China's robot beat Usain Bolt's record. Now comes the hard part

For five days in August, China staged the most expensive blooper reel in industrial history and called it the 11th World Robot Conference. More than 300 companies showcased 3,000-plus products along an outdoor street where apron-clad humanoid robots prepped fast foods, cotton candy and coffee orders for attendees. That 90 per cent of these showcases were Chinese, when only a decade ago Japan was considered the global leader in robotics, has triggered the usual insecurity of the West. Another 'race' that China is winning, they will tell you, as they solemnly file this as 'a frontier of US-China rivalry', with the sincerity of someone so deep in thought that you don't stop to question what the victory markers in this race even are. To be fair, there indeed was a race: A 100-m sprint where a robot first beat Usain Bolt's record, didn't stop at the finish line, crashed into a padded wall at full speed and needed to be carried off by handlers. And in the boxing ring, most robots heroically landed a swing several inches off their target. One machine raised its leg to kick and fell over first. The conference also featured robots dancing, performing tai chi, playing table tennis, tug of war, floor gymnastics, and all three soccer formats. All this while a deluge of state support and subsidies lures several players into the sector, artificially raising competition as they chip away at each other's market shares. Unitree Robotics, a company that rose to global fame last year when its robots danced to traditional Chinese folk music at the Spring Festival Gala, timed its stock-market debut immaculately. As the company's robots did more coordinated backflips, its shares rose nearly sixfold on the opening day. Indeed, technical capabilities in the sector are getting better. Only a few years ago, most robots didn't finish a 100-m sprint, and many that did needed a human chaperone, battery swaps and some remote assistance. Evidently the hardware is improving, but the positive returns from large-scale commercial deployment are still some time off. Robots are not yet nimble, dexterous, or intelligent enough to pull off even the most basic human tasks reliably. Last year, a robot took 17 minutes to pick up nine pieces of trash, a task that any human could easily have done in one-fourth the time, for instance. Training data for almost all tasks is still insufficient. But this is the bet the Chinese government is making: As China's population ages and shrinks, such a physical manifestation of AI will add productivity and contribute to economic growth. The truth buried under the Chinese state's pageantry and an onlooker's fearmongering (what if China deploys these machines on the border with India?) is mundane and, in its own way, more impressive. The field of robotics is advancing fast; China is working to indigenise the supply chain but struggles at the intelligence layer, and the pratfalls are exactly what you would expect from any hardware industry moving quickly: Expensive and embarrassing. Still, it will attract more talent into the sector, creating a flywheel effect that advances China's capabilities. This will look less like a breakthrough and more like incremental progress on metrics that are useful but also boring, and not amenable to visual entertainment. But until then, for technology enthusiasts, more entertainment is on its way. Reddy is a researcher at Takshashila Institution, Bengaluru. Views are personal