AI-driven large models showcase great application potential

(Xinhua) 14:20, January 23, 2024

BEIJING, Jan. 23 (Xinhua) -- Large models based on algorithms, computing power and big data have become the mainstream of AI development, and have continuously empowered China's economic development.

As of October 2023, 238 big models have been launched in China, according to official data. Meanwhile, an increasing number of universities, institutes and innovative enterprises are jumping on the bandwagon of AI large models.

Large language models have garnered significant attention in recent years, with a proliferation of products attracting a substantial user base based on their functionalities such as language understanding, logical reasoning, knowledge question and answer, and text generation.

China's leading AI company iFLYTEK unveiled its AI-powered large language model dubbed "Spark Desk" last year. The number of its users had exceeded one million just within 14 hours after it was launched.

The multitask model covers key features including AI-generated content, multi-language understanding, logical reasoning, mathematics and coding.

China's tech giant Baidu rolled out its large language model and ChatGPT-like product dubbed Ernie Bot in March, 2023. It can provide a wide range of services, such as welcome speeches, speech drafts, business plans, instructions, flow charts and mind maps, covering many aspects of work and life.

As of December 28, 2023, the user base of Ernie Bot had exceeded 100 million, making it China's first large model product to provide paid services for consumers.

The research and development (R&D) of large models represent a key focus of future scientific endeavors, albeit with high entry barriers and related challenges. Only by fully integrating into various industries, can the large models serve the real economy more effectively and achieve sustainable development.

In recent years, China's self-developed AI-driven large models have showcased great application potential in the real economy, such as smart mining, drug R&D, meteorology, government affairs, finance, intelligent manufacturing and railway management.

In the manufacturing sector, robots equipped with large models help workers improve their efficiency. In the transportation field, traffic management departments can use large models to optimize traffic flow and road conditions.

In the realm of drug R&D, large models can expedite the process of new drug research and development, and achieve highly efficient, innovative, and personalized drug design and discovery by using natural language processing, knowledge graphs, and molecular modeling.

As a deep learning model with hundreds of billions of scale parameters and complex structures, a large model has extremely high requirements for the carrying capacity and efficiency of computing power.

Many cities with strong demands for AI are increasing the supply of computing power, such as Beijing, Shenzhen and Shanghai, also dubbed as the "first echelon" of China's AI development. These cities have successively issued relevant policies to support the layout of computing power infrastructure.

According to a white paper on China's computing power development index, the intelligent computing power developed rapidly to account for 59 percent of the nation's computing power in 2023.

(Web editor: Zhang Kaiwei, Liang Jun)


Related Stories