Zhang Zhiyong, a deputy to the National People s Congress, promoted the localization of AI hardware

Mondo Social Updated on 2024-03-05

On March 5, the second session of the 14th National People's Congress opened at the Great Hall of the People in Beijing. The reporter of Nandu Bay Finance Society learned that during the two sessions, Zhang Zhiyong, a representative of the National People's Congress and chairman of China Tower, submitted suggestions on large models, new energy vehicles, edge computing power and other content.

At present, technological breakthroughs represented by artificial intelligence models are widely used in various fields, promoting industrial upgrading and transformation, and triggering a worldwide research boom. Although there is a certain gap between China and the United States in terms of basic theories and original models, China is a big manufacturing country and has a sound industrial category and the urgent needs of high-end, digital and green industries, which provides a broader practical space for the innovation and development of large models.

In the "Suggestions on Promoting the High-quality Development of Industry Large Models", Zhang Zhiyong said that with the successive introduction of support policies at all levels, the development of large models will be encouraged from the aspects of computing power support, scenario opening, technological breakthroughs, and product ecology. However, at this stage, problems such as insufficient computing resources, low data quality, insufficient scenario openness, high application deployment costs, and model data security risks have formed certain constraints on the development of industry large models.

In this regard, Zhang Zhiyong put forward six suggestions: first, improve the supply capacity of computing resources; the second is to improve the level of industry data supply; the third is to accelerate the construction of innovative application scenarios; Fourth, reduce the cost of application deployment; fifth, strengthen the security of model data; Sixth, strengthen industrial policy support and personnel training. Among them, in order to improve the supply capacity of computing resources, Zhang Zhiyong suggested the establishment of a public scheduling platform for computing resources to promote the collaborative sharing of computing resources and improve the utilization efficiency of computing resources. At the same time, we will enhance our independent innovation capabilities, increase support for domestic hardware at the levels of policy guidance, brand promotion, and industry procurement, and promote the 100% localization of AI deep learning frameworks and AI hardware computing. Accelerate the construction of a single ultra-large-scale industry intelligent computing center that can provide training and inference integration.

At present, the echelon layout of "center + edge" of China's computing network is taking shape, especially with the rise of massive terminal connections and scenario-based applications, more than 75% of the data will be generated and processed on the edge side in the future, and the importance of edge computing infrastructure is gradually highlighted. Zhang Zhiyong mentioned in the "Suggestions on Further Promoting the Collaborative Deployment and Application of Edge Computing Power" that from the research situation, the current computing infrastructure construction is mainly concentrated in large-scale centralized data centers and computing hub nodes, and the construction of edge computing infrastructure has just started, and there are several challenges such as the lack of overall planning of edge computing infrastructure, the uneven matching of edge computing power supply and demand, the insufficient application of edge computing power, and the lack of co-construction and sharing of edge computing infrastructure.

Combined with the current development status and existing problems of computing power, Zhang Zhiyong put forward suggestions: strengthen the overall planning and construction of edge computing infrastructure, explore the establishment of a computing power supply and demand docking mechanism, increase the cultivation of innovative applications of edge computing power, and promote the co-construction and sharing of edge computing infrastructure.

With the continuous promotion of the "dual carbon" strategy, the development of new energy vehicles, electrochemical energy storage, electric bicycles and other fields continues to accelerate, driving the vigorous development of China's new energy lithium battery industry. "The best departments at all levels have done a lot of work, but at present, there are still weak links in the supervision of the whole life cycle of new energy lithium batteries in China, and the governance capacity is not fully matched with the scale and status of China's new energy industry. Zhang Zhiyong pointed out.

The use of information technology to monitor the operation of new energy vehicles and power batteries in real time is an effective means to strengthen supervision. Judging from the current progress, the full coverage of real-time monitoring and traceability of lithium batteries has not yet been realized; The existing platforms are not fully connected, and the application of information sharing and data mining is insufficient; In three aspects, such as the lack of coordination between upstream and downstream enterprises, it is also necessary to continue to make up for shortcomings, strengths and weaknesses.

To this end, Zhang Zhiyong pointed out in the "Suggestions on Further Strengthening the Monitoring and Traceability of the Whole Life Cycle of New Energy Lithium Batteries" that it is possible to achieve full coverage of real-time supervision of lithium batteries; Strengthen the docking of existing platform construction and deeply explore the value of data; Consolidate the main responsibilities of all parties in the market, encourage linkage and cooperation, and strengthen the digital guarantee of the industry.

Written by: Nandu Bay Finance Society reporter Kong Xueshao.

Related Pages