Baidu is said to have cited the cost of membership and recent financial pressures for the move. But as relations between the US and China worsen, the departure comes amid growing challenges for companies and people in the two countries to collaborate, or find common ground, when it comes to critical technologies like AI.
In a statement, Baidu said it “shares the vision of the Partnership on AI and is committed to promoting the ethical development of AI technologies. We are in discussions about renewing our membership, and remain open to other opportunities to collaborate with industry peers on advancing AI.”
Terah Lyons, executive director of PAI, says the company cited a weaker financial outlook for the decision, adding: “Baidu remains committed to our mission and hopes to be able to resume membership in 2021.”
Still, the withdrawal coincides with increasing criticism of Chinese AI companies, and a more hostile attitude in Washington. The US government has tightened trade restrictions on China’s top AI firms for allegedly supplying technology used for state surveillance. This has not affected Baidu so far.
“The whole issue of China and AI became a hot potato” after Baidu joined PAI, says Paul Triolo, practice head of geotechnology at the Eurasia Group, a think tank. No other Chinese companies have joined.
PAI declined to discuss Baidu’s dues as a member. The organization’s annual report from last year said it projected revenue of $8.1 million. That suggests its 23 for-profit members each contributed in the low to mid six figures. Baidu reported 2019 revenue of $15.4 billion.
Baidu, often referred to as China’s Google, operates the country's biggest web search engine, and has diversified in recent years to focus on AI and related areas including self-driving cars. The company’s Apollo platform for automated vehicles is used by numerous automotive companies in China and elsewhere.
When Baidu joined PAI, in October 2018, it seemed to signal the potential for international collaboration around issues raised by applications of AI, including potential misuses of facial recognition, the social impact of algorithmic bias, and the safety and policy challenges of autonomous vehicles. “We cannot have a comprehensive and global conversation on AI development unless China has a seat at the table,” Lyons said at the time. More broadly, PAI has supported engaging with Chinese companies.
PAI was created in 2016 with backing from Amazon, Facebook, Google, and IBM. They were joined later by Apple as well as organizations including the UN and Human Rights Watch.
US-China relations have gone into a tailspin since Baidu joined. The US government has taken increasingly aggressive actions against Chinese tech companies, partly motivated by fears around China’s rising technological capabilities and their implications for America’s standing in the world.
Last October, the US banned six AI companies (not including Baidu) from doing business with US firms, citing their role in alleged human rights violations. In March, the government tightened restrictions aimed at telecommunications giant Huawei, which is also heavily invested in AI, to prevent it from obtaining the latest microchips.
On Monday, the US and other OECD countries also launched a collaboration called Global Partnership on AI (GPAI), aimed at “ensuring that artificial intelligence is used responsibly, respecting human rights and democratic values.”
Increasingly, the US and China seem locked in a technological Cold War centered around advanced technologies including AI, 5G, and advanced microchips. There is a sense in Washington that advances in these areas will serve as a foundation for economic and military power in coming decades, with American values at stake.
Most controversially, some Chinese AI companies have been tied to surveillance and repression of Muslims in Xinjiang, a province in northwest China. As many as a million ethnic Uighurs, Kazakhs, and other groups have reportedly been rounded up and interned in “reeducation camps” in recent years. A May 2019 report published by Human Rights Watch, a group affiliated with PAI, cited Baidu, among others, for allegedly supplying technology to police in Xinjiang.
Michael Kratsios, chief technology officer of the US, said in a statement: “The United States remains committed to the development of trustworthy AI and stands against authoritarian uses of this technology. Chinese tech companies that have been widely implicated in human rights abuses clearly do not share this vision.”
Yet the resulting decoupling of the countries’ tech industries may have unforeseen consequences. The two countries clearly depend on one another, as well as on other nations, for AI training and talent. Collaboration between US and Chinese AI researchers remains common. Some AI researchers argue that counties with different cultural attitudes towards the technology and its use can, and should, still cooperate on standards and norms of governance.
A tough US stance on Chinese tech could backfire by causing Beijing to double down on authoritarian uses of AI, contributing to the spread of such uses worldwide.
Triolo argues that it’s better to engage with Chinese companies than to decouple. He says Chinese companies can influence Chinese regulators, and offer visibility on how technology is developing in China. “There is this meme in Washington that we are part of this zero-sum battle with China for these advanced technologies,” he says. “Keeping China out of these dialogues is a mistake, it’s dangerous.”