If you are reading this on the internet, then you’ve probably heard of ChatGPT. The artificial intelligence (AI) chatbot, made by California-based company OpenAI, has taken the world by storm. More than 100 million people around the world are estimated to have used it in the two months since it launched, and Microsoft has made a “multiyear, multibillion dollar investment” in OpenAI to integrate the company’s technology into its search engine, Bing (with some initially frightening effects). Google is developing its own chatbot named Bard.
The furor around ChatGPT and similar alternatives has prompted a scramble in China’s tech sector to join the party. Baidu, China’s leading search engine, said it plans to roll out its “Ernie Bot” in March while other Chinese tech giants, such as Alibaba and JD.com, announced chatbots of their own.
China has spent years developing its artificial intelligence capabilities, outlining its ambitions in a 2017 plan that sought to make the country a “global innovation center” in the field by 2030. Those goals have engendered competition but also collaboration between U.S. and Chinese scientists and companies, including through labs that the likes of Baidu, Tencent, and Alibaba have set up in Silicon Valley. But as the relationship between the world’s two largest economies becomes increasingly adversarial—particularly on the technology front—the room to work together on technologies such as artificial intelligence may rapidly be shrinking. The Biden administration’s restrictions last October on the export of advanced semiconductor chips to China, aimed in part at China’s AI capabilities, was a major broadside in a relationship that has only become more acrimonious.
The 2030 goal was set months after what could be described as Beijing’s AI “Sputnik moment,” when a Google AI program beat a Chinese grandmaster at the ancient board game wei qi, also known as Go. “That really set off China’s AI revolution,” said Sam Howell, who researches technology and national security at the Center for a New American Security.
But defining the chatbot moment in overly nationalistic terms can be misleading, said Graham Webster, a researcher and editor in chief of the DigiChina Project at Stanford University’s Cyber Policy Center. ChatGPT is a far cry from the kind of tech that has traditionally been developed in U.S. government labs. “I don’t think that one should view ChatGPT as a win for the United States in some kind of race,” Webster said. “OpenAI is a company, but it is not the United States. … It has its own ambition and values that may sometimes align with most Americans and may sometimes not, and it’s simply not directed by national interest.”
But national wrinkles still play a big part. OpenAI’s Chinese counterparts will have a different calculus because of Beijing’s often restrictive tech regulations, its vast censorship apparatus, and its broader government control of the private sector. In China, generative AI will have to hew to Beijing’s rules on how the product affects national security, Webster said.
“So there’s an automatic responsibility, essentially, to conduct what you could consider content moderation or censorship—it’s frankly both,” Webster added.
There’s also a major difference in how China approaches regulation of artificial intelligence, according to Paul Triolo, senior vice president for China and technology policy lead at the Albright Stonebridge Group, adding that the United States tends to approach AI as an added element of already regulated sectors. “Chinese officials on the other hand view AI as a critical technology that requires a regulatory framework, both to control the negative aspects of AI algorithm deployment but also to provide technology companies with a clear sense of where the regulatory lines will be drawn to encourage innovation in the sector,” Triolo said.
ChatGPT has not been made available in China, and Beijing has reportedly restricted Chinese tech firms from using or offering its services to the public. The Chinese government is attempting to justify ChatGPT restrictions and broader censorship in China by depicting it as part of the broader U.S.-China rivalry, said Angeli Datt, who researches Chinese censorship at free expression group PEN America.
“When really, it’s the sheer fact that uncensored and free information outside of the firewall counters Chinese Communist Party propaganda, misinformation, and disinformation” about any number of sensitive issues, like human rights in Xinjiang, Datt said.
An official at the Chinese Embassy in Washington said Beijing is focused on keeping AI useful and safe.
“China is committed to building a community with a shared future for humanity in the domain of AI and efforts to advocate a people-centered approach and the principle of AI for good, ensure that AI is safe, reliable, controllable, and capable of better empowering global sustainable development and enhancing the common well-being of all humanity,” the official wrote via email.
Ironically, Baidu’s chatbot Ernie was trained on English-language information from Wikipedia and Reddit, which are both blocked in China. “It’s going to be a difficult balancing act for Chinese companies to find ways to use generative AI technology without running afoul of Chinese censors,” said Paul Scharre, who recently wrote Four Battlegrounds: Power in the Age of Artificial Intelligence.
In any event, companies operating in China’s tech space are used to navigating top-down restrictions. “All of these companies are very experienced at navigating government regulation and censorship requirements on their other products,” Webster said. “So it’s certainly an extra barrier, but it’s one that’s familiar to them.”
Beijing will also likely use AI to produce propaganda, Scharre said. “They’re basically fake news generators, so they can be used to generate state propaganda or fake news at speed and scale that may not be possible with people.”
More impactful than Beijing’s censorship regarding China’s ability to develop its AI ecosystem could be U.S. export controls last year that all but cut off the sale of the most advanced semiconductor chips—and the equipment needed to make them—to Chinese companies. Those chips are crucial to the development of advanced AI models, and the export restrictions could be the “biggest single impact of U.S.-China technology competition on China’s AI development,” Triolo said.
“The restrictions will take time to bite, as Chinese organizations and firms have stockpiled some chips. But over time, if Chinese companies cannot find workarounds, they will be at an increasing disadvantage in terms of training AI algorithms on cutting-edge hardware,” Triolo added. “This is particularly the case with large language models, which are best trained on the most advanced hardware.”
Not unlike the fight over global leadership of other advanced technologies, such as 5G mobile networks, there is more at stake than leadership in technologies that can write clever texts on demand.
“Competition over AI—and emerging tech broadly—carries consequences that extend beyond the digital domain. A country that leads in AI is going to reap significant economic benefits and also gain a national security advantage,” Howell said. “What’s at stake here is not just who leads in artificial intelligence but also who sets the rules for how it’s going to be used around the world—so democratic values and the concept of open societies are at stake in the AI race.”
What does ChatGPT itself think about China’s AI capabilities?
“China has made significant investments in artificial intelligence (AI) in recent years, and its capabilities in this area are rapidly advancing,” the chatbot told Foreign Policy, referencing China’s aforementioned 2030 benchmark. “China’s strengths in AI include natural language processing, computer vision, and machine learning. The country has made significant progress in developing AI-powered applications for various industries, such as healthcare, transportation, and finance,” it added. It wasn’t all praise though: The bot also called out China’s “policies and initiatives aimed at collecting and analyzing large amounts of data, which has raised concerns about potential misuse and infringement on privacy.”
“Overall, China’s AI capabilities are rapidly advancing, and the country is poised to become a major player in the global AI landscape,” ChatGPT said. “However, it will be important for China to address concerns about privacy and ensure that AI is developed and used responsibly.”