Wings and Hilton. Completing 50% of a task well is straightforward, even 80%, he said. Getting to a 99.9% completion rate is where engineering and training get difficult.
Keenon has shipped more than 100,000 robots and expects to surpass 150,000 units by the end of next year, Wan said.
Prepared with AlphaScala editorial tooling from the source reporting linked above. Indexable analysis may include a cited Alpha Score value. Publishing checks screen each story before release. Educational coverage, not personalized advice.