怎样的数据才算“高质量”？南京玄武：全国首笔具身智能数据集交易的背后

Core Insights - The article highlights the transition of the artificial intelligence industry from a "model-driven" approach to a "data-driven" one, emphasizing the increasing importance of high-quality datasets as a scarce resource for AI technology implementation [1][4] Group 1: Company Overview - Zhujing Intelligent Technology Co., Ltd. has developed a "embodied intelligence dataset" that was successfully traded at the Jiangsu Data Exchange, marking a significant milestone in the field [1][3] - The dataset includes approximately 25,000 structured data entries covering various scenarios such as office work, supermarkets, catering, and housekeeping, with each data entry lasting about 10 seconds and varying in size from tens to hundreds of megabytes [3][4] Group 2: Market Demand and Value - High-quality data products are becoming the focal point of market competition, characterized by high application value, high knowledge density, and high technical content [4][6] - Companies purchasing these datasets gain access to thoroughly cleaned and annotated data, significantly reducing the time and cost associated with building data collection environments and ensuring data quality [4][8] Group 3: Ecosystem Development - Jiangsu Province aims to build high-quality datasets across key sectors, with a target of 321 datasets and a total data scale exceeding 93PB by October 2025, equivalent to 93 million two-hour movies [6][11] - The region is fostering a data element industry ecosystem by establishing key infrastructures such as the Jiangsu International Data Port and the Jiangsu Data Exchange, promoting the understanding of data asset value among enterprises [6][8] Group 4: Standardization and Future Pathways - Standardization is viewed as a critical pathway for the construction of high-quality datasets, addressing practical pain points in data application and ensuring effective value realization [10][11] - The national strategy for high-quality dataset management includes a comprehensive framework focusing on infrastructure, construction entities, and application scenarios to support AI model development [11][13]