The development of LLM models, which are trained using AI deep learning techniques to mine huge amounts of data and text, has entered an "accelerated" phase since 2020, according to a report by research institutes under China's Ministry of Science and Technology .
Illustration photo: Reuters
The report, released on Sunday, said that in 2020, Chinese institutions issued just two LLMs compared to 11 in the US, but in 2021 a total of 30 LLMs were issued in each country.
US institutions are expected to issue a total of 37 LLMs in 2022, compared with 28 in China, according to the report. China has issued 19 LLMs so far this year, while the US has added just 18 of these smart models.
“China and the US lead the distribution of large language models worldwide , with more than 80% of the models being published by the two countries,” the report concluded. “The US consistently leads the world in the number of large language models.”
The report comes as China's AI industry faces significant challenges, with US-imposed export controls making it difficult for Chinese organizations to access semiconductors used to train LLMs, as well as other advanced computing tasks.
The report analyzed 79 LLMs developed in China, noting that although 14 provinces and regions have developed the technology, joint development projects have yet to “meet the demand.”
After OpenAI released ChatGPT, Chinese tech giants from Alibaba to surveillance company Sensetime and search engine giant Baidu rolled out their own versions of chatbots powered by general AI technologies.
Hoang Ton (according to Reuters)
Source
Comment (0)