GENEVA (AP) 鈥 Top envoys from the U.S. and China huddled in closed-door talks in Geneva on Tuesday to discuss ways to ensure that emerging artificial intelligence technologies don't become existential risks.
The talks, which Presidents Joe Biden and Xi Jinping agreed to launch in last 2023, to open up bilateral dialogue between the world鈥檚 two biggest economies 鈥 and increasingly, geopolitical rivals 鈥 on a fast-moving technology that already has consequences for trade, lifestyles, culture, politics, national security and defense and much more.
U.S. technology experts say the meeting 鈥 led on the American side by high-level White House and State Department officials 鈥 could offer a glimpse into Beijing鈥檚 thinking about AI amid a generally tight-lipped Chinese approach to the technology.
Co-founder Jason Glassberg of Casaba Security in Redmond, Washington, an expert on new and emerging threats posed by AI, handicapped the meeting as a get-to-know-you that will likely yield few concrete results, but get the two sides talking.
鈥淲hat鈥檚 most important right now is that both sides realize they each have a lot to lose if AI becomes weaponized or abused,鈥 Glassberg said in an e-mail. 鈥淎ll parties involved are equally at risk. Right now, one of the biggest areas of risk is with deepfakes, particularly for use in disinformation campaigns.鈥
鈥淭his is just as big of a risk for the PRC as it is for the U.S. government,鈥 he added, referring to the People鈥檚 Republic of China.
鈥淚t鈥檚 vitally important that the United States and China begin frank discussions about how to improve AI safety," said Paul Scharre, an AI expert at the Center for New American Security think tank. "The stakes are high and the consequences for AI-related accidents could be severe. "
He noted that the United States pledged in 2022 to always maintain a human in the loop for nuclear weapons use. But China's military has not done the same.
鈥淓nsuring strict human control over nuclear weapons seems a low bar to clear for agreement on military AI,鈥 said Scharre, author of 鈥淔our Battlegrounds: Power in the Age of Artificial Intelligence.鈥 "As the world鈥檚 leading economic, military, and technological powers, agreement between the United States and China on how to manage AI risks could help set the stage for other nations to follow suit.鈥
It was not immediately clear why the meeting was taking place in Geneva, though the internationally-minded Swiss city bills itself as a hub of diplomacy and U.N. and international institutions.
The Geneva-based International Telecommunications Union 鈥 a U.N. agency currently headed by American Doreen Bogdan-Martin and previously run by China鈥檚 Houlin Zhao 鈥 is set to host its annual 鈥淎I for Good鈥 conference in the city later this month.
The meeting is the first under an intergovernmental dialogue on AI agreed upon during between Xi and Biden in San Francisco six months ago.
The U.S. government has sought to set some guardrails around the technology while fostering its growth, seeking a possible boon for economic output and jobs.
Western experts have suggested that China鈥檚 government, meanwhile, has in part kept a lid on AI applications because of its real or potential applications for military and surveillance activities under the ruling Communist Party.
U.S. officials suggested they would lay out ways to mitigate possible risks from the technology by creating voluntary commitments with the sector鈥檚 leading companies and requiring safety tests of AI products.
___
AP Technology Writer Frank Bajak in Boston contributed.
___
This story was first published on May 14, 2024. It was updated on May 15, 2024, to correct the spelling of Doreen Bogdan-Martin鈥檚 name. She is the head of the Geneva-based International Telecommunications Union.
