According to the website of the Ministry of Industry and Information Technology on July 2, the Ministry of Industry and Information Technology and other four departments issued the "Guidelines for the Construction of a Comprehensive Standardization System for the National Artificial Intelligence Industry (2024 Edition)". It is proposed that by 2026, the linkage level of standards and industrial scientific and technological innovation will continue to improve, and more than 50 new national standards and industry standards will be formulated, leading the high-quality development of the artificial intelligence industry to accelerate the formation of a standard system. There are more than 1,000 enterprises that have carried out the publicity, implementation and promotion of standards, and the results of the innovation and development of standard service enterprises have become more prominent. Participated in the formulation of more than 20 international standards to promote the global development of the artificial intelligence industry.
Guidelines for the Construction of a Comprehensive Standardization System for the National Artificial Intelligence Industry (2024 Edition)
In order to thoroughly implement the deployment requirements of the Party Central Committee and the State Council on accelerating the development of artificial intelligence, implement the "National Standardization Development Outline" and the "Global Artificial Intelligence Governance Initiative", further strengthen the systematic planning of artificial intelligence standardization work, accelerate the construction of a standard system that meets the needs of the high-quality development of the artificial intelligence industry and the high-level empowerment of "artificial intelligence +", and consolidate the supporting role of standards in promoting technological progress, promoting enterprise development, leading industrial upgrading, and ensuring industrial safety. To better promote the new industrialization empowered by artificial intelligence, this guide is specially formulated.
First, the status quo of industrial development
Artificial intelligence is a basic and strategic technology leading a new round of scientific and technological revolution and industrial transformation, and is becoming an important engine for the development of new quality productivity, accelerating the deep integration with the real economy, comprehensively empowering new industrialization, and profoundly changing the industrial production model and economic development form, which will play an important supporting role in accelerating the construction of a manufacturing power, a network power and a digital China. The artificial intelligence industry chain includes four parts: the basic layer, the framework layer, the model layer, and the application layer. Among them, the basic layer mainly includes computing power, algorithms and data, the framework layer mainly refers to the deep learning framework and tools used for model development, the model layer mainly refers to technologies and products such as large models, and the application layer mainly refers to the application of artificial intelligence technology in industry scenarios. In recent years, the mainland's artificial intelligence industry has achieved rapid development in technological innovation, product creation and industrial application, forming a huge market scale. With the accelerated iteration of new technologies represented by large models, the artificial intelligence industry has shown new characteristics such as breakthroughs in innovative technology groups, integrated development of industry applications, and in-depth collaboration of international cooperation, and it is urgent to improve the standard system of the artificial intelligence industry.
Second, the general requirements
Guided by Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era, fully implement the spirit of the 20th National Congress of the Communist Party of China and the Second Plenary Session of the 20th Central Committee of the Communist Party of China, conscientiously implement the deployment requirements of the Central Economic Work Conference and the National Conference on the Promotion of New Industrialization, completely, accurately and comprehensively implement the new development concept, coordinate high-quality development and high-level security, accelerate the empowerment of new industrialization, and improve the top-level design of artificial intelligence standards with the goal of seizing the development opportunities of the artificial intelligence industry, strengthen the coordination of standard work of the whole industry chain, and coordinate the promotion of standard research. Formulation, implementation and internationalization to provide solid technical support for promoting the high-quality development of the mainland's artificial intelligence industry.
By 2026, the linkage level between standards and industrial scientific and technological innovation will continue to improve, and more than 50 new national standards and industry standards will be formulated, accelerating the formation of a standard system that leads the high-quality development of the artificial intelligence industry. There are more than 1,000 enterprises that have carried out the publicity, implementation and promotion of standards, and the results of the innovation and development of standard service enterprises have become more prominent. Participated in the formulation of more than 20 international standards to promote the global development of the artificial intelligence industry.
Adhere to innovation-driven. Optimize the linkage mechanism between industrial scientific and technological innovation and standardization, accelerate the research of key common technologies in the field of artificial intelligence, and promote the efficient transformation of advanced and applicable scientific and technological innovation achievements into standards.
Insist on applying traction. Adhere to the main body of the enterprise and market orientation, face the application needs of the industry, strengthen the iteration of innovation achievements and the construction of application scenarios, and collaboratively promote the integration and application of artificial intelligence and key industries.
Adhere to industrial synergy. Strengthen the coordination of the standardization work of the whole industrial chain of artificial intelligence, strengthen the collaboration of cross-industry and cross-field standardization technical organizations, and create a standardized model for the integration and development of large and medium-sized enterprises.
Adhere to openness and cooperation. Deepen exchanges and cooperation in international standardization, encourage mainland enterprises and institutions to actively participate in international standardization activities, and work together with upstream and downstream enterprises in the global industrial chain to jointly formulate international standards.
Third, the construction of ideas
(1) Artificial intelligence standard architecture
The AI standard architecture includes seven parts, including basic commonality, basic support, key technologies, intelligent products and services, enabling new industrialization, industry applications, and security/governance, as shown in Figure 1. Among them, the basic common standards are the basic, framework, and overall standards of artificial intelligence. The basic support standards mainly standardize the technical requirements of data, computing power, algorithms, etc., and lay a solid technical foundation for the development of the artificial intelligence industry. Key technical standards mainly regulate the technical requirements of artificial intelligence text, voice, images, as well as human-machine hybrid augmented intelligence, agents, cross-media intelligence, embodied intelligence, etc., to promote the innovation and application of artificial intelligence technology. The intelligent product and service standard mainly regulates the intelligent product and service model formed by artificial intelligence technology. The enabling new industrialization standard mainly regulates the technical requirements for artificial intelligence technology to empower the whole process of manufacturing industry and the technical requirements for intelligent upgrading of key industries. Industry application standards mainly standardize the technical requirements of artificial intelligence to empower various industries, and provide technical support for artificial intelligence to empower industry applications and promote the development of industrial intelligence. Security/governance standards mainly regulate the requirements of AI security and governance, and provide security guarantees for the development of the AI industry.
Figure 1 Architecture diagram of the AI standard
(2) The framework of the AI standard system
The AI standard system framework is mainly composed of seven parts: basic commonality, basic support, key technologies, intelligent products and services, enabling new industrialization, industry applications, and security/governance, as shown in Figure 2.
Figure 2 Framework diagram of the AI standard system
IV. Priority Direction
(1) Basic common standards
The basic common standards mainly include AI terminology, reference architecture, testing and evaluation, management, sustainability and other standards.
1. Terminology standards. Standardize the conceptual definitions of AI-related technologies and applications, and provide reference for the formulation of other standards and AI research, including standards such as definitions, scopes, and examples of AI-related terms.
2. Reference architecture standards. Standardize the logical relationships and interactions of AI-related technologies, applications, and systems, including standards such as AI reference architectures, AI system life cycles, and stakeholders.
3. Test evaluation criteria. Standardize the index requirements for testing and evaluation of the maturity of AI technology development, the adaptability between AI architectures, the development level of the industry, and the intelligent capabilities of enterprises, including the maturity assessment of AI-related service capabilities, the general testing guidelines, evaluation principles and level requirements of AI, and the framework and evaluation requirements of enterprise intelligent capabilities.
4. Management standards. Standardize the management requirements and evaluations of personnel and organizations involved in the entire life cycle of AI technologies, products, systems, services, etc., including management requirements for AI organizations, AI management systems, classification methods, rating processes, and other standards.
5. Sustainability standards. Standardize the technical framework, methods and indicators of AI impacting the environment, and balance industrial development and environmental protection, including the open source basic framework of AI software to promote ecological sustainability, the energy efficiency evaluation of AI systems, and standards for AI and resource utilization, carbon emissions, and waste parts disposal.
(2) Foundation support standards
The basic supporting standards mainly include basic data services, smart chips, smart sensors, computing equipment, computing centers, system software, development frameworks, software and hardware collaboration, and other standards.
1. Basic data service standards. Standardize the requirements for data services in the process of AI R&D, testing, and application, including standards for data collection, data labeling, data governance, and data quality.
2. Smart chip standards. Standardize the general technical requirements related to smart chips, including standards such as smart chip architecture, instruction set, unified programming interface and related test requirements, chip data formats and protocols.
3. Smart sensor standards. Standardize the interface protocols, performance evaluation, test methods and other technical requirements of single-modal and multi-modal new sensors, including the architecture, instructions, data formats, information extraction methods, information fusion methods, function integration methods, performance indicators and evaluation methods of intelligent sensors.
4. Calculate equipment standards. Standardize the technical requirements and test methods for computing devices such as AI acceleration cards, AI acceleration modules, and AI servers, as well as enabling software, including the virtualization method of AI computing devices, the interface protocol and test method of AI acceleration modules, and the access protocol, function, performance, and energy efficiency test methods of enabling software, as well as operation and maintenance requirements.
5. Hash center standards. Standardize the technical requirements and evaluation methods for infrastructure such as large-scale computing clusters, new data centers, intelligent computing centers, basic network communications, computing power networks, and data storage for artificial intelligence, including standards such as infrastructure reference architectures, computing capability assessments, technical requirements, stability requirements, and business service interfaces.
6. System software standards. Standardize the software and hardware technical requirements of the AI system layer, including software and hardware compiler architecture and optimization methods, AI operator libraries, chip software runtime libraries and debugging tools, AI software and hardware platform computing performance and other standards.
7. Develop framework standards. Standardize the technical requirements related to the AI development framework, including the functional requirements of the development framework, the interface protocol with the application system, the expression and compression of neural network models, and other standards.
8. Software and hardware collaboration standards. Standardize the adaptation requirements between hardware such as smart chips and computing devices and software such as system software and development frameworks, including the adaptation requirements between smart chips and development frameworks, the interaction protocols, execution efficiency, and collaborative performance of software and hardware collaborative tasks such as artificial intelligence computing task scheduling and distributed computing.
(3) Key technical standards
Key technical standards mainly include machine learning, knowledge graphs, large models, natural language processing, intelligent speech, computer vision, biometric recognition, human-machine hybrid augmented intelligence, agents, swarm intelligence, cross-media intelligence, embodied intelligence and other standards.
1. Machine learning standards. Standardize machine learning training data, data preprocessing, model expression and format, model effect evaluation, etc., including standards for self-supervised learning, unsupervised learning, semi-supervised learning, deep learning, and reinforcement learning.
2. Knowledge graph standards. Standardize the description, construction, operation and maintenance, sharing, management, and application of knowledge graphs, including knowledge representation and modeling, knowledge acquisition and storage, knowledge fusion and visualization, knowledge computing and management, knowledge graph quality evaluation and interconnection, knowledge graph delivery and application, knowledge graph system architecture and performance requirements, and other standards.
3. Large model standard. Standardize the technical requirements for large model training, inference, deployment, and other links, including standards such as general technical requirements for large models, evaluation indicators and methods, service capability maturity assessment, and generated content evaluation.
4. Natural language processing standards. Standardize the technical requirements and evaluation methods for language information extraction, text processing, and semantic processing in natural language processing, including standards such as syntax analysis, semantic understanding, semantic expression, machine translation, automatic summarization, automatic question answering, and language large models.
5. Intelligent voice standards. Standardize technical requirements and evaluation methods for front-end processing, speech processing, voice interfaces, data resources, etc., including standards such as deep synthesis forgery detection methods, full-duplex interaction, and large speech models.
6. Computer vision standards. Standardize technical requirements and evaluation methods for image acquisition, image/video processing, image content analysis, 3D computer vision, computational photography, cross-media fusion, etc., including standards for function, performance, and maintainability.
7. Biometric standards. Standardize technical requirements for biometric sample processing, biometric data protocols, equipment, or systems, including standards for biometric data exchange formats, interface protocols, and so forth.
8. Human-machine hybrid augmentation of intelligent standards. Standardize multi-channel, multi-modal, and multi-dimensional interaction paths, modes, methods, and technical requirements, including standards for brain-computer interface, online knowledge evolution, dynamic adaptation, dynamic recognition, human-computer collaborative perception, and human-computer collaborative decision-making and control.
9. Agent Standards. Standardize the technical requirements for agent instances and basic functions and application architecture with the general large model as the core, including standards such as agent reinforcement learning, multi-task decomposition, reasoning, prompt word engineering, agent data interface and parameter range, human-machine collaboration, agent autonomous operation, and multi-agent distributed consistency.
10. Swarm Intelligence Standards. Standardize the technical requirements and evaluation methods for control, formation, perception, planning, decision-making, and communication of swarm intelligence algorithms, including standards for autonomous control, cooperative control, task planning, path planning, collaborative decision-making, and networking communication.
11. Cross-media intelligence standards. Standardize the technical requirements for text, image, video, audio and other multimodal data processing basics, conversion analysis, fusion applications, etc., including standards for data acquisition and processing, modal conversion, modal alignment, fusion and collaboration, and application expansion.
12. Embodied Intelligence Standards. Standardize standards such as multimodal initiative and interaction, autonomous behavior learning, simulation, knowledge reasoning, embodied navigation, and group embodied intelligence.
(4) Standards for smart products and services
Intelligent product and service standards mainly include intelligent robots, intelligent vehicles, intelligent mobile terminals, digital humans, intelligent services and other standards.
1. Intelligent robot standards. Standardize the technical requirements for the application of artificial intelligence in the field of robotics, including standards for intelligent cognition and intelligent decision-making of robots.
2. Standards for intelligent launch vehicles. Standardize the technical requirements for perception, recognition and prediction, coordination and game, decision-making and control, and evaluation of intelligent vehicles, including standards for environmental fusion perception, intelligent identification and prediction, intelligent decision-making control, and multi-mode testing and evaluation.
3. Smart mobile terminal standards. Standardize the technical requirements for the application of artificial intelligence in the field of mobile terminals, including image recognition, face recognition, intelligent voice interaction, and standards for information accessibility and age-appropriate related to intelligent mobile terminals.
4. Digital human standards. Standardize technical requirements for digital humans, such as appearance, action generation, speech recognition and synthesis, and natural language interaction, including standards for the assessment of digital humans' basic abilities, multimedia synthesis and rendering, basic data collection methods, and identification and recognition methods.
5. Smart Service Standards. Standardize services based on AI technologies such as large models, natural language processing, intelligent speech, and computer vision, including standards such as technical requirements and evaluation methods for model-as-a-service platforms, as well as AI application services for specific scenarios, such as standards for intelligent software development, intelligent design, and intelligent anti-counterfeiting.
(5) Enabling new industrialization standards
The enabling new industrialization standards mainly include intelligent standards for the whole process of manufacturing such as R&D and design, pilot test verification, manufacturing, marketing services, and operation management, as well as intelligent upgrading standards for key industries.
1. R&D design standards. Develop standards for cross-domain knowledge integration, generation of new design models, and human-machine collaborative R&D and design.
2. Pilot test verification standards. Focusing on high-precision, full-process simulation models, we develop intelligent virtual pilot test standards and verification standards for new technology applications in complex industrial scenarios.
3. Manufacturing standards. Develop standards for intelligent production process, production line monitoring and maintenance.
4. Marketing service standards. Focusing on the improvement of marketing service efficiency, we will develop standards for intelligent customer service, digital humans, 3D models of goods, and user experience.
5. Operational management standards. Focusing on the improvement of intelligent operation management capabilities, relevant standards for supply chain management, data management, and risk management are developed.
6. Intelligent upgrade standards for key industries. Focusing on the raw material industry, we will carry out the development of standards such as large-scale model smooth connection production line data, optimized online monitoring and control, and process improvement. Focusing on the consumer goods industry, we carry out the development of standards such as demand forecasting and personalized customization. Focusing on the equipment industry, we will develop standards for intelligent equipment perception, interaction, control, collaboration, and independent decision-making.
(6) Industry application standards
Carry out standard research in the fields of smart city, scientific and intelligent computing, smart agriculture, smart energy, smart environmental protection, smart finance, smart logistics, smart education, smart medical care, smart transportation, and smart cultural tourism.
(7) Safety/governance standards
Security/governance standards mainly include security, governance and other standards in the field of artificial intelligence.
1. Safety standards. Standardize the security requirements for the entire life cycle of AI technologies, products, systems, applications, and services, including standards for basic security, data, algorithm, and model security, network, technology, and system security, security management and services, security testing and evaluation, security labeling, content identification, and product and application security.
2. Governance standards. Combined with the actual needs of AI governance, standardize the requirements for AI technology research and development and operation services, including technical requirements and evaluation methods for AI robustness, reliability, and traceability, and AI governance support technologies; Standardize the ethical governance requirements for the entire life cycle of artificial intelligence, including ethical risk assessment of artificial intelligence, technical requirements and evaluation methods for ethical governance such as fairness and explainability of artificial intelligence, and standards for ethical review of artificial intelligence.
5. Safeguard measures
(1) Improve organizational construction. Establish and improve the standardization technology organization in the field of artificial intelligence, coordinate the advantages of all parties involved in production, education, research and application, and all links in the industrial chain, coordinate the construction of artificial intelligence standards, and jointly build an advanced and applicable artificial intelligence industry standard system.
(2) Build a talent team. Encourage standardization research institutions to cultivate and introduce high-end standardization talents, and strengthen special training for standardization practitioners. Encourage enterprises, universities, research institutions, etc. to include standardized talents in the scope of professional ability evaluation and incentives, and build a standardized talent echelon.
(3) Strengthen publicity and promotion. Guide industry associations, standardization technology organizations, national technical standard innovation bases, etc., to carry out publicity and training on the artificial intelligence standard system and key standards for enterprises, guide enterprises to meet standards in R&D, design, production, management, testing and other links, and continue to improve the efficiency of standards to help the high-quality development of the industry.