Companies, universities and research institutes in China have launched some 130 LLMs between 2019 and July of this year, according to a report by CCID Consulting Co. Ltd. Sixty-four new models were launched in China between January and July this year, accounting for 67% of all LLMs launched globally in the same period.
 |
Amid the flurry of activity, two areas have emerged as bright spots for adoption of LLMs by Chinese tech giants: incorporation into cloud services and industry applications.
In June, Alibaba Group Holding Ltd.’s cloud computing arm announced that its intelligent assistant has been integrated with its LLM Tongyi Qianwen, while Tencent Holdings Ltd.’s conference app has released an AI meeting assistant powered by its LLM Hunyuan for trial use.
LLMs that target specific industry needs include Huawei Technologies Co. Ltd.’s Pangu, released in July, which focuses on areas such as mining. JD.com that same month launched ChatRhino for retail and logistics businesses.
In the area of governance, Chinese regulators have been quick to catch up. In July, seven government agencies including the Cyberspace Administration of China and the National Development and Reform Commission issued final guidelines for regulating generative AI.
The rules were China’s first national legislation covering AI and cover the research, development and use of products with generative AI functions.
With providers only allowed to launch their generative AI products on the market after regulators approve both their security review and algorithm registration, the rules are a way for the government to promote the development and use of AI while trying to reduce risks, including the spread of fake news, infringement of personal information and data security.
Read more
China’s AI Charge Part 1: The Battle Begins to Build a ChatGPT Challenger
 |
| Companies showcase a variety of artificial intelligence products on Oct. 12 at the 4th Shenzhen International Artificial Intelligence Exhibition in Shenzhen, South China’s Guangdong province. Photo: VCG |
Challenges, curbed investor enthusiasm
LLMs in China are still in their infancy and competition comes down to technological advancement, while first-mover advantage doesn’t necessarily result in victory, one executive at a Software as a Service (SaaS) company that partners with both Alibaba and Baidu previously told Caixin.
However, anticipated challenges mean not everyone is diving in headfirst. Some industry insiders anticipated that regulators would tighten their grip over the issuance of licenses granted to new models, amid potential concerns of an overcrowded market and uncertainties about the models’ application and profitability.
The sheer number of LLM application licenses has led some provinces to set quotas on approvals, Caixin has learned. There won’t be too many new LLMs approved by regulators in the future, one industry source said.
Investor enthusiasm has also faded somewhat in the later half of the year due to the current low return on investment for general-purpose LLMs and cautious investors amid the sluggish performance of many AI companies on the stock market.
For example, investment firm Temasek International Pte. Ltd. hasn’t made a single investment on the mainland in artificial-intelligence generated content (AIGC) stocks, a term commonly used for shares related to generative AI, due to their lack of fundamentals, its China President Wu Yibing told Caixin in a July interview.
The U.S. also in October amped up export restrictions on more cutting-edge AI semiconductor and chip manufacturing equipment aimed at further blocking China’s access to American technology.
The move came after Huawei’s launch of its advanced 5G Mate 60 smartphone during U.S. Commerce Secretary Gina Raimondo’s visit to China in late August, which sparked celebration in the country and showed that China’s pursuit of advanced technology hasn’t been stopped by U.S. sanctions.
“It will take increasingly more time for LLM development in China as larger data sets and more model parameters are required for model training, while computing power growth is handicapped,” wrote analysts at JP Morgan in a November report.
China’s LLM market could see a consolidation as computing power becomes scarce and potentially cost prohibitive, with bigger domestic LLM operators that already possess large amount of AI chips being better positioned while smaller LLM developers facing cost issues in 
training their models, they said.
Source: Caixin Global
