OpenAI CEO Calls for Collaboration With China to Counter AI Risks

“China has some of the best AI talent in the world,” Altman said. “So I really hope Chinese AI researchers will make great contributions here.”

OpenAI doesn’t make available its services, including ChatGPT, in China.

Altman and Geoff Hinton, a so-called godfather of AI who quit Google to warn of the potential dangers of AI, were among more than a dozen American and British AI executives and senior researchers from companies including chip maker Nvidia and generative AI leaders Midjourney and Anthropic on the speaker list at the conference. 

Chinese speakers at the conference came from top universities and companies including U.S.-blacklisted telecom company Huawei Technologies, search giant Baidu and speech-recognition firm iFlytek, which the U.S. sanctioned in 2019 saying it aided the Chinese government in the surveillance of ethnic Uyghurs and other Muslim minorities in Xinjiang.

“This event is extremely rare in U.S.-China AI conversations,” said Jenny Xiao, a partner at venture-capital firm Leonis Capital and who researches AI and China. “It’s important to bring together leading voices in the U.S. and China to avoid issues such as AI arms racing, competition between labs and to help establish international standards,” she added.

The U.S. imposed sanctions last October on China to prevent it from accessing the most popular cutting-edge chips needed for AI development. In Washington, anxieties of China challenging U.S. dominance have loomed large in regulatory discussions. 

By some metrics, China now produces more high-quality research papers in the field than the U.S. but still lags behind in “paradigm-shifting breakthroughs,” according to an analysis from The Brookings Institution. In generative AI, the latest wave of top-tier AI systems, China remains one to two years behind U.S. development and reliant on U.S. innovations, China tech watchers and industry leaders have said. 

The Chinese government has made developing AI a priority in recent policy statements, while also pushing ahead with regulation to make sure it conforms to China’s heavily censored internet.

The competition between Washington and Beijing belies deep cross-border connections among researchers: The U.S. and China remain each other’s number one collaborators in AI research, according to a tracker from the Center for Security and Emerging Technology, a Washington-based think tank.

During a congressional testimony in May, Altman warned that a peril of AI regulation is that “you slow down American industry in such a way that China or somebody else makes faster progress.”

At the same time, he added that it was important to continue engaging in global conversations. “This technology will impact Americans and all of us wherever it’s developed,” he said.

The annual China conference, hosted since 2019 by the Beijing Academy of Artificial Intelligence, a government-sponsored research nonprofit sometimes loosely referred to as the OpenAI equivalent of China, has become one of the country’s most prominent forums for bringing together Chinese and Western researchers.

The conference is geared toward a technical audience. This year’s proceedings covered a breadth of topics ranging from the latest large language models and next-generation semiconductor design to applications of AI in life sciences and self-driving cars.

Altman delivered the opening keynote for a session dedicated to AI safety and alignment, a hotly contested area of research that aims to mitigate the harmful impacts of AI on society. Hinton will deliver the closing talk for the same session later Saturday, also dialing in remotely.

During the question-and-answer session after Altman’s prepared remarks, Zhang Hongjiang,

“Open-source could benefit AI safety,” Zhang said to laughter in the audience, adding that BAAI open-sources its models to increase transparency and give people a sense of control into AI development.

“Over time you should expect us to open-source more models in the future,” Altman said but added that it would be important to strike a balance to avoid abuses of the technology.

Altman has been on a global tour to meet with world leaders, students and developers, including in India, Israel and across Europe. He has emphasized cautious regulation as European regulators consider the AI Act, viewed as one of the most ambitious plans globally to create guardrails that would address the technology’s impact on human rights, health and safety, and on tech giants’ monopolistic behavior.

Chinese regulators have also pressed forward on enacting strict rules for AI development that share significant overlap with the EU act but impose additional censorship measures that ban generating false or politically sensitive speech.

Chinese tech giants and startups have rushed to create their ChatGPT equivalents and other generative AI technologies. Authorities in China have detained some individuals in recent weeks for allegedly using generative AI tools to produce fake news or defraud individuals. 

On Friday, thousands of participants packed into the main conference venue, alongside tens of thousands of viewers online, for the opening keynotes from Meta chief AI scientist Yann LeCun and MIT professor Max Tegmark.

Tegmark, who attended in person, strode onto the stage smiling and waved at the crowd before opening with a few lines of Mandarin.

“For the first time now we have a situation where both East and West have the same incentive to continue building AI to get to all the benefits but not go so fast that we lose control,” Tegmark said, after warning the audience about catastrophic risks that could arise from careless AI development. “This is something we can all work together on.”

Write to Karen Hao at [email protected]