China approves first batch of AI language models

 

SOURCE / ECONOMY

Authorities strengthen regulation to prevent risks, ensure digital safety

By GT staff reporters Published: Aug 31, 2023 

A total of 11 Chinese artificial intelligence (AI) large models, including Baidu's ERNIE Bot, reportedly have obtained approval for release to the general public starting from Thursday as Chinese technology companies ramp up innovations to compete with ChatGPT. 

Analysts and industry insiders said that China's AI sector has developed at an unprecedented speed so far this year despite the intensifying US blockade, and they expressing confidence in China's ability to break the US-led Western stranglehold in the high-tech sector.

On Thursday, Baidu took the lead by announcing that its AI-generated language model ERNIE Bot is now fully open to the general public.

Moreover, Baidu will launch a suite of new AI-native apps that allows its users to fully experience the four core abilities of generative AI - understanding, generation, reasoning and memory, the company said in a statement sent to the Global Times.

Boosted by the news, Baidu's shares in Hong Kong closed up by 2.12 percent at HK$139.4 ($17.77), while semiconductor-related shares such as East China's Jiangsu-based China Wafer Level CSP Co rose by the daily limit of 10 percent to 21.34 yuan ($2.93).

For its part, China's leading AI company iFlytek plans to upgrade its general large model, comprehensively benchmarking ChatGPT, on October 24. The new model is expected to surpass ChatGPT in Chinese capability and be equivalent to it in English capability. It is expected to be benchmarked against GPT-4 in the first half of 2024, according to a press release sent to the Global Times.

"Overall, the development of China's AI sector has witnessed progress at an unprecedented speed so far this year," Liu Dingding, an independent tech analyst, told the Global Times on Thursday.

Thanks to policy support, China entered a rapid development period for large AI models starting in 2020, and the total number of published large AI models in China and the US now accounts for more than 80 percent of the global total, Liu said.

He said that Chinese AI language models are running neck-and-neck with those developed in the US, with China even leading the US in applications driven by the country's 5G, mobile internet and other digital infrastructure.

China ranked second in the Global AI Innovation Index for a third straight year in 2022, as it has been accelerating efforts to create a talent pool, build data infrastructure and make technological breakthroughs.

Liu Qingfeng, chairman of iFlytek, said at the China Entrepreneurs Forum 2023 on August 25 that Huawei's graphics processing units can match Nvidia's A100 performance.

"We are collaborating with Huawei to build a cluster advantage in the training of ultra-large language models, and we have confidence in breaking the stranglehold in computing power and striving to benchmark with OpenAI in terms of the application of general AI," Liu said.

"The achievements in China's AI sector, especially AI chips, show that it is wishful thinking for the US to contain China's technological rise by restricting high-tech chip exports," Xiang Ligang, director-general of the Beijing-based Information Consumption Alliance, told the Global Times on Thursday.

Chinese tech giant Huawei surprised the market by unveiling its Mate60 Pro smartphone on Tuesday, which the company said is the most powerful of the series, with new breakthroughs in aspects including satellite communications, AI and Kunlun Glass, a screen that is high performing on the inside and toughen on the outside.

Huawei's rollout of a cutting-edge product despite severe US restrictions shows that the US crackdown is doomed and instead it will prompt Chinese firms to develop faster, Xiang said, noting that China has become stronger despite escalating US efforts at containment.

The era in which the US could contain Chinese companies' development with political measures has gone, and the US needs to ponder the China-US relationship. Does the US want to pursue globalization and cooperation, and what are the world's two largest economies competing over, Xiang asked, urging the US to refrain from hegemonic actions but instead strengthen sci-tech innovations to maintain global competitiveness.

While the boom of generative AI technology has created new opportunities for economic and social development, there are potential risks such as the spread of fake information, the infringement of personal information and data safety, Liu said, noting that the authorities are stepping up regulation of AI language models for the healthy development of the market.

An interim regulation on the management of generative AI services, released by the Cyberspace Administration of China and several other authorities, took effect on August 15. 

The move is aimed at promoting the sound development of generative AI and its standard applications, safeguarding national security and the public interest, and protecting the legitimate rights and interests of citizens, legal entities and organizations, according to the CAC.