China accelerating development of large language models
A staff member (1st L) introduces an AI character recognition system to visitors at the 2022 World Artificial Intelligence Conference (WAIC) in east China's Shanghai, Sept. 1, 2022. (Xinhua/Fang Zhe)
At the recently held 2023 Zhongguancun Forum, a generalist large language model (LLM), using visual cues to perform arbitrary segmentation tasks received wide attention.
The language model, developed by the Beijing Academy of Artificial Intelligence (BAAI), exemplifies China's commitment to seizing the opportunities brought about by LLM development.
Zhao Zhiyun, head of the Institute of Scientific and Technical Information of China (ISTIC), said China's LLM technology has developed rapidly along various technical routes in recent years, adding that some pre-trained ones with industry influence have emerged in various fields including natural language processing and machine vision.
China's general-purpose LLMs application is expanding from office and entertainment to fields such as healthcare, the industrial sector and education, according to a report released by the ISTIC.
The report also stated that 14 provincial-level regions in China are currently working on the development of LLMs among which 38 are in Beijing and 20 in south China's Guangdong Province.
With ChatGPT leading a new wave of global artificial intelligence (AI) LLM technology development, China is pooling AI innovation resources to carry out research in LLM algorithms and key technologies.
According to a plan recently released by Beijing to promote the innovation and development of AI, the core industry scale of AI in the city is expected to reach 300 billion yuan (about 41.88 billion U.S. dollars) by 2025.
Shanghai released a slew of measures to support the participation of private companies in the construction of AI infrastructure such as data and computing power.
Shanghai's Xuhui District is actively cultivating and introducing a number of research teams to build a highland for LLMs ecology cluster and innovative applications.
Shenzhen in south China's Guangdong Province has proposed to integrate funding resources in AI with a scale of 100 billion yuan to shore up intelligent computing power cluster supply, innovation ability of key core technologies and products, and industrial agglomeration level.
Experts also emphasized the importance of maintaining a calm and composed approach toward the country's AI development.
"China has become the world's largest AI application market, but it still lags behind in basic technology and talents," said Dai Qionghai, dean of the School of Information Science and Technology of Tsinghua University.
Dai called for strengthening the cultivation of AI talent and basic research, and beefing up the original innovation.
Universities, scientific research institutions, and companies are all involved in the development of LLMs, but joint research and development between academia and industry is still insufficient, according to the ISTIC report.
Zhao said efforts should be made to promote scenario-driven development and build specialized models in finance, medical care, electric power and other fields to achieve high-quality application breakthroughs.
The application scenarios and data can help upgrade LLM technology, Zhao added.