China is playing catch-up in a ChatGPT world, Chinese lawmaker says

[Hamartia Antidote]

https://www.scmp.com/news/china/science/article/3213197/china-playing-catch-chatgpt-world-chinese-lawmaker-says

• The country needs to advance its own cognitive development models, which are far behind the best competitors, iFlytek founder says
• Getting there means firming up its software and hardware foundations, he says

Large language products will “likely be the biggest leap in artificial intelligence technology”, according to one NPC deputy. Photo: Shutterstock

China should speed up the development of its own ChatGPT-like language models to benefit industry and the public alike, a top Chinese legislator has urged.
Hailing large language products as “likely the biggest leap in artificial intelligence technology”, Liu Qingfeng, founder and chairman of voice recognition firm iFlytek, said China’s cognitive intelligence models still lagged considerably behind the best chatbots in the world.

Given the enormous industrial opportunities and vast social impacts of such models, as well as the ever-growing hi-tech race with the United States, China must have a strategy to catch up to and even overtake the US in some areas, Liu told China Electronics News at the National People’s Congress in Beijing on Saturday.
He said that in the future, ChatGPT would be not only a chatbot but an artificial intelligence assistant revolutionising life from the way information was distributed to human-computer interaction, “allowing everyone to become more creative on the shoulders of AI”.

Liu said that while research institutions and businesses in China had released a number of large language models, they still needed to work out systematic issues such as general-purpose pre-training and reinforcement learning from human feedback.

To quickly close the gap with the US, state-level special projects should be set up to support a long-term, steady innovation system to bring together industry and academia – including newly established national laboratories – to tackle the development and upgrade of cognitive intelligence models, he said

Meanwhile, the country could quickly do a series of pilot projects and apply such models to the education, medical, and legal systems, he said.
To develop and run China’s own large language models, the nation must strengthen its software and hardware foundations, such as information storage, computing power and even operating systems, and become more self-reliant in these technologies.

He said that just as ChatGPT’s developers trained their system on massive internet resources and electronic books from American public libraries, China should establish a national data resource platform to gather and share basic, high-quality data needed for building large models.

Liu said China should also research the possible social disruptions AI might induce, including job losses, and how the government could guide such transitions.

ChatGPT has been a hot topic during this year’s “two sessions”, the country’s biggest annual political gathering.

Last Sunday, Science and Technology Minister Wang Zhigang said such products were “difficult to achieve”.

Wang also raised ethical concerns, saying that China must “wait and see” when it came to catching up with ChatGPT, as such large language models often resulted in unpredictable or inaccurate responses.