China’s biggest tech companies are barreling into generative A.I. in an effort to catch up to OpenAI’s ChatGPT, even as the country’s state media warns investors of a “valuation bubble.”
The Alibaba Group announced on Tuesday that it would be installing its “Tongyi Qianwen” chatbot into its smart speakers and its office chat software, with plans to integrate the new technology into all its services in the “near future.” The Chinese company is also offering access to the new program to enterprise clients for “beta testing.”
A.I. and cloud computing could drive a “watershed moment” for tech, said Daniel Zhang, the Alibaba Group’s CEO. Zhang took personal control of the company’s cloud computing division after a service outage last December, and is remaining as its CEO after last month’s organizational shakeup.
Alibaba’s announcement follows a similar reveal from Chinese A.I. developer SenseTime, which unveiled its own chatbot, called SenseNova, on Monday. CEO Xu Li showed the bot writing an email, telling a story and writing code after user prompts. “We can provide a supermarket of A.I. big models,” he told reporters. The company also demonstrated tools that could generate images and use human movements to animate a digital avatar.
Xu predicted that A.I. programs would soon take on the lion’s share of future A.I. development, doing 80% of the work, while human programmers offer direction and clean up the generated results.
SenseTime is one of China’s so-called “A.I. dragons”, alongside CloudWalk, Yitu and Megvii, which are generally seen to be the country’s champions for the new technology. The U.S. restricted sales of some advanced technologies to SenseTime in 2019, and later barred U.S. investors from investing in the Chinese company.
Several other Chinese firms have announced ChatGPT clones in recent weeks. Baidu, which views A.I. as a way to recapture ground lost to competitors like Alibaba and Tencent, showed off its chatbot, called Ernie Bot, in a recorded demonstration in mid-March.
Game developer NetEase also plans to integrate generative A.I. into one of its video games in June, suggesting the technology can help the game’s characters react to the player’s actions. “If you tell an NPC [non-playable character]: ‘your home is on fire’, the NPC will rush to his home quickly so that you don’t have to fight with the NPC as a task in the game,” the company posted on WeChat in February.
Excessive hype
However, as China’s tech giants flog their generative A.I. plans, Chinese state media is warning the country’s tech sector against getting too excited about artificial intelligence.
On Monday, the Economic Daily, a state-run newspaper, published a commentary piece noting that Chinese tech stocks had surged due to predictions about A.I., despite the lack of any commercially-available products. The piece said the sector had “signs of a valuation bubble,” and warned against “excessive hype.”
China’s regulators may be starting to figure out how to regulate the new industry. On Tuesday, Chinese regulators revealed draft rules on “A.I.-generated content,” which would require a security evaluation from companies wanting to provide such services.
Cyberspace Administration of China published draft rules over AIGC (ChatGPT-like services) for public comment, which, as it stands now, basically requires a license (申报安全评估 asks for a security evaulation from CAC) to provide AIGC to the public. https://t.co/lZhUzbnnJu pic.twitter.com/QoX6mm4LHd
— Zichen Wang (@ZichenWanghere) April 11, 2023
Catching up
In 2021, Chinese institutions and researchers published the most papers in A.I. journals, according to a recent report from the Stanford Institute for Human-Centered Artificial Intelligence. China also led in terms of citations by other publications. That perceived strength in A.I. motivated some of the Biden administration’s semiconductor controls last year, barring the export of some advanced chips from companies like Nvidia and AMD to Chinese companies.
Yet the success of ChatGPT appears to have caught Chinese tech companies off-guard. Experts also suggest that generative A.I. might also run afoul of Chinese internet controls on content. Last year, the MIT Technology Review reported that ERNIE-ViLG, an image generating program developed by Baidu, would not generate images of Tiananmen Square.
OpenAI has not made ChatGPT available in either mainland China or the semi-autonomous city of Hong Kong. China has since blocked ChatGPT’s website in the country, and reportedly told both Tencent and Alibaba not to offer access to OpenAI’s service.