The Washington Post-US-China talks on AI risks set to begin in Geneva

2024-05-12T15:09:28.796Z

The United States and China will hold their first high-level talks over the risks of artificial intelligence on Tuesday in Geneva, as the two governments seek to prevent disastrous accidents and unintended war amid an arms race for the emerging technology.

“We’re focused on how both sides define risk and safety here,” a senior Biden administration official told reporters last week, speaking on the condition of anonymity to discuss expectations for the talks.

Seth Center, the State Department deputy envoy for critical and emerging technology, and Tarun Chhabra, senior director for technology and national security at the National Security Council, will lead the U.S. delegation, the administration official said. China will be represented by officials from the Foreign Ministry and the National Development and Reform Commission, the nation’s central economic planning agency.

“AI” — a catchall term for a range of advanced computing capabilities — has loomed large in the U.S.-China rivalry, with both governments elevating it to a priority. Sophisticated computing algorithms can give a nation an edge in areas as diverse as warfare, economic output and the creation of soft-power cultural products. Researchers say AI can also be leveraged for disinformation campaigns and cyberattacks.

AI holds particular allure for militaries and intelligence agencies for its potential to help them sift through more raw data within seconds than a human could in a lifetime. Officials say the wars of the future will increasingly be fought with AI helping to make complex decisions in the heat of the moment.

The Biden administration imposed sanctions on China in October aimed at slowing its AI development by restricting its access to advanced chips, the brains of computing systems. U.S.-China tensions further flared last month after President Biden signed into law a U.S. ban on the popular short-video platform TikTok unless it sells itself to a non-Chinese buyer.

Chinese Embassy spokesperson Liu Pengyu said in a statement that the dialogue between the United States and China on AI would have effects for the future of not only the two countries, but other nations as well.

“The two sides have the responsibility to engage in candid dialogue,” he said.

Biden administration officials tempered expectations for concrete outcomes of this week’s talks, saying they are not seeking to release a joint statement or cooperate with China on AI research. An official dismissed the idea that the U.S. chips sanctions might be revisited, saying that Washington will not negotiate “national security measures.”

China’s AI development continues to lag behind the United States, but its high-tech companies like Huawei Technologies, Alibaba and Baidu have made significant strides. China may have an edge in certain aspects, such as the manual data-labeling process that is used to train AI models, due to the country’s lower labor costs.

In a commentary for the Brookings Institution think tank, scholars Graham Webster and Ryan Hass suggested these talks could produce a better shared understanding of what constitutes permissible military use of AI, and agreements on what kinds of data can be shared across borders for training AI models.

Administration officials did not say if the talks will touch on TikTok or Huawei, the China-based telecommunications giant that has faced U.S. sanctions. Both companies have made forays into AI algorithms in recent years.

Simply keeping an open line of communication may be enough of a result from the talks for now. With intensified distrust and hostility, officials on both sides say preventing the de facto cold war from accidentally turning hot — whether from an AI mishap or human bungling — should be a policy priority.

The term “artificial intelligence” dates to the 1950s, with research on intelligent computers going back further. These algorithms reached a new level of sophistication by 2015, when Google’s DeepMind division unveiled an AI program that could beat the world’s top players of Go, a classic Chinese board game considered one of the most complex games to strategize.

Under the Obama administration, the National Science and Technology Council produced a report in 2016 identifying AI as a strategic focus, recommending that federal agencies intensify their investment in the technology and keep tabs on rival nations. China set out its own national blueprint in 2017, aiming to be a world leader in AI by 2030.

The launch of ChatGPT in 2022 brought this AI development race into the public view, with the chatbot making it clear to lay consumers just how advanced and broadly useful this technology had become. Biden issued an AI executive order in October, launching a whole-of-government push to ensure the United States remains the world leader in the technology.

Even as AI holds enormous innovative promise, one risk that government officials find troubling is the potential for the vast data troves hooked up to the back ends of the AI algorithms to be hacked by adversaries. Another risk is of military accidents due to malfunction of automated systems.

There also are all sorts of thorny ethical questions as governments decide exactly what they program AI systems to do, and what margins of error they will allow. The Israeli army’s use of AI algorithms to identify individuals as bombing targets in Gaza, as reported by +972 magazine, has generated controversy in recent weeks.