China is making significant strides in the development and application of large AI models across various industries, showcasing its prowess in the field during the 28th session of the Conference of the Parties to the United Nations Framework Convention on Climate Change (COP28) held in Dubai. With “FuXi-Subseasonal,” a cutting-edge AI model for weather forecasts, taking the spotlight, China stands as the second-largest contributor to the global development of large models.
China’s growing dominance in large models
China’s rapidly expanding portfolio of domestically developed large models is making waves on the international stage. According to a recent report by the Institute of Scientific and Technical Information of China, affiliated with the Ministry of Technology, the country ranks second globally in terms of the number of large models developed domestically. These models have garnered considerable influence in various industries.
Tao Qing, spokesperson for China’s Ministry of Industry and Information Technology (MIIT), emphasized the transformative power of AI, particularly large models, in driving innovation, extensive application, and global competition. The growing adoption of AI is reshaping production modes and economic structures across the manufacturing sector.
Revolutionizing traffic management with AI
One notable application of large models is in traffic management, as seen in the Beijing Economic-Technological Development Area (Beijing E-Town). Collaborating with Chinese tech giant Baidu, the area has harnessed spatiotemporal data, including roadside perception and internet data, to implement comprehensive signal control. Large traffic models have been used to assess traffic congestion possibilities, leading to automated optimization of signal timing.
This initiative has significantly reduced the average time it takes for vehicles to pass through traffic light intersections in the area by a remarkable 28 percent. It serves as a prime example of AI’s potential to enhance everyday life.
Vertical domains embrace large models
China’s large models have found their place in a wide array of vertical domains, from transportation to culture and tourism, education, and healthcare. These specialized large models bring domain-specific expertise to the table, offering targeted solutions to industry-specific problems.
Ctrip, a leading online travel agency in China, introduced “Ctrip Wendao,” a large model tailored to the tourism industry. This model supports natural language interaction, aiding users with hotel and tourist attraction reservations, trip planning, and other travel-related decisions.
iFLYTEK, a prominent AI and intelligent speech company, launched the “IFLYHEALTH” app, based on their medical large model, Spark Desk. This app offers features such as health self-assessment, medical examination report interpretation, and health record management, providing valuable health consulting services to users.
Additionally, Nanjing Agricultural University unveiled a large language model (LLM) for ancient books, capable of performing lexical analysis, information extraction, and even generating intelligent poetry. This LLM holds promise for applications in ancient book collation and digital entertainment.
Proliferation of large models in China
As of October 2023, over 250 companies and academic institutions in China have developed large models with more than 1 billion parameters each. This proliferation highlights the pivotal role of large models in various sectors. China’s status as the world’s second-largest economy, coupled with its abundant digital resources, positions it favorably for sophisticated and diverse large model applications.
Empowering large models with computing power
The development of large models relies heavily on substantial computing power for training and inference. Recognizing this, China is actively promoting its east-to-west computing resource transfer project, redistributing computing resources to the western regions of the country. This initiative aims to provide a conducive environment for intelligent computing centers in these regions.
In November 2023, the Qinghai Kunlun Artificial Intelligence Computing Center was inaugurated in northwest China’s Qinghai province, utilizing the Sanjiangyuan national big data base for resource management and operations. This marks a significant step in bolstering computing resources for AI development.
Furthermore, in December 2023, northwest China’s Ningxia Hui autonomous region embarked on its first large-scale computing center cluster construction project. Powered by domestic full-function graphics processing units (GPUs), this cluster is set to deliver computing power for large model training, inference, 3D graphics rendering, and more.
China has ambitious plans to achieve a total computing power of over 300 EFLOPS by 2025, with a target of 35 percent of this power dedicated to AI computing. This is part of an action plan for the high-quality development of computing power infrastructure, jointly issued by six Chinese authorities, including MIIT, in October 2023.
China’s rapid progress in the development and application of large AI models across various industries is reshaping the country’s economic landscape. These large models are demonstrating their ability to bring about transformative changes, from optimizing traffic management to revolutionizing healthcare and cultural preservation. With a robust commitment to computing power infrastructure, China is poised to maintain its leading position in the AI revolution, paving the way for innovative solutions and enhanced efficiency across sectors