OpenAI CEO Calls for Cooperation with China to Combat AI Risks

Calling in from abroad to a packed conference in Beijing on Saturday to thunderous cheers from the audience, Mr. Altman said he wanted to use artificial intelligence (AI) in the face of an intensifying race for supremacy between the United States and China. He stressed the importance of collaboration between US and Chinese researchers to reduce system risk. in technology.

“China has the best AI talent in the world, so I sincerely hope that Chinese AI researchers will make a great contribution here,” Altman said.

OpenAI cannot use its services, including ChatGPT, in China.

Altman and Jeff Hinton, the so-called godfather of AI, who quit Google to warn of the potential dangers of AI, have been working with chipmaker Nvidia and generative AI leader Midjourney & General. was among a dozen American and British AI executives and senior researchers at companies in There is humanity in the conference speaker list.

A Chinese speaker at the conference faces US sanctions in 2019 for aiding the Chinese government in surveillance of US-blacklisted telecom companies Huawei, search giant Baidu, Uyghurs and others. They came from top universities and companies, such as iFlytek, a speech recognition company that has A Muslim minority in the Xinjiang Uighur Autonomous Region.

“This kind of event is extremely rare in the US-China AI conversation,” said Jenny Xiao, a partner at venture capital firm Leonis Capital who studies AI and China. “It’s important to bring together the major voices of the United States and China.” To help establish international standards, such as the AI ​​arms race, the competition among laboratories,” she added.

Last October, the United States imposed sanctions on China to block its access to cutting-edge chips needed to develop AI. In Washington, fears of a Chinese challenge to US dominance have dominated the regulatory debate.

According to a Brookings Institution analysis, China now publishes more high-quality research papers in the field than the United States, according to some indicators, but is a “paradigm-changing breakthrough.” still lags behind. When it comes to layering AI systems, China is still one to two years behind the US in development and relies on US innovations, Chinese technology watchers and industry leaders say.

While the Chinese government has made AI development a priority in recent policy statements, it is also advancing regulations to ensure compliance with China’s heavily censored internet.

The competition between Washington and China underscores the deep cross-border ties among researchers. The US and China remain each other’s biggest collaborators in AI research, according to trackers at the Center for Security and Emerging Technologies, a Washington-based think tank.

In May’s congressional testimony, Altman warned that the danger of AI regulation “is to slow down American industry in a way that China or anyone else can advance faster.”

At the same time, he added that it is important to continue the global dialogue. “This technology, wherever it is developed, will impact the American public and all of us,” he said.

The annual China Conference, hosted since 2019 by the Beijing Academy of Artificial Intelligence, a government-backed research non-profit organization that is sometimes loosely called China’s equivalent of OpenAI, brings together Chinese and Western researchers. It has become one of the most prominent forums in China to meet

This conference is aimed at a technical audience. This year’s proceedings covered a wide range of topics, from the latest large-scale language models and next-generation semiconductor design to AI applications in life sciences and self-driving cars.

Altman delivered the opening keynote for a session dedicated to AI safety and coordination, a hotly debated area of ​​research aimed at mitigating the harmful effects of AI on society. Hinton will deliver the closing address for the same session remotely later Saturday.

During a Q&A after Altman’s remarks, BAAI Chairman Hongjiang Zhang pressed Altman whether OpenAI intended to open source its research again. OpenAI, which had released code to encourage more research to build on its work, has increasingly blocked visibility into its technical systems, citing safety concerns.

“Open source can benefit the safety of AI,” said Zhang, eliciting laughter from the audience, and BAAI to increase transparency and give people a sense of control over AI development. added that it is open sourcing the model.

“In the future, we should expect to open-source more models over time,” Altman said, but there is a balance to be made to avoid abuse of the technology. added that it was important.

Altman has traveled around the world, including India, Israel and all of Europe, meeting with world leaders, students and developers. He said that as European regulators consider AI legislation, it is among the most ambitious in the world to build guardrails to address the impact of technology on human rights, health, safety and the monopolistic behavior of tech giants. It is considered one of the schemes and emphasizes prudent regulation.

Chinese regulators are also enforcing strict rules on AI development. The regulation largely overlaps with EU law, but imposes additional censorship measures that prohibit the production of false or politically sensitive speech.

Chinese tech giants and startups are rushing to develop ChatGPT equivalents and other generative AI technologies. In recent weeks, Chinese authorities have detained some individuals on suspicion of using generative AI tools to create fake news and defraud individuals.

Thousands of attendees flocked to the conference’s main venue on Friday for the opening keynotes by Meta’s chief AI scientist Yann Lucan and Massachusetts Institute of Technology professor Max Tegmark, with tens of thousands of online viewers also gathered.

Attending in person, Tegmark appeared on stage with a smile on his face, waved to the audience, and opened with a few lines of greeting in Mandarin.

After warning his audience of the catastrophic risk, Tegmark said, “Both the East and the West have the same incentives to continue building AI to reap all the benefits, but not so rapidly that it loses control.” This is the first time we’ve come to a situation where we can’t move forward.” This can be caused by careless AI development. “This is something we can all work together.”

Email Karen Hao (karen.hao@wsj.com).