Artificial intelligence and facial recognition technology are emerging as major tools for the Chinese government to repress dissent, and for local police agencies to both monitor people and reduce crime. At the same time, Chinese firms are developing AI software for commercial customers, according to a recent discussion hosted by the Center for Strategic and International Studies.

How this affects China and the world was the topic for a Zoom discussion July 27 titled “The AI-Surveillance Symbiosis in China: A Big Data China Event.”

The Center for Strategic and International Studies trustee chair in Chinese Business and Economics and the Stanford Center on China’s Economy and Institutions hosted the third feature of their new collaboration, Big Data China. It highlighted the work of professors Noam Yuchtman of the London School of Economics and David Yang of Harvard University.

Chinese Business and Economics trustee chair director Scott Kennedy moderated a panel discussion that included Ilaria Mazzocco, a CSIS fellow with the trustee chair in Chinese Business and Economics, Emily Weinstein of Georgetown’s Center for Security and Emerging Technology, Paul Mozur of The New York Times, and trustee chair non-resident senior associate Paul Triolo.

“Artificial intelligence makes people think about a lot of things, but in some ways it’s a huge opportunity. It’s unbelievably impressive,” Kennedy said. “And for mortals like me, oftentimes very difficult to understand. It’s also connected to some of the most entrepreneurial parts of China. But there’s also lots of concerns about artificial intelligence as it intersects with China’s political system. We are worried about economic competitiveness in the West, national security, human rights.”

Yuchtman said since the end of the Cold War, many observers in the West believed that democracy would sustain innovation and citizens in less-democratic regimes got richer, they would demand more democratic rights.

“However, in recent years, the emergence of artificial intelligence, seen as the technological basis for a potential fourth Industrial Revolution, seems poised to challenge this reassuring consensus," he said. "A key reason for this is that AI is data-intensive technology. Since the deep learning revolution, access to large amounts of data has played a critical role in driving AI innovation, from translation to chess mastery to facial recognition technology. 

“We believe that there exists the possibility of a mutually reinforcing relationship between autocracy and AI innovation in particular," he added. "Why is that? Well, autocrats engage in intensive monitoring of their populations. They always have. And especially in contemporary autocratic societies, they collect massive amounts of data on their citizens.”

This emerging technology can be used for many purposes, including identifying people, making predictions about how individuals react, and controlling behavior. Companies that provide these services could be rewarded with lucrative government contracts, he said.

“Consider firms providing some sort of facial recognition AI services to the government,” Yuchtman said. “They’ll analyze surveillance video feeds and, most directly, they'll use those video surveillance feeds to provide services to the police force, let’s say, to help the police suppress political unrest.

“But that same data isn’t just used to make predictions for political control, it’s also useful as an input into improving facial recognition algorithms," he added. "The firm that provides AI services to the government can use that data to develop new facial recognition technology with government data as an input into innovation, and this innovation isn’t just narrow innovation for government use. It can be much broader. Trained algorithms for facial recognition can be used to identify a protester, but they can also be used to identify customers in a retail context.”

Yang said the research studied 8,000 recognition firms that have emerged in China in the last decade, and 10,000 contracts issued by local public security agencies.

They focused on political unrest events that took place across China from 2014 until today. As such events occur, local police agencies obtain more AI technology and additional high-resolution surveillance cameras.

“It was obvious that this was a very important story to bring to Washington, but the policy implications are quite complex,” Masako said. “I think it came down in part to the old question of does it mean that we need to run faster? Does it mean that we need to slow down China, and what does it mean for human rights?”

Masako said because the AI industry is globally integrated, she doubts China can dominate it. But it does make it apparent that countries must invest in talent and technology.

Weinstein termed the research “super-fascinating,” and said she really looked forward to digging more into it.

“China over the past five or 10 years has actually started developing actual AI majors at different Chinese universities and has set up specific laboratories or research centers,” she said.

The Chinese are usually fairly explicit about their strengths and their weaknesses and vulnerabilities, Weinstein said.

“And what we've seen is particularly in artificial intelligence and emerging technologies, the part that China is missing now is usually falling somewhere in the basic research level, the part where for a while we thought of China, particularly in the West, particularly in the United States, we always thought of China as a country that wasn’t necessarily capable of innovating, it kind of had this copycat culture where it could buy things or steal things and try to take them apart and rebuild them," she said. 

"And that was certainly the case,” she added. “But we’ve moved into a point now where China is shifting out of that copycat mentality and they're actually able to innovate in its own unique way.”

Triolo said he has attempted to talk with people across China regularly on what's happening there, although the COVID-19 pandemic has made that challenging.

“I think it's important to note that there's no top-down AI mastermind in China that's commanding researchers and companies that serve as part of some centralized and nefarious AI strategy for global domination,” he said.

There are concerns across the globe about the intrusive nature of AI, Triolo said.

“And then the other issue that we’ve touched a little bit upon is the effectiveness of the technology,” he said. “How is it used? What advantages does it bring? China already had a pretty good public security apparatus and capabilities before AI.”

Triolo said he thinks companies rushed into the emerging technology but fewer will do so now.

Cameras, alarms and lists

Mozur is a Pulitzer Prize-winning journalist focused on technology and geopolitics in Asia. He has covered extensively the buildup of China's surveillance capabilities, chronicling how police there have used artificial intelligence to profile minorities domestically and sold cameras and monitoring software abroad.

“At a local government level, we all know that China’s covered in cameras, but the way this is working now is that you can’t watch everybody all the time with the camera," he said. "What police do at a local level is they make lists of people that they consider their biggest problem-makers. Oftentimes, this includes people with a criminal background of theft or something like that. But also, oftentimes it includes dissidents.”

Using AI technology, police are no longer watching 10,000 people, they are looking for around 300. 

“What you do is you have a back-end software, basically take a note every time a camera detects one of these 300 people or thinks it detects one of these 300 people," Mozur said. "And it uses not just the face but the clothes they're wearing and other things to basically draw a kind of map of where they've been throughout any given day. 

“What you can then do is go back and search those maps. If you run into somebody that, say two weeks later, they're protesting, you want to go back, you can basically go back to that software and consult it," he added. "These lists are called blacklists. Most Chinese people are not aware they exist. And if they are, they don't really know who's on them.”

Police like the lists because they strongly believe it has greatly reduced theft and other crimes. Political activists, however, are closely watched and alarms are set up to alert police to their movements. The technology also can be used to perform racial profiling.

People go to ingenious lengths to try to avoid this surveillance.

“We found out for a recent report a man who has 60 years of petitioning, which was just a brilliant find, and this guy was extremely clever,” Mozer said. “His name is Zhang Yuqiao and he talks about, 20, 30 years ago, it was fairly easy to make it to Beijing. You basically would just get out of town, get onto the local roads, and then you can hire cars, take buses. You were on your way.

“Now, in a recent trip to get to Beijing, he turned off his phone, left at night, took a car, paid by cash, got to the local capital, got a train ticket to the wrong destination, which was Beijing, got off before because he believed that his buying the tickets would alert the police to pick him up when he got to Beijing, then took a bus, got off on a bus, took another car, paid money for it, got out before a checkpoint where they check IDs for buses, took another private car and then got into line with other petitioners at dawn,” he said. “So this is the level of kind of evasion that it takes now to get to Beijing.”

But now, whenever he turns off his phone, police come to his home.

This is becoming a centralized effort, he said, but it works better in some cities and regions than in other areas. Police are at times unhappy with responding to so many alarms, but this appears to be a tool that will be used extensively. With crime dropping, most Chinese residents will accept it.

“And the only moment they actually see the political impact of it is when a whole group of people get angry about some injustice, take to the streets, and all of a sudden it’s effective at going after very specific people and stopping that protest,” Mozer said. “So there is this kind of thing where it's a very pleasant thing until it’s not, and you don’t really know what’s going on in the background until sometimes it’s too late.”

He said this technology and these practices may spread across the planet. There is evidence that has already started.

“In Ecuador we saw the kind of classic China thing where you put a camera outside the house of a person, a dissident that you want to watch. I think we’re going to see more of that around the world,” Mozer said. “And I don’t really think there’s anything anybody can do to stop it. I think the best thing the U.S. can do is try to set a good example and try to encourage better practices in this and I guess that’s about it.”