Written by Josh Yeh
HONG KONG (Reuters) – The Biden administration plans to put guardrails on U.S.-developed artificial intelligence (AI) models that power popular chatbots such as ChatGPT to protect the technology from countries such as China and Russia, Reuters reports. Reported.
But China has built a domestic generative AI industry over the past year, encouraging its companies to avoid foreign technology.
Here's how much China currently relies on U.S. AI models and the potential impact of Washington's plans.
How can OPENAI's AI models be used in China?
OpenAI's core AI services, such as ChatGPT and DALL-E image generator, have not been officially deployed in mainland China. An OpenAI spokesperson told Reuters last year that it was not possible in certain countries due to local “circumstances.”
However, many companies and engineers use proxy tools such as virtual private networks (VPNs) to mask their network addresses and access OpenAI's services.
As a result, many Chinese companies have been able to build software and applications based on OpenAI's model. Chinese companies also frequently benchmark their AI models against his OpenAI AI models.
OpenAI has blocked Chinese companies from accessing its services. Last December, OpenAI suspended the account of TikTok's Chinese owner ByteDance after technology website The Verge reported that ByteDance had developed its own AI using OpenAI's technology.
In Hong Kong, a special administrative region of China, access to OpenAI's AI models is also restricted, but not sealed. Although OpenAI's services are not available there, Microsoft, an OpenAI investor and partner, has made his Copilot, a generative AI service built with OpenAI's latest technology, publicly available. By partnering with Microsoft, Microsoft companies will also have access to his OpenAI AI models.
Do Chinese AI models use US technology?
The Commerce Department's move is aimed at exporting proprietary or closed-source AI models, where the software and data used for training are kept secret, officials told Reuters. Open source models go beyond export controls.
However, China relies heavily on a number of open source models developed in the West, such as Metaplatforms' Llama series.
In March, the Beijing Institute of Artificial Intelligence, a high-level research institution, revealed that the majority of China's domestically produced AI models are actually built using Meta's Llama model, making this a key challenge for China's artificial intelligence. He was quoted by Chinese state media as saying that AI development.
The institute told then-Chinese Premier Li Qiang that China had a “serious lack of autonomy” in the region.
01.AI, one of China's hottest AI unicorns founded by former Google executive Li Kaifu in November 2023, announced that its AI model Yi-34B is built on Meta's Llama. This was discovered by some AI engineers and faced a huge backlash. system.
That said, many Chinese technology companies such as Baidu, Huawei, and iFlytek are working on developing their own “completely proprietary” AI models. Some of them claim that their models are now as powerful as OpenAI's latest GPT4 models in some areas.
What is Beijing's position on US AI models?
Chinese authorities have emphasized the need for the country to develop its own “controllable” AI technology, following orders from Chinese President Xi Jinping to become technologically self-sufficient.
In February last year, state newspaper China Daily said in a post on Chinese microblogging site Weibo that ChatGPT had been “used in the US government's efforts to spread disinformation and manipulate the global narrative for its own geopolitical interests. “There is a possibility that we can lend a helping hand.”
The country is also active in rolling out regulations around the use of generative AI, requiring government approval before services can be made available to the public. As of January, China had approved more than 40 AI models for public use, none of which were foreign AI models.
Last April, a senior Hong Kong government official also said there were no plans to allow the use of ChatGPT within the Hong Kong government.
The Chinese government's positive sentiment toward U.S. generative AI technology is primarily aimed at comparing how far China lags behind the U.S. in AI development, rather than encouraging U.S. AI technology.
At the country's annual parliament last March, the minister used a soccer analogy to explain ChatGPT's significant lead over Chinese AI products.
Chinese Minister of Science and Technology Wang Zhigang said of Argentine superstar Lionel Messi, “Soccer requires dribbling and shooting, but it is not easy to be as good as Messi.''
(Reporting by Josh Yeh; Editing by Sharon Singleton)