Beijing Becomes the City with the Largest Number of AI Large Models in China
-
Yesterday, at the 2023 Artificial Intelligence Computing Conference (AICC), the 'Beijing Artificial Intelligence Industry Large Model Innovation Application White Paper (2023)' was released. The white paper shows that Beijing is currently the region with the best innovation foundation, the most concentrated talent resources, the strongest R&D innovation capabilities, and the most active product iteration in the field of artificial intelligence in China. It now has 122 large model innovation teams, accounting for about half of the national total, and ranks first in the country in terms of the number of large models.
Globally, countries around the world are actively promoting the development and application of large models. The white paper shows that the total number of general-purpose large models released by the United States and China currently accounts for 80% of the global total, making them leaders in the field of large model technology. Nationally, Beijing was the earliest to lay out large models and is currently the region with the best innovation foundation, the most concentrated research teams, the strongest technical capabilities, and the most active model iteration in China. As of October 2023, there are 254 manufacturers, universities, and research institutions in China with large models of more than 1 billion parameters, distributed in more than 20 provinces/regions, of which 122 are in Beijing.
Beijing has a broad market space for large model applications and has taken the lead in carrying out diversified application practices. For example, Xianyuan Technology, focusing on the needs of Beijing Yiqing Technology Group in market demand tracking and new product development, has created a consumer-oriented product business large model demonstration application, forming services such as intelligent reverse customization of products, predictive production, intelligent scheduling, and intelligent marketing, accelerating the 'digital and intelligent' upgrade in the consumer manufacturing field. 360 Company, focusing on the needs of banks in digital transformation, has created a large model for the financial field, reducing the average office document processing time for employees by 40%.
The white paper also further highlights the challenges currently faced in the application and implementation of large models, such as insufficient computing power resources, high-quality data becoming a major bottleneck in the industry application of large models, the gap between large model hallucination problems and large-scale implementation, and the phenomenon of 'bandwagoning' and homogenization in large model applications.
During the conference, a series of innovative achievements in the field of artificial intelligence in Beijing, including the first batch of Chinese internet corpora, were released. To meet the urgent need for high-quality datasets in the rapid development of large models, the Institute for Artificial Intelligence Research and Innovation (BAAI) jointly announced the establishment of the 'Chinese Internet Corpus' with TRS and CASW, providing safe and reliable corpus resources for the development of the domestic large model industry. The first phase of 104GB of corpus data was released at the conference. At the same time, BAAI, together with 17 large model companies including Alibaba Cloud, ThunderSoft, and Ant Group, released the 'Chinese Internet Corpus Co-construction and Sharing Initiative,' advocating six principles including cooperation and sharing, security and compliance, and high-quality data to promote the healthy and sustainable development of the artificial intelligence industry.
The rapid iteration of large models cannot be separated from sufficient computing power support. During the conference, the '2023-2024 China Artificial Intelligence Computing Power Development Evaluation Report' was released, announcing the ranking of Chinese cities in terms of artificial intelligence computing power. Beijing ranked first for the sixth consecutive year.
To further strengthen the city's AI computing infrastructure foundation, the Beijing Artificial Intelligence Public Computing Platform was officially launched at the conference. Leading AI model companies including Zhipu AI and Zidong Taichu have become the first batch of signed tenants. The platform will soon achieve high-quality utilization of public computing resources to accelerate the output of major scientific and technological achievements.