On September 4, 2024, the National AI Standardization Committee (SAC/TC28/SC42) convened a week-long meeting in Qingdao, unveiling the latest progress in China’s AI standardization efforts. Key updates include:
- SAC/TC28/SC42 has initiated, developed, and advanced over 30 national standards and launched 34 sector standards. These standards address foundational resources like intelligent computing power chips and training datasets, as well as critical technologies like industrial vision and autonomous driving. The SC has also collaborated with relevant industry authorities to advance the development of standards for large models in sectors such as steel, government services, healthcare, and finance.
- The national standard Artificial Intelligence – Performance Testing Methods for Server Systems is now in its final approval stage and is expected to be released soon.
- To create a unified, standardized benchmark evaluation system for large models, SAC/TC28/SC42 has developed the national standard Artificial Intelligence – Large-Scale Models – Part 2: Evaluation Metrics and Methods, which has completed testing and validation and is now undergoing technical review.
- To implement the Guidelines for Developing National Comprehensive Standardization System of Artificial Intelligence Industry (2024 Edition), which outlines tasks to empower new industrialization and industry applications, SAC/TC28/SC42 has drafted 12 AI sector standards in collaboration with leading organizations in industries such as steel, government services, mining, transportation, and healthcare. The standards include:
- AI – Large Model Training Data Processing Workflow and Quality Evaluation Model
- AI – Technical Specifications for High-Speed Imaging Analysis System
- AI – Technical Specifications for Industrial Vision Model Training and Inference Platform
- AI – Technical Specifications for Computer Vision-oriented Autonomous Learning System
- AI – Technical Requirements for Visual Large Model System in Logistics Parks
- AI – Technical Specifications for Deepfake Voice Identification System
- AI – Testing Methods for Neural Network Model Compression Performance
- AI – Neural Network Model Conversion Guidelines
- Technical Requirements for AI Visual Data Automatic Labeling System
- AI – Technical Requirements for Large Models in Steel Industry
- AI – Technical Requirements for Knowledge Graph Construction in Steel Production Processes
- AI – Technical Requirements for Intelligent Decision-Making Data Processing and Model Construction in Steel Production and Smelting Process
During the meeting, SAC/TC28/SC42 also established four working groups dedicated to AI application standards in the steel, logistics, communications, and construction industries.
Looking ahead, SAC/TC28/SC42 outlined plans to strengthen the development and promotion of standards for AI hardware and software, large model evaluation, and AI applications in critical sectors such as steel, telecommunications, construction, healthcare, logistics, power, coal, and petrochemicals.
