The race between the US and China in artificial intelligence has entered its next stage: commercialization.
Over the past decade, China’s AI development has surpassed that of the US by many measures, including in fundamental research. But the arrival of ChatGPT — the essay-writing, exam-passing chatbot developed by American startup OpenAI — has left Chinese players scrambling to catch up.
“Artificial intelligence technology in China is more advanced than in the US, and I think ChatGPT could start evening that out, with Microsoft leading the way,” said Dan Ives, managing director at Wedbush Securities. “But ultimately, that’s going to put more pressure on the Chinese players to advance their AI algorithms.”
Microsoft announced last week that it will team up with OpenAI to improve its search engine Bing in a bid to make up ground it lost to Google years ago.
The software giant said in late January that it is investing “multiple billions” in OpenAI, with media reports putting the total figure at $10 billion for a nearly 50% stake in the San Francisco-based startup.
“The fact that [OpenAI has] cracked the nut on the AI formula, and now Microsoft has bought 50% of it, that’s accelerated this AI arms race, both in the US as well as globally and within China,” Ives said.
Google, which leads in the search engine market in the US and many parts of the world, was quick to respond, unveiling its Bard AI chatbot on Feb. 6.
Chinese players were not far behind in announcing their own initiatives. The same week, Alibaba Group Holding, Tencent Holdings, Baidu, NetEase and JD.com all unveiled plans to test and launch their own ChatGPT-like services in the near future.
On the domestic front, the sudden rush of Chinese companies into the ChatGPT space is unlikely to bring any major disruption, according to Alicia Yap, Hong Kong-based analyst at Citi.
Unlike in the US, where Bing could potentially use ChatGPT to challenge Google, Baidu has no such rival to worry about in China, she said. The company, which plans to release its own “Ernie Bot” in March, is not only the local leader in search but also relatively advanced in terms of AI capability.
With big tech names flocking to the new AI race, investors around the world are also taking notice. But their views on the potential of ChatGPT-style innovations are mixed.
Gloria Zhang, an investment manager at venture fund DCM, said such tech is in its early stages and people are still experimenting with different possible uses, but she adds it has the potential to drastically increase productivity and transform industries.
“I think it will be a huge paradigm shift to a lot of existing industries, such as e-commerce and social media,” she said.
Others are less sure. The popularity of ChatGPT is “definitely going to develop some derivative businesses, but it’s more just excitement right now,” said Jeffrey Lee, a partner at NLVC, an early-stage venture capital fund based in Silicon Valley. “For the first-time consumers … this is affirming what they’ve imagined and what Hollywood imagined AI is going to be like.”
But behind that excitement are real-world problems with the technology that still need to be ironed out, he said.
One is the “self-reinforcing” problem. A large language model like ChatGPT is like an echo chamber, Lee explained, that only reinforces the biases in the data set it is trained with. This could lead to societal and regulatory problems if, for example, a chat bot spreads misinformation in the process of answering questions, as it does not have a proper fact-checking mechanism.
“The second issue is the business applicability of GPT. What is the business purpose of this?” Lee said.
While an AI chatbot might help provide better customer service, for example, the cost of maintaining the tech might not necessarily help keep overall costs down and profit up. Moreover, the ad-based revenue model of Google and most other search engines could be significantly disrupted.
“How is this going to fundamentally improve their business model and their business at the same time? How is it not going to hurt their business?” Lee asked.
Ke Yan, a research analyst at Singapore-based DZT Research, pointed out other hurdles to developing and monetizing apps like ChatGPT, some unique to Chinese-speaking markets. “It requires a large data set for training, which only bigger companies can afford. Then it has the issue of web scraping to supply content for training, which might violate data privacy and copyright laws,” he said.
“For Chinese companies, I think there is another layer of challenge: The Chinese language is less structured than the English language, hence it is harder to train ChatGPT to produce things that look like human output,” he added.
Language is not the only hurdle Chinese companies are facing. Washington’s semiconductor ban and other trade sanctions threaten to put Chinese tech companies at a disadvantage in this latest leg of the AI race.
The most recent ban, announced on Oct 7, bars American companies from shipping certain grades of advanced chipmaking equipment to any China-based client without a license. It also restricts suppliers from supporting the development of China’s cutting-edge artificial intelligence and supercomputer chips.
“I think the poker game is around the chips,” said Ives at Wedbush. “That is a big wild card that could ultimately slow down AI in China.”
Fearing further US sanctions, big Chinese tech players stocked up enough advanced chips over the past few years to prepare for the worst-case scenario, but further development in AI technology could be severely hindered if they cannot access next-generation chips, people familiar with the matter at both Baidu and Tencent told Nikkei Asia.
NLVC’s Lee said Chinese tech companies may have enough advanced chips on hand to start launching AI services similar to ChatGPT, but “probably not at scale.”
This article first appeared on Nikkei Asia. It has been republished here as part of 36Kr’s ongoing partnership with Nikkei.