China Change

Home » Analyses and Opinions » The Danger of AI Collaboration With China

The Danger of AI Collaboration With China

Matthew Robertson, October 11, 2018

 

AI, Cambricon-1A.png

 

China’s rapidly expanding interest in researching and applying artificial intelligence has been widely noted. Last year, the Chinese government published a plan to become a world leader in the field by the end of the next decade; billions of dollars are being funnelled into AI startups; and China is competing head-to-head with industry in the United States on the cutting edge of the field.

What makes AI developments in China so different from those in the United States, however, is that as with any technology, if it can be used by the Chinese Communist Party to strengthen its grip on power or further its panoptistate, it will.

This is almost a truism, of course, and military adoption of new technologies applies just as well to the U.S.

The real difference is that in China, exploitation of new technologies is almost always attendant with human rights abuses. The area of AI may end up becoming a particularly grim demonstration of this principle, if current trajectories continue.

And researchers from the West may have already given China a significant helping hand. Witness the case of the French research institute Inria’s (French Institute for Research in Computer Science and Automation) collaboration with the Chinese Academy of Sciences, and in particular its assistance in developing the technology underpinning one of China’s AI ‘unicorns,’ Cambricon (寒武纪).

Cambricon, the company featured in Science’s February 2018 profile of the burgeoning AI sector in China, was supported and spun out of the Chinese Academy of Sciences’ Institute of Computing Technology (中科院计算所). Their flagship AI chip, the Cambricon-1A, hit the market last year and has been incorporated into Huawei smartphones, among other products.

It was one of Inria’s researchers, Olivier Temam, who was instrumental in helping to lay the technical foundations of Cambricon’s breakthroughs.

“Cambricon’s founding team came from academia, and I myself was a professor and doctoral tutor at the Computing Institute of Chinese Academy of Sciences,” writes Chen Tianshi (陈天石), Cambricon’s co-founder and CEO. He goes on to thank Inria and and long-time academic collaborators Chen Yunji (陈云霁) and Olivier Temam.

Temam’s LinkedIn profile describes him as a senior research scientist at Inria from September 2004 to May 2014.

The three — Chen Tianshi, Chen Yunji, and Olivier Temam — have collaborated on a dozen journal and conference papers, including many that won best paper awards. With names like “DianNao: A Small-Footprint High-Throughput Accelerator for Ubiquitous Machine-Learning” and “ShiDianNao: Shifting Vision Processing Closer to the Sensor.”

It is papers like this that underlie innovations in AI chip development that Cambrion built its company on.

The chip architecture has another highly useful feature for China’s security authorities: use in image recognition systems for filtering and processing the bucketloads of data collected by the Communist Party’s pervasive surveillance apparatus.

VOA quoted a Cambricon employee in June 2018 commenting on a provider of surveillance cameras to the Party, Hikvision:

“A staff member at Cambricon, another Chinese company that provides the government technical support for security needs, told VOA that major video surveillance companies in the Chinese market are working with the government and that government authorities can access the information from any company at any time.”

The engineer remarked that surveillance technology, in attempting to identify ethnic minorities, might “consider beards, facial, and head accessories.”

There is as yet, at least as far as China Change could discover, no public evidence that Cambricon’s chips have been used to drive surveillance technologies.

Its chips, however, have been listed as among those that Hikvision could make easy recourse to.

Chinese tech websites have also listed “public security surveillance” (安防监控) as among the capabilities of the Cambricon-1A and 1H8 chips.

When the subject is reported in the Chinese media, surveillance technology is just another one of the potentially lucrative uses that Cambricon can exploit, alongside self-driving cars and cloud computing.

Along with Inria, MIT has also begun collaborating with iFlytek, a Chinese company whose AI technologies are being deployed by the security apparatus.

“Authorities are collaborating with iFlytek, a Chinese company that produces 80 percent of all speech recognition technology in the country, to develop a pilot surveillance system that can automatically identify targeted voices in phone conversations,” Human Rights Watch wrote in an October 2017 report.

HRW shows clearly how iFlytek has marketed the security uses of its technology, including deep relationships with official entities that have helped the authorities build the Golden Shield Project, one of the key components of China’s surveillance apparatus.

The MIT relationship with iFlytek was announced in June 2018.

Olivier Temam did not respond to an email requesting comment. Since July 2018 he has worked at Google — a somewhat ironic move given the context.

Under the guidance of Google’s former AI chief, Fei-Fei Li (李飞飞), the company opened an AI lab in China. Aside from meeting Google’s own research needs, the institution will without doubt also help fertilize China’s own AI ambitions and talent.

Li, born in Beijing in 1976, immigrated to the United States at 16 and went on to gain a BS from Princeton and a PhD from Caltech. She was appointed as AI leader and Chief Scientist at Google while on sabbatical from Stanford where she was a full professor, and while there became among the most outspoken opponents of Google cooperating with the Pentagon.

“Weaponized AI is probably one of the most sensitized topics of AI — if not THE most. This is red meat to the media to find all ways to damage Google,” Li wrote in an internal email seen by The New York Times.

But Silicon Valley, for one reason or another, does not seem to be as intent on opposing all forms of cooperation with the Chinese Communist Party and its appurtenances.

Google recently discontinued its contract with a Pentagon artificial intelligence program, saying that “we couldn’t be assured that it would align with our AI Principles,” but it has shown no qualms developing the Dragonfly search engine, in cooperation with the Chinese government, which would aid the official internet censorship regime. Meanwhile, Google has been recommending security keys manufactured by a Chinese company that has deep ties with the PLA and government.

 

 

Matthew Robertson is a Research Fellow in China Studies at the Victims of Communism Memorial Foundation. He was previously a translator and editor with China Change.

 


Related:

Google Recommends Product From a Chinese Company with Communist Party and Military Ties for its ‘Advanced Protection Program’, Matthew Robertson, August 23, 2018.

 


Support Our Work

cropped-China-Change-Logo.jpg

At China Change, a few dedicated staff bring you information about human rights, rule of law, and civil society in China. We want to help you understand aspects of China’s political landscape that are the most censored and least understood. We are a 501(c)(3) organization, and your contribution is tax-deductible. For offline donation, or donor receipt policy, check our “Become a Benefactor” page. Thank you.

 

Leave a Reply

Follow Blog via Email

Enter your email address to follow this blog and receive notifications of new posts by email.

%d bloggers like this: