Xinhua Headlines: Out of the robotic class: China’s embodied AI handles real-world tasks

BEIJING, Sept. 28(Xinhua)– Graduation day is no longer simply for people.

Thirty robotics comprised the very first class at a robotic school in Hangzhou, east China’s Zhejiang Province. After more than 2 months of training, 4 have actually finished and landed official tasks.

One works as a museum guide in Hangzhou. Another provides trips at the robotic school. A 3rd carries out evaluations at a Nestlé production workshop in Tianjin. The 4th has actually taken a trip to the Republic of Korea for piano efficiencies.

Their release highlights an essential difficulty in bringing embodied AI into the real life: training robotics might need more than specific robotic makers or users can supply. It requires shared facilities where makers can be trained, evaluated and enhanced at scale.

In Xiong’an, a “city of the future” being developed about 105 km southwest of Beijing, various kinds of robotics are discovering day-to-day jobs such as securing the garbage, arranging racks, and moving items at a devoted training center.

Human trainers train robotics in major reproductions of genuine offices and public areas. They utilize virtual-reality headsets and exoskeletons to run robotics as they carry out jobs in simulated environments.

Preparing a human for a task is no simple job, and neither is training a robotic. “It took more than 1,000 repetitions just to teach the robot the basic movement of picking something up,” stated Liu Rui, a trainer with a digital-urban-technology subsidiary of China Xiong’an Group.

Came the complete series: selecting up a piece of garbage, standing upright and putting it into a bin. Knowing those 3 actions took another week.

Every effort likewise produces training information.

Trainers identify the information, break motions down into smaller sized actions and determine items in each frame. This assists robotics acknowledge what they see and choose how to act.

By the end of June, China had actually finished or brought into operation more than 70 robotic training centers, with over 40 others under building or on the drawing board, according to the China Academy of Information and Communications Technology.

This broadening network is significantly part of China’s wider AI and robotics community. In 2025, more than 6,200 AI business were running in the nation, with the core market valued at over 1.2 trillion yuan (around 173.9 billion U.S. dollars), according to the Ministry of Industry and Information Technology.

China likewise established more than 400 total humanoid robotic items in 2025, representing over half of the international overall, according to main information. In the very first half of 2026, Chinese producers represented 97 percent of humanoid robotic deliveries worldwide, according to a report launched at the 2026 World Robot Conference.

Jiang Lei, primary researcher at the National and Local Co-Built Humanoid Robotics Innovation Center, stated embodied AI robotics had actually moved from “zero to one,” Scaling from “one to 10” stayed a difficulty.

The traffic jams, he stated, depend on requirements, engineering management and pilot-scale screening. At the heart of all of it is the capability to adjust robotics to various real-world situations.

China’s huge production base provides robotic makers not just a big market, however likewise a large range of real-world and simulated environments in which to train, test and improve their makers.

A DATA MEGAFACTORY

The increase of robotic schools and training bases is turning training information into a financial resource– a crucial input for bringing embodied AI into the real life.

Big language designs and multimodal AI systems have actually advanced quickly by making use of built up written product and huge quantities of web information. Embodied AI, by contrast, should discover to act in the real world, which needs information produced through real-world interactions.

A factory, storage facility or dining establishment might require various levels of force, accuracy and motion. With a lot of situations to cover, getting adequate top quality training information stays pricey and tough.

Much of that information has actually never ever been methodically gathered, processed or kept, market experts state.

When China Xiong’an Group worked to incorporate its digital platform more carefully with the physical city in 2025, leading AI design business approached the group to buy premium datasets amounting to 10s of countless hours.

At the robotic school in Xiong’an, training datasets are utilized to train AI designs and fed into a metropolitan information area, where they can later on be shared and traded, stated Cao Yue, a supervisor at the group’s digital-urban-technology subsidiary.

Because sense, the robotic school is likewise a “megafactory” producing the information fuel for embodied AI training.

At a nationwide pilot base for embodied AI applications in Hangzhou, business can access calculating power, information, design services and situation screening through a public technical platform.

“Companies once spent tens of millions of yuan on training data alone to develop a single real-world application. Now, by joining the base, startups can tap national-level innovation resources in one place, significantly reducing R&D costs at the pilot stage,” stated Zhang Haiwei, creator of a robotics innovation business in Hangzhou.

The concept is uncomplicated: rather of every business developing its own training school, information pipeline and screening environment, shared centers can spread out the expense and make it much easier to turn laboratory models into devices that can really work.

THE HUMAN QUESTION

As more robotics move into the office– some after training at devoted robotic schools– another concern is coming forward: What occurs to human employees?

Yu Jiang, a scientist at the Chinese Academy of Sciences, stated technological transformations might remove some tasks, however likewise develop brand-new ones.

“When cars replaced horse-drawn carriages, for example, some coachmen became drivers,” he stated.

Xiong Rong, a teacher at Zhejiang University, stated robotics would progressively take control of recurring, exhausting and high-risk work, while individuals concentrate on management, innovation and development.

Previously this month, China included 11 professions to its list of formally acknowledged occupations, consisting of embodied AI robotic application service technicians and AI representative designers.

Specialists kept in mind that long-lasting knowing and employment reskilling are vital for employees to adjust to a brand-new work community that integrates human judgment, compassion and imagination with robotic accuracy and endurance. ■