Illustration: Liu Xidan/GT

Since its launch four months ago, the data collection and training facility at the Beijing Innovation Center of Humanoid Robotics has reportedly gathered more than 3 million data entries. This accumulation is not merely a statistical milestone; it provides a valuable insight into the internal workings of China's humanoid robotics ecosystem.

CCTV News reported on Tuesday that the facility has generated more than 300,000 open-source data entries in just the few months since its launch. A staff member at the center was quoted as saying that it could cut data collection costs for small and medium-sized robotics companies by more than 50 percent. This development underscores the growing accessibility of valuable data resources for smaller players in the industry, helping to lower barriers to entry and stimulate innovation within China's robotics sector.

To better understand the scale of this data collection, let's zoom in on a specific task at the innovation center - fruit washing. In preparation for the day when humanoid robots may assist with household chores, the center is conducting numerous repetitive actions to gather data, which is then used to train large-scale models.

According to a report by jiemian.com on March 21, an operator controls two robotic arms to pick up a piece of fruit, turn on the faucet, scrub off any dirt, and place the fruit back into a basket. Each robotic arm has six joints - three for horizontal movement and three for vertical movement - that align with the joints of a robot performing the same task. This process is repeated 500 times to generate 500 data entries, all of which help the model understand what it means to "clean fruit."

Fruit washing is just one of many data collection scenarios at the innovation center. According to jiemian.com, the center has also developed a range of other scenarios based on client needs, including changing diapers, organizing a toolbox, and transferring bottled water from the refrigerator to the checkout counter. Taken together, these diverse tasks position the center as a "data factory," capable of generating large volumes of training data on a production scale. 

One of the key advantages of centralized data production is cost reduction. As mentioned by CCTV News, small and medium-sized robotics companies can cut data collection costs by more than 50 percent. This significant saving makes it easier for these businesses to access valuable data and rapidly scale their development.

Small and medium-sized enterprises are often one of the driving forces of innovation in emerging industries. The humanoid robotics sector is expanding rapidly, and its growth hinges on the participation of such companies. By lowering the barriers to entry, this centralized data collection model not only opens the market to more businesses but also fosters greater innovation and activity within the industry.

Such data collection centers are not limited to Beijing. For example, take Zhangjiang Robot Valley in Shanghai's Pudong district, where a "super data factory" has been established, allowing humanoid robots to operate simultaneously. The facility has been described as a "university for robots." Inside the data center, a wide range of real-life environments - from homes and supermarkets to milk tea shops and fruit stands - are recreated, allowing robots to learn how to handle the complexities of human society. 

Moreover, data itself is evolving from a production factor into a tradable service. In cities such as Chengdu, in Southwest China's Sichuan Province, a growing number of companies are reportedly focusing on data collection, labeling, and training services, emerging as important link in the artificial intelligence (AI) value chain. This development means that data is no longer just an internal resource for companies, but is increasingly becoming a new type of infrastructure that can be supplied at scale.

The emergence of "data factories" and their associated upstream and downstream supply chains are becoming an integral part of the infrastructure supporting the development of humanoid robotics and other AI sectors. By providing cost-effective data, these centers help integrate more small and medium-sized enterprises, contributing to a vibrant and innovative industry.

At the Beijing innovation center, the repetitive task of a robot washing fruit may appear mundane, but it encapsulates a deeper logic behind China's AI-driven growth. China's AI industry is experiencing rapid, inclusive growth, with high cost efficiency enabling AI to quickly penetrate various sectors. This affordability likely stems from multiple factors, with the development of "data factories" playing a role in providing scalable and accessible resources to support industry growth.

The author is a reporter with the Global Times. bizopinion@globaltimes.com.cn