20 January 2020

Export Controls Threaten the Future of AI Outposts in China


For some time, American companies including Microsoft, Google, and IBM have established research labs in China to tap into local AI talent and to keep track of technological trends. Now, as tensions and restrictions continue to ramp up, some observers wonder if the days of those outposts may be numbered.

The US Commerce Department imposed new export controls on artificial intelligence software last week, a measure apparently designed to prevent US companies from shipping AI technology that could train Chinese military drones or teach intelligence systems to interpret aerial imagery. But it remains to be seen how broadly the rules are interpreted, and they are unlikely to be the last to land on American AI algorithms, datasets, and chips. The Commerce Department is, in fact, still weighing further AI software controls, and the Trump administration is increasingly scrutinizing how Silicon Valley interacts with China.


“I don’t think any US government official expects to keep China from developing AI—they are doing quite well—but they don’t want companies like Google or Microsoft helping them,” says James Lewis, senior vice president of the Center for Strategic & International Studies, a Washington, DC, think tank.

Lewis says US companies might respond to the new export controls by looking to shrink their operations in China. “If you start to see Microsoft and Google beginning to retrench, then you’ll know which way we are going,” he says.

AI has become a lightning rod for economic and political tension between the US and China over the past few years. The field is racing ahead rapidly and there are big economic benefits and strategic advantages to be gained on both sides. China’s ambitious national AI plan, announced in July 2017, led local governments and businesses to pour billions into AI projects. Both countries are working hard to harness the technology for military ends.

The export rules are also the latest spanner to be thrown into the machinery of US-Chinese relations. It follows restrictions placed on business dealings with Chinese AI companies, tighter visa vetting for Chinese researchers, and public pushback over US companies acquiescing to Beijing’s demands.

“I don’t think any US government official expects to keep China from developing AI—they are doing quite well—but they don’t want companies like Google or Microsoft helping them.”

JAMES LEWIS, CENTER FOR STRATEGIC & INTERNATIONAL STUDIES

Google established an AI research center in Beijing in 2017, noting at the time that a significant number of AI research landmarks and papers now came from China. The company said the team there would do basic AI research, organize conferences, and collaborate with Chinese researchers. “China is home to many of the world's top experts in artificial intelligence (AI) and machine learning,” reads the blog post announcing the lab.

The decision seemed especially significant given that Google pulled its search engine from the country in 2010 in protest over government censorship and spying. This year, Google was forced to scale back efforts to build a modified version of its search product suitable for the Chinese market amid protest from employees, politicians, and the public. It is unclear how many people currently work at Google’s lab in Beijing.

Other US tech companies, like Facebook, are effectively blocked from operating in China, and have watched home-grown rivals, like the mobile app WeChat, grow to dominate the market in recent years. Efforts to replicate and replace US technology are now intensifying as the US government seeks to block technology transfer. In December, for instance, China’s central government reportedly ordered officials to remove foreign PCs and software from their offices within three years.

Lewis says companies that have established a foothold in China may be forced to accept that the tech superpowers are now decoupling. “Someone at one of these companies told me they have maybe a seven-year lifespan left in China,” he says. “Although a lot depends on how quickly the Chinese can build competing products.”

US outposts in China are more than just a symbol of cooperation. Bill Gates recognized China’s potential as a research hub in 1998, establishing the first big US research lab there, Microsoft Research Asia. The lab quickly proved its worth, and over time it has contributed to key commercial projects including Windows, Office, Xbox, and Bing. The lab was also quick to adopt advances in AI from the US, demonstrating a near real-time English to Chinese translation system in 2012. Apple also has several research facilities in China. Amazon established a research lab in Shenzhen, a manufacturing powerhouse in southern China, this year.

As China’s tech industry matured over the past few years, several of its big companies, including Baidu, DiDi Chuxing, Tencent, and Alibaba, have created outposts in the US with a heavy focus on AI. Facing new scrutiny and tougher economic headwinds at home, however, some may consider scaling back.

“Almost all American and Chinese companies are reviewing their R&D risks,” says Nina Xiang, managing editor of China Money Network and an expert of China’s tech industry. “Chinese companies have closed their US labs if they aren’t essential. It could also be partially related to cost cutting too, as all these companies are tightening their belts.”

As China’s tech industry produces more innovations of its own, it may be increasingly important for US companies to be able to borrow ideas and innovations. The White House strategy seems to ignore the fact that there might be a competitive disadvantage to cutting off ties to the country.

“Technological isolationism would be a counterproductive strategy for the US to adopt,” says Jeff Ding, an academic at Oxford University who writes a newsletter on Chinese AI. “Google, Microsoft Research Asia, and other multinationals with R&D labs in China aren't doing offshore work out of charity or some desire to help the Chinese military. They want to be plugged into global innovation networks and adopt technological advances from abroad into home bases.”

Perhaps surprisingly, given rising friction and barriers, some hold out hope that the US and China can find common ground, arguing that decoupling might be bad for innovation itself.

“Advances in machine learning benefit everyone, especially in fields like health,” says Daniel Castro, vice president of Information Technology and Innovation Foundation, a science and technology policy think tank backed by the industry. He suggests the real race in AI may not be in research at all. “Many AI advances are quickly being commoditized—so who develops the technology is somewhat less important than who adopts it.”

No comments: