Microsoft-backed OpenAI has kept its hit ChatGPT app off-limits to users in China, but the app is attracting huge interest in the country, with firms rushing to integrate the technology into their products and launch rival solutions.
While residents in the country are unable to create OpenAI accounts to access the artificial intelligence-powered (AI) chatbot, virtual private networks and foreign phone numbers are helping some bypass those restrictions.
At the same time, the OpenAI models behind the ChatGPT programme, which can write essays, recipes and complex computer code, are relatively accessible in China and increasingly being incorporated into Chinese consumer technology applications from social networks to online shopping.
The tool’s surging popularity is rapidly raising awareness in China about how advanced US AI is and, according to analysts, just how far behind tech firms in the world’s second-largest economy are as they scramble to catch up.
“There is huge excitement around ChatGPT. Unlike the metaverse which faces huge difficulty in finding real-life application, ChatGPT has suddenly helped us achieve human-computer interaction,” said Ding Daoshi, director of Beijing-based internet consultancy Sootoo. “The changes it will bring about are more immediate, more direct and way quicker.”
OpenAI or ChatGPT itself is not blocked by Chinese authorities but OpenAI does not allow users in mainland China, Hong Kong, Iran, Russia and parts of Africa to sign up.
OpenAI told Reuters it is working to make its services more widely available.
“While we would like to make our technology available everywhere, conditions in certain countries make it difficult or impossible for us to do so in a way that is consistent with our mission,” the San Francisco-based firm said in an emailed statement. “We are currently working to increase the number of locations where we can provide safe and beneficial access to our tools.”
In December, Tencent Holdings’ WeChat, China’s biggest messaging app, shut several ChatGPT-related programmes that had appeared on the network, according to local media reports, but they have continued to spring up.
Dozens of bots rigged to ChatGPT technology have emerged on WeChat, with hobbyists using it to make programmes or automated accounts that can interact with users. At least one account charges users a fee of CNY 9.99 ($1.47 or roughly Rs. 120) to ask 20 questions.
Tencent did not respond to Reuters’ request for comments.
ChatGPT supports Chinese language interaction and is highly capable of conversing in Chinese, which has helped drive its unofficial adoption in the country.
Chinese firms also use proxy tools or existing partnerships with Microsoft, which is investing billions of dollars in its OpenAI, to access tools that allow them to embed AI technology into their products.
Shenzhen-based Proximai in December introduced a virtual character into its 3D game-like social app who used ChatGPT’s underlying tech to converse. Beijing-based entertainment software company Kunlun Tech plans to incorporate ChatGPT in its web browser Opera.
SleekFlow, a Tiger Global-backed startup in Hong Kong, said it was integrating the AI into its customer relations messaging tools.
“We have clients all over the world,” Henson Tsai, SleekFlow’s founder said. “Among other things, ChatGPT does excellent translations, sometimes better than other solutions available on the market.”
Censorship
Reuters’ tests of ChatGPT indicate that the chatbot is not averse to questions that would be sensitive in mainland China. Asked for its thoughts on Chinese President Xi Jinping, for instance, it responded it does not have personal opinions and presented a range of views.
But some of its proxy bots on WeChat have blacklisted such terms, according to other Reuters checks, complying with China’s heavy censorship of its cyberspace. When asked the same question about Xi on one ChatGPT proxy bot, it responded by saying that the conversation violated rules.
To comply with Chinese rules, Proximai’s founder Will Duan said his platform would filter information presented to users during their interaction with ChatGPT.
Chinese regulators, which last year introduced rules to strengthen governance of “deepfake” technology, have not commented on ChatGPT, however, state media this week warned about stock market risks amid a frenzy over local ChatGPT-concept stocks.
The Cyberspace Administration of China, the internet regulator, did not respond to Reuters’ request for comment.
“With the regulations released last year, the Chinese government is saying: we already see this technology coming and we want to be ahead of the curve,” said Rogier Creemers, an assistant professor at Leiden University.
“I fully expect the great majority of the AI-generated content to be non-political.”
Chinese rivals
Joining the buzz have been some of the country’s largest tech giants such as Baidu and Alibaba who gave updates this week on AI models they have been working on, prompting their shares to zoom.
Baidu said this week it would complete internal testing of its “Ernie Bot” in March, a big AI model the search firm has been working on since 2019.
On Wednesday, Alibaba said that its research institute Damo Academy was also testing a ChatGPT-style tool.
Duan, whose company has been using a Baidu AI chatbot named Plato for natural language processing, said ChatGPT was at least a generation more powerful than China’s current NLP solutions, though it was weaker in some areas, such as understanding conversation context.
Baidu did not reply to Reuters’ request for comments.
Access to OpenAI’s GPT-3, or Generative Pre-trained Transformer, was first launched in 2020, an update of which is the backbone of ChatGPT.
Duan said potential long-term compliance risks mean Chinese companies would most likely replace ChatGPT with a local alternative, if they could match the U.S.-developed product’s functionality.
“So we actually hope that there can be alternative solutions in China which we can directly use… it may handle Chinese even better, and it can also better comply with regulations,” he said.
© Thomson Reuters 2023
