Five departments, including the central network information office, issued the guidelines for the construction of national new generation artificial intelligence standard system

In order to strengthen the standardized top-level design in the field of artificial intelligence, promote the technology research and development and standard formulation of artificial intelligence industry, and promote the healthy and sustainable development of the industry, recently, the National Standardization Management Committee, the central network information office, the national development and Reform Commission, the Ministry of science and technology, and the Ministry of industry and information technology jointly issued the following contents. < / P > < p > in order to implement the decision-making and deployment of the Party Central Committee and the State Council on the development of artificial intelligence, promote the continuous self optimization of artificial intelligence technology in the open-source and open industrial ecology, give full play to the leading role of standards in basic generality, ethics, security and privacy, guide the formulation, revision and coordination of national standards, industry standards and group standards of artificial intelligence to form standards To lead the new pattern of comprehensive and standardized development of artificial intelligence industry. < p > < p > fully implement the spirit of the 19th CPC National Congress and the second, third and fourth plenary sessions of the 19th CPC Central Committee, implement the decision-making and deployment of the CPC Central Committee and the State Council on the development of a new generation of artificial intelligence, and combine market driven and government guidance, in accordance with the principle of “overall planning, classified implementation, market driven, urgent needs first, cross-border integration, collaborative promotion, independent innovation, and open cooperation”, Based on the domestic demand and taking into account the international situation, we should establish a new generation of national artificial intelligence standard system, and strengthen the top-level design and macro guidance of standards. Accelerate the transformation of innovative technology and application to standards, strengthen the implementation and supervision of standards, and promote the deep integration of innovation achievements and industry. Focus on the coordination and matching with relevant standard systems such as intelligent manufacturing, industrial Internet, robot and Internet of vehicles. Deepen the international exchange and cooperation of artificial intelligence standards, pay attention to the coordination of international and domestic standards, give full play to the supporting and leading role of standards for the development of artificial intelligence, and escort the high-quality development. < p > < p > by 2021, clarify the top-level design of artificial intelligence standardization, study the general rules of standard system construction and standard development, clarify the relationship between standards, guide the orderly development of artificial intelligence standardization work, and complete the pre research work of more than 20 key standards such as key general technology, key field technology and ethics. < p > < p > by 2023, the artificial intelligence standard system will be initially established, focusing on the development of data, algorithms, systems, services and other key urgent need standards, and taking the lead in promoting key industries and fields such as manufacturing, transportation, finance, security, home furnishing, pension, environmental protection, education, medical health, justice, etc. Build artificial intelligence Standard Test and verification platform to provide public service capability. < p > < p > Artificial Intelligence standard architecture includes eight parts, such as “a basic Commonness”, “B supporting technology and products”, “C basic software and hardware platform”, “d key general technology”, “e key field technology”, “F products and services”, “G industry application” and “H safety / ethics”, as shown in Figure 1. < p > < p > among them, a basic common standards include terminology, reference architecture and test evaluation, which are located on the far left side of the standard architecture of artificial intelligence and support other parts of the standard architecture. < / P > < p > C basic software and hardware platform standards mainly focus on Intelligent chips, system software, development framework and other aspects to provide infrastructure support for artificial intelligence. < / P > < p > 0 > 0 > 0 > 0 D key general technical standards mainly focus on machine learning, knowledge mapping, brain like intelligent computing, quantum intelligent computing, pattern recognition and other aspects to provide general technical support for artificial intelligence applications; < / P > < p > e key technical standards mainly focus on natural language processing, intelligent speech, computer vision, biometric recognition, virtual reality / augmented reality, human In terms of computer interaction, it provides technical support in the field of artificial intelligence application; < / P > < p > G industry application standard is located at the top level of artificial intelligence standard architecture, which is oriented to the specific needs of the industry, refines other standards to support the development of various industries; < / P > < p > the framework of artificial intelligence standard system is mainly composed of basic commonness, supporting technologies and products, and basic software and hardware Platform, key general technology, key domain technology, product and service, industry application and safety / ethics are eight parts, as shown in Figure 2. Terminology Standard. It is used to unify the concepts, technologies and application scenarios of artificial intelligence, and provide support for the formulation of standards for other parts and artificial intelligence research of enterprises, including relevant definitions, categories and examples of artificial intelligence terms. < p > < p > 2. Refer to architecture standards. To standardize the logical relationship and interaction of artificial intelligence related technology, application and value chain, and provide positioning and direction suggestions for the development of artificial intelligence related standards. < p > < p > 3. The common requirements of testing and evaluation are extracted around the maturity of artificial intelligence technology development, industry development level and enterprise capability. It includes AI related service capability maturity assessment, AI generic testing guidelines, evaluation principles and level requirements, enterprise capability framework and evaluation requirements. The supporting technologies and product standards mainly include big data, Internet of things, cloud computing, edge computing, intelligent sensors, data storage and transmission equipment, as shown in Figure 4. < p > < p > 1. Big data standard. To standardize the data storage, processing, analysis and other supporting technical elements related to big data in the process of artificial intelligence development and application, including big data system products, data sharing and opening, data management mechanism, data governance and other standards. < p > < p > 2. To standardize the key technical elements of perception and execution involved in the development and application of artificial intelligence, and provide support for the collection, interaction and interconnection of all kinds of artificial intelligence perception information. It includes intelligent network interface such as intelligent perception equipment standard, interface and interoperability between perception equipment and artificial intelligence platform, integrated model standard of perception and execution, multimodal and situation awareness standard, etc. Cloud computing standard. To standardize the artificial intelligence oriented cloud computing platform, resources and services, and provide support for the storage, calculation and sharing of artificial intelligence information. It includes virtual and physical resource pooling and scheduling, intelligent computing platform architecture, intelligent computing resource definition and interface, application service deployment and other standards. Edge calculation standard. Standardize the end-to-end computing equipment, network, data and applications involved in artificial intelligence applications. Including data transmission interface protocol, intelligent data storage, end-to-end collaboration, end-to-end cloud collaboration and other standards. Intelligent sensor standard. To standardize high-precision sensors, new MEMS sensors, etc., to provide standard support for the development of artificial intelligence hardware, including sensor interface, performance evaluation, test methods and other standards. Smart chip standard. To standardize the intelligent computing chip, new sensing chip and related underlying interface to provide computational support for artificial intelligence model training and reasoning. Including instruction set and virtual instruction set, chip performance, power test requirements, data exchange format, chip operating system design and testing standards. System software standard. To standardize the software and hardware optimization compiler, artificial intelligence operator library, artificial intelligence software and hardware platform computing performance, and promote the collaborative optimization of software and hardware platform. The key general technical standards mainly include machine learning, knowledge mapping, brain like intelligent computing, quantum Intelligent Computing and pattern recognition, as shown in Figure 6. Machine learning standard. We should standardize different types of models, training data, knowledge base, expression and evaluation of supervised learning, unsupervised learning, semi supervised learning, integrated learning, deep learning and reinforcement learning. Knowledge mapping standard. To standardize the structural form, interpretation process and deep semantics of knowledge description, and solve the uncertainty of knowledge representation granularity and mode. Brain like intelligent computing standard. To standardize the basic model, performance and application of brain like computing algorithm, provide a new computing architecture for artificial intelligence system, and improve the ability of artificial intelligence to deal with complex problems. It includes brain like intelligent computing reference architecture, brain feature mechanism computing model modeling and expression, algorithm requirements and performance evaluation based on biological mechanism modeling, general technical requirements of brain like intelligent computing hardware equipment. Quantum intelligent computing standard. To standardize the basic model, performance and application of quantum computing algorithm, and to provide support for improving the computing capacity of artificial intelligence. It includes quantum computing model and algorithm, high performance and high bit rate quantum artificial intelligence processor, real-time quantum artificial intelligence system that can interact with the external environment. The technical standards in key fields mainly include natural language processing, intelligent speech, computer vision, biometric recognition, virtual reality / augmented reality, human-computer interaction, etc., as shown in Figure 7. Natural language processing standards. The technical requirements on the basis of natural language processing, information extraction and text content analysis are specified, and the consistency of data, analysis method and semantic description in the process of computer understanding and expressing natural language is solved. Natural language processing standard includes language information extraction, text processing, semantic processing and application expansion. Intelligent speech standard. To standardize the technology and method of human-computer language communication to ensure the accuracy, consistency, efficiency and usability of speech recognition, speech synthesis and their applications. The intelligent speech standard includes five parts: speech facilities, speech processing, speech recognition, speech synthesis and speech interface. Computer vision standard. It specifies the technical requirements for the detection, identification and tracking of the target by computer and visual perception equipment, and solves the problems of consistency and interconnection in the collection, processing, identification, understanding and feedback of pictures or videos. Computer vision standard includes three parts: visual facilities, data and models, image recognition and processing. Biometric recognition standard. To standardize the technical requirements of personal identification by computer using the inherent physiological or behavioral characteristics of human body, and solve the problem of consistency of biological characteristics description, data and interface. Virtual reality / augmented reality standard. To provide users with the general technical requirements of visual, tactile, auditory and other multi sensory information consistency experience. Human computer interaction standard. To standardize the multi-channel, multi-mode and multi-dimensional interaction ways, modes, methods and technical requirements of human and information system, solve the problems of integration, coordination and efficient application of multi-modal interaction such as voice, gesture, somatosensory, brain computer and so on, and ensure high reliability and security interaction mode. Human computer interaction standard includes intelligent perception, dynamic identification and multimodal interaction. Intelligent robot standard. Combined with the work deployment, in terms of service robots, the standards of hardware interface, safe use, multimodal interaction mode, function set, application operating system framework of service robot, and general requirements of service robot cloud platform are improved; in terms of industrial robots, standardization work is focused on path dynamic planning of industrial robots and design of collaborative robots 。 < p > < p > 2. General standard body for application of artificial intelligence technology in the field of intelligent vehicles