China Builds 120,000 High-Quality Datasets, Exceeding 1565 Petabytes
China has significantly expanded its high-quality data resources, announcing the construction of 120,000 datasets across various sectors. As of the end of June this year, the total volume of these datasets, covering fields such as scientific research, industrial manufacturing, healthcare, and education, has surpassed 1565 petabytes (PB). This represents a growth of over 60% compared to the first quarter of the year. The current data volume is approximately 547 times the total digital resources of the National Library of China, providing substantial support for the development of artificial intelligence. Concurrently, China's data annotation industry is experiencing rapid growth, effectively supporting the creation of these high-quality datasets. Seven pilot cities, including Chengdu, Shenyang, Hefei, Changsha, Haikou, Baoding, and Datong, have achieved a data annotation scale exceeding 119 petabytes (PB), with 140,000 individuals employed in data annotation. Moving forward, the National Data Administration plans to further develop a high-quality data supply system tailored for artificial intelligence. This strategy involves using model applications to guide data supply and employing data to drive model iteration, aiming to establish a value loop for the paid use of datasets.
The substantial growth in China's high-quality datasets and the parallel expansion of its data annotation industry highlight a strategic national focus on leveraging data as a core asset for AI development. This initiative appears designed to create a robust domestic ecosystem for AI innovation, potentially reducing reliance on foreign data sources and fostering indigenous technological advancement. The emphasis on a 'value loop' for paid dataset usage suggests a move towards commercializing data assets and establishing market-based mechanisms for data exchange. This approach could incentivize further data production and refinement, while also raising questions about data accessibility, intellectual property rights, and the potential for market concentration within the AI sector in the coming decade.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.