The Beijing Municipal People's Government issued a notice on measures for the development of general artificial intelligence

Source: Official website of Beijing Municipal People's Government

Beijing Government Office [2023] No. 15

The people's governments of all districts, all committees, offices, and bureaus of the municipal government, and all municipal institutions:

With the approval of the municipal government, the "Several Measures for Promoting the Innovation and Development of General Artificial Intelligence in Beijing" are hereby issued to you, please implement it carefully in light of your actual situation.

General Office of Beijing Municipal People's Government

May 23, 2023

(This piece is publicly released)

Several Measures of Beijing to Promote the Innovation and Development of General Artificial Intelligence

In order to implement the "Implementation Plan for Beijing to Accelerate the Construction of a Globally Influential Artificial Intelligence Innovation Source (2023-2025)" and give full play to the guiding role of the government and the catalytic role of the innovation platform,

1. Improve the ability to coordinate and supply computing power resources

(1) Organize commercial computing power to meet urgent needs

Focus on giving full play to the advantages of the city's computing power resources, implement the computing power partnership plan, accelerate the collection of existing computing power by strengthening cooperation with cloud vendors, and clarify supply technical standards, software and hardware service requirements, computing power supply scale and support measures, and provide innovative The main body provides diversified high-quality inclusive computing power to ensure the computing power demand for artificial intelligence technology innovation and product research and development.

(2) Efficiently promote the construction of new computing power infrastructure

Incorporate new computing power construction projects into the computing power partnership plan, accelerate the construction of Beijing artificial intelligence public computing power center and Beijing digital economy computing power center in Haidian District and Chaoyang District, form a large-scale advanced computing power supply capacity, and support hundreds of billions of parameters A large number of large-scale language models, large-scale visual models, multi-modal large models, scientific computing large models, large-scale fine neural network simulation models, and brain-inspired neural network research and development.

(3) Build a unified multi-cloud computing power scheduling platform

In response to the demand for flexible computing power, through the construction of a multi-cloud computing power scheduling platform, unified management and unified operation of heterogeneous computing power environments are realized, which facilitates innovation subjects to seamlessly, economically and efficiently run various artificial intelligence computing tasks on different cloud environments. Further optimize the direct connection basic optical transmission network between this city and Tianjin, Hebei Province, Shanxi Province, Inner Mongolia Autonomous Region and other regional computing power clusters, and improve the integrated dispatching capability of computing power in the surrounding Beijing area.

2. Improve the supply capacity of high-quality data elements

(4) Collection of high-quality basic training data sets

Organize relevant institutions to integrate and clean Chinese pre-training data to form a safe and compliant open basic training data set; continue to expand multi-modal data sources, build high-quality text, pictures, audio, video and other large-scale model pre-training corpora, support in Data trading institutions established in accordance with the law carry out data circulation and transactions.

(5) Planning and building a data training base

Accelerate the construction of a pilot demonstration area for the basic system of data, explore the establishment of a data training base, promote the high-level opening of data elements, and improve the scale and quality of the city's artificial intelligence data annotation database. Encourage Internet platforms that carry out content information services to provide high-quality corpus data for application by innovative entities. Explore commercial scenario cooperation based on data contribution and model application.

(6) Build a crowdsourcing service platform for fine-grained annotation of datasets

In the form of crowdsourcing services, build a platform for fine-grained labeling of datasets, develop an intelligent cloud service system, and integrate related tool applications. Encourage and organize professionals from different disciplines to participate in labeling multimodal training data and instruction data to improve the quality of data sets. Research the platform incentive mechanism to promote the sustainable and healthy development of the platform.

3. System construction of large models and other general artificial intelligence technology systems

(7) Carry out research on large-scale model innovation algorithms and key technologies

Focusing on model construction, training, tuning alignment, inference deployment, etc., actively explore the innovation of basic model architecture, research large-scale model efficient parallel training technology and tuning methods such as cognitive reasoning, instruction learning, and human intention alignment, and develop support for tens of billions of parameters Efficient compression of model reasoning and end-to-side deployment technology form a complete and efficient technical system and encourage the ecological construction of open source technology.

(8) Strengthen the research and development of large model training data collection and management tools

Focusing on the links of "acquisition, storage, management, research, and use" of training data, research Internet data full-scale real-time update technology, multi-source heterogeneous data integration and classification methods, build data management platform related systems, and develop data cleaning, labeling, classification, Algorithms and tools such as annotation and content moderation.

(9) Build an open service platform for large model evaluation

Encourage third-party non-profit organizations to build multi-modal and multi-dimensional basic model evaluation benchmarks and evaluation methods; research artificial intelligence-assisted model evaluation algorithms, and develop multi-dimensional foundations including versatility, efficiency, intelligence, and robustness Model evaluation tool set; build an open service platform for large model evaluation, establish a fair and efficient adaptive evaluation system, and realize automatic adaptive evaluation of large models according to different goals and tasks.

(10) Build a large model basic software and hardware system

Support the development of a large-model distributed training system to realize efficient and automatic parallel training tasks. Develop a new generation of artificial intelligence compiler suitable for model training scenarios to realize automatic generation and automatic optimization of operators. Promote the extensive adaptation of artificial intelligence training reasoning chips and framework models, develop artificial intelligence chip evaluation systems, and realize automatic evaluation of basic software and hardware.

(11) Exploring new paths for general artificial intelligence

Develop a basic theoretical system for general artificial intelligence, strengthen basic theoretical research on the mathematical mechanism of artificial intelligence, autonomous collaboration, and decision-making, and explore new paths for general artificial intelligence such as general agents, embodied intelligence, and brain-like intelligence. Support value and causality-driven general agent research, create a unified theoretical framework system, rating standards, and testing platform, develop operating systems and programming languages, and promote the application of the underlying technical architecture of general agents. Promote the research and application of embodied intelligence systems, and make breakthroughs in the perception, cognition, and decision-making technologies of robots under complex conditions such as open environments, generalized scenarios, and continuous tasks. Support the exploration of brain-inspired intelligence, study core technologies such as brain neuron connection patterns, coding mechanisms, and information processing, and inspire new artificial neural network model modeling and training methods.

4. Promote the application of general artificial intelligence technology innovation scenarios

(12) Promoting demonstration applications in the field of government services

Focusing on government affairs consultation, policy services, handling upon receipt of complaints, government affairs, etc., use the advantages of artificial intelligence in terms of semantic understanding, autonomous learning and intelligent reasoning to improve the intelligent question and answer level of the government affairs consultation system and strengthen the "Beijing Policy" platform specification Management and precise service capabilities, assisting the citizen service hotline to efficiently respond to citizens' demands, and promoting precise guidance and efficient approval of government affairs.

(13) Explore demonstration applications in the medical field

Support qualified research-based medical institutions to extract intelligent guidance, auxiliary diagnosis, intelligent treatment and other scenarios, fully mine medical literature, medical knowledge graphs, medical images, biological indicators and other multi-modal medical data, and develop together with artificial intelligence innovation subjects Intelligent application realizes accurate identification and prediction of symptoms, signs and special diseases, and improves the intelligent level of disease diagnosis, treatment, prevention and whole course management.

(14) Explore demonstration applications in the field of scientific research

Develop scientific intelligence and accelerate scientific research in the field of new materials and innovative drugs empowered by artificial intelligence technology. Support laboratories in the fields of energy, materials, and biology to set up special scientific research cooperation projects, carry out joint research and development with artificial intelligence innovation subjects, fully mine experimental data in the fields of materials, proteins, and molecular drugs, develop scientific computing models, and develop new alloy materials, protein sequences, and innovations Sequence prediction of medicinal chemical structures shortens the cycle of scientific research experiments.

(15) Promoting demonstration applications in the financial field

The system layout "revealing the list and leading" project will promote financial institutions to further open industry application scenarios; support fintech innovation subjects to focus on intelligent risk control, intelligent investment advisory, intelligent customer service and other links, and develop financial professional long-text accurate analytical modeling technology and complex decision-making logic Integrating technology with model information processing to support investment decision-making in the financial field.

(16) Explore demonstration applications in the field of autonomous driving

Support autonomous driving innovators to develop multi-modal fusion sensing technology, based on vehicle-road coordination data and vehicle driving multi-sensor fusion data, improve the multi-dimensional perception prediction performance of autonomous driving models, effectively solve the long-tail problem of complex scenes, and help improve the pan- ability. Support the opening of the vehicle-road collaborative autonomous driving data set in the construction of the Beijing High-Level Autonomous Driving Demonstration Zone 3.0 project. Carry out cloud-controlled autonomous driving model tests based on low-latency communication, and explore the path of new technologies for autonomous driving.

(17) Promoting demonstration applications in the field of urban governance

Support artificial intelligence innovation subjects to meet the needs of smart city construction scenarios, take the lead in applying large-scale model technology in the construction of urban brains, accelerate the research and development of multi-dimensional perception system fusion processing technology, and realize unified perception, correlation analysis and situation prediction of smart city bottom-level business, and provide urban governance Decision-making provides more comprehensive and comprehensive support.

V. Explore and create an inclusive and prudent regulatory environment

(18) Continue to promote the innovation of regulatory policies and regulatory processes

Explore the creation of a stable and inclusive regulatory environment, encourage innovation subjects to adopt safe and reliable software, tools, computing and data resources, and carry out independent innovation, promotion and application, and international cooperation of basic technologies such as artificial intelligence algorithms and frameworks. Strive to carry out pilot projects in the core area of the Zhongguancun National Independent Innovation Demonstration Zone, and promote the implementation of inclusive and prudent supervision pilots.

(19) Establish a normalized service and guidance mechanism

For artificial intelligence-related Internet information services with public opinion attributes or social mobilization capabilities, carry out normalized contact services, guide innovation subjects to introduce technical tools for security testing, apply for security assessments in accordance with regulations, and perform procedures such as algorithm filing.

(20) Strengthen network service security protection and personal data protection

Guide innovation entities to strengthen network and data security management, implement the responsibilities of network security, data security, and personal information protection entities, and strengthen the construction of security management systems and work implementation. Encourage innovative entities to carry out data security management certification and personal information protection certification, implement a data cross-border transmission security management system, and comprehensively improve network security and data security protection capabilities.

(21) Continue to strengthen scientific and technological ethics governance

Strengthen research on artificial intelligence ethics and security norms and social governance practices. Build a public service platform for scientific and technological ethics governance in the field of general artificial intelligence, serve government supervision, and promote industry self-discipline. Carry out scientific and technological ethics review and related business training, and strengthen the awareness of scientific and technological ethics norms of all responsible entities. Carry out in-depth education and publicity on science and technology ethics, and build a good atmosphere for artificial intelligence science and technology ethics.