Why you should care
Washington fears that China could be using private sector proxies to steal secrets from the West.
The two men posing for photographs in a Nanjing conference room could not have more different backgrounds. On one side was Mao Yongqing, head of the 28th Research Institute of China Electronics Technology Group Corp. (CETC), which develops electronic warfare technology for the People’s Liberation Army (PLA). On the other was Yin Shiming, vice president of cloud computing at Baidu, one of China’s privately owned internet groups.
Mao is one of a small group of state cadres entrusted by China’s leader Xi Jinping with pushing the military into the era of artificial intelligence. Yin is an engineer who built his expertise at some of the most important Western tech companies, including Apple. But at the ceremony this year, the two men smiled and lifted a red silk scarf to unveil a bronze plaque that declared CETC and Baidu to be partners in a “joint lab for intelligent command and control technology” — the facilities that are used to direct military operations.
Mao lauded the deal as an implementation of “military-civil fusion,” an instruction by the Chinese Communist Party that new technologies developed by the private sector must be shared with the military, which Xi had written into the constitution last year. Yin said CETC and Baidu should “work hand in hand to link up computing, data and logic resources to further advance the application of new-generation AI technologies in the area of defense.”
The Chinese drive for this form of military-civil fusion is the source of nightmares for Western governments and one of the motivations for the increasingly confrontational approach U.S. President Donald Trump is taking toward China.
While old-fashioned warships are deployed in the South China Sea, the armed forces of both countries are also investing heavily in a new generation of weapons that they hope will give them a military edge in the coming years. The systems they are working on, including various forms of semiautonomous weapons, aim to take advantage of recent advances in robotics, quantum computing and AI.
Liberal societies facilitate the development of AI, but the Chinese state’s single-minded pursuit … puts this same openness and freedom at risk.
Elsa Kania, Center for a New American Security
The fear in Washington is that the close collaboration between China’s private sector and the PLA, including allegedly underhand efforts by the state to get hold of new U.S. technologies, is helping give Beijing an advantage in this incipient arms race.
In a speech last month that crystallized the toughening U.S. approach, Vice President Mike Pence accused the Chinese authorities of stealing “cutting-edge military blueprints” and said that “Beijing has prioritized capabilities to erode America’s military advantages on land, at sea, in the air and in space.”
An Australian think tank warned last week that China had sent thousands of scientists affiliated with its armed forces to Western universities — many disguising their military connection — as part of its effort to build a web of research collaboration that could boost Beijing’s military technology development.
As a result the AI research world, which has developed over the past couple of decades in an environment of international cooperation and free flow of ideas, is coming under the sort of scrutiny normally reserved for the weapons industry because of its potential use in both military and civilian spheres.
“The dual use of this [AI] technology is perfectly aligned,” says Sean Gourley, CEO of Primer, an AI-related startup based in San Francisco. “Image recognition can be used for selfies or for targeting.”
The Trump administration is actively seeking policy tools that would allow it to monitor and control the flow of potentially dual-use technologies out of the U.S. In a speech last month, Christopher Ford, assistant secretary of state for international security and nonproliferation, warned that technologies transferred to China by private companies could be used to threaten U.S. national security.
“Based on the explicit premises of the [Chinese Communist Party’s military-civil fusion] strategy, if any given technology is in any way accessible to China, and officials there believe it can be of any use to the country’s military … one can be quite sure that the technology will be made available for those purposes,” Ford said.
The controversy over AI is part of a broader concern about reciprocity in dealings with China. U.S. officials say that while the American economy is open to outsiders, large parts of the Chinese economy are shut to foreigners.
AI has developed in a different way from many earlier technologies as it is often explored in open collaboration between researchers using widely shared software tools. In the process, scientists and companies in both China and the West have become interwoven. Large numbers of young Chinese study and work in related disciplines in Europe and America. As many as 25 percent of graduate students in science, technology, engineering and math in the U.S. are Chinese citizens, according to an estimate from the Pentagon. Western technology companies are heavily invested in China and Chinese firms have been a growing source of funding for AI startups in Silicon Valley.
Analysts argue that this cooperation is often beneficial to both sides and that if the U.S. wants to retain its technological leadership, it will need to continue attracting talent and funds from overseas, especially China.
But they also point out that this collaborative relationship is unequal: Western societies are generally liberal and open, while in the Chinese system individual researchers and private companies can frequently be made a tool of the state — and the military.
“The entanglement in AI creates a dual-use dilemma,” says Elsa Kania, a fellow at the Center for a New American Security. “Our open and liberal societies facilitate the development of AI, but the Chinese state’s single-minded pursuit of these technologies puts this same openness and freedom at risk.”
In its attempt to develop new high-tech weapons, the Pentagon is trying to work more closely with Silicon Valley, setting up a West Coast office three years ago called the Defense Innovation Unit to help it engage with startups. However, it cannot demand the same loyalty of private companies that Beijing enjoys. Earlier this year, Google said it would not continue an AI project with the Pentagon after protests by members of its staff about the idea of applying the new technologies to weapons.
A study published last week by the Australian Strategic Policy Institute, a think tank partly funded by Canberra’s Defense Ministry, found that the PLA had sent up to 3,000 scientists to universities in Western countries during the past decade, sometimes under false pretenses, to extract know-how, often in AI-related disciplines, that could then be applied to the development of new military capabilities.
Beijing has also made its mark with massive funding of AI startups in Silicon Valley. Chinese entities funded 10 to 16 percent of all venture capital deals between 2015 and 2017, according to the Defense Innovation Unit. Some of these investments come with board representation and eventual technology transfer deals, the analysts say.
Chinese companies have also been involved in a series of corporate partnerships that could provide insights helpful to the military. Baidu has been at the forefront of such collaboration with an artificial intelligence lab in Silicon Valley established in 2014. Huawei, the telecoms equipment maker, promised $1 million for AI research at the University of California, Berkeley. Microsoft has been engaged in AI-related research and development in China for more than a decade.
CETC has partnered with the University of Technology, Sydney on AI projects. Australian officials say this raises concerns as the Chinese state company is exploring — for example, in its new joint lab with Baidu — how to put AI to use in command and control so that algorithms, rather than human soldiers, could make battlefield decisions.
The role of private companies means that industry “is moving center stage in geopolitics,” says Su Tzu-yun, a senior official at the Institute for National Defense and Security Research, a think tank backed by Taiwan’s Defense Ministry and National Security Council. “Twenty years ago, China’s impact was the China price, but in the future, it will be the China specs,” he says, suggesting that Beijing’s role is changing from a manufacturer of cheap goods to a creator of global standards.
Rattled by the challenge from China, the U.S. has started to look for new ways to slow the flow of technology. In the past few months, the Trump administration has started to overhaul regulatory tools that could help rein in China’s adaptation of Western technologies for military use.
Ford says technology policy has tried to enforce a clear demarcation between civil and military use of products. “Military-civilian fusion, of course, makes a hash of such distinctions: [its] very point is to ensure the free flow of technology and material between civilian and military enterprises,” he says. The U.S. now needs “new approaches and procedures that are alive to the magnitude of the tech-transfer problem as it exists with China.”
Earlier this year, the administration broadened the mandate of the Committee on Foreign Investment in the U.S. beyond outright acquisitions to give it oversight of minority and early-stage investments. It even introduced a “catchall” provision that could allow the committee to block any Chinese investment in Silicon Valley.
The U.S. is also putting together a list of key military components and acquisition programs that are vulnerable to technology theft or sabotage.
The next, and significantly more difficult, step is modernizing export controls. Under the Export Control Reform Act, which became law in August, the U.S. Commerce Department is set to begin asking for public comments on which emerging technologies should not be available for transfer abroad.
“First they will go after robotics, electric vehicles, AI and internet of things stuff,” says a senior industry official in Washington involved in the discussions. But industry experts say the administration’s target of having a list by the end of the year is far too ambitious.
“They really don’t know what emerging technologies to go after and how,” says the industry expert.
The task is slightly easier for so-called foundational technologies — more mature areas with a key role in national security, which the administration plans to tackle next year.
“If you really want to slow China down in its push for dual-use technology, the key is not export controls on AI itself. It’s the precision tool makers for semiconductors, that’s the chokepoint,” says one U.S. diplomat. Observers believe Washington could opt for limiting export of machinery used for spotting faults in wafer manufacturing, high-end equipment made by U.S. companies Applied Materials and LAM.
For that approach to work, however, Washington would need to consult with governments in Japan and some European countries that also have companies making such machinery. Diplomats see this as a big challenge as long as Trump has so many disagreements over trade with countries other than China.
Gourley, whose company is backed by In-Q-Tel, a venture capital group linked to the CIA, says that the easiest part of AI technology for the U.S. government to limit via export restrictions would be hardware, especially chips that were “optimized for particular jobs.”
However, the hardest area would be the algorithms that are at the heart of machine learning. Export controls that apply to software have not “traditionally been applied outside of cryptography software and weapons design software,” Gourley says. It might be possible “to expand this to include AI libraries — though because of the generality of AI algorithms it may be hard to draw a line between military and nonmilitary applications.”
China sees artificial intelligence as essential to warfare
If official pronouncements are to be believed, the People’s Liberation Army (PLA) is racing to leapfrog the U.S. military with the help of artificial intelligence.
Military Science, a Chinese periodical on military strategy and doctrine, anticipates a transition from “informationized” warfare, where the military operates using IT, to what is refers to as “intelligentized” warfare, with AI playing a key role. Beijing’s New Generation Artificial Intelligence Development Plan, unveiled in July 2017, envisions the country becoming the world’s AI leader by 2030, and explicitly lists defense as one of the areas of application.
But despite the rhetoric, the Chinese armed forces’ experiments with AI are only just beginning. Analysts believe that while Chinese military strategists are frequently exploring and debating the potential impact of AI on weaponry and doctrine, the PLA leadership has yet to reach consensus on the issue.
“So far, their main interests appear to be in harnessing AI for command and control, and to use augmented and virtual reality to make exercises more real — an important feature for a military that has essentially not fought since the [1979 Sino-Vietnamese War],” says Elsa Kania, a fellow at the Washington-based Center for a New American Security.
Regarding weapons systems, PLA scientists are looking into AI-enabled processing of images collected by drones as well as data mining for missile target recognition and navigation. Kania observes particularly strong PLA interest in autonomous weapons.
This area of unmanned warfare, some believe, will be key in the race to prevent what is ever more frequently described as an AI military arms race from getting out of hand.
OZY partners with the U.K.'s Financial Times to bring you premium analysis and features. © The Financial Times Limited 2019.