The police officers walked past the display of robot dogs and into the ornate Beijing hotel ballroom, as a welcome video showed animated androids conducting warfare in outer space.
For the next few hours, the officers, who had come from all across China, sat and listened as tech executives and government researchers took the stage to describe how they could use artificial intelligence to perfect their surveillance of Chinese society.
One executive said his company was partnering with the police in ethnic-minority regions to deploy its speech recognition software, which he said could decipher more than 200 Chinese dialects and minority languages, in the name of national security.
A researcher for the Ministry of Public Security said that robots could reduce the need for police officers to conduct street patrols, by being trained to look out for things like protest banners.
The chief executive of another company, which describes itself as using big data to make cities safer, suggested its software could help the police understand a person’s habits, connections and even state of mind, based on that person’s medical history, online shopping habits, interactions with neighbors and use of smart home appliances.
It was hard to tell how many of the systems advertised at the conference, which was hosted by an industry group, actually worked, and how many were being hyped up with the aim of winning contracts from the many high-ranking police officials in attendance.
But it was clear that the companies and the officials shared a common goal: a Chinese state that is ever more capable of knowing, predicting and controlling its citizens.
That vision of an all-seeing state has unnerved Washington, which has cited China’s use of advanced computing to enable surveillance and repression as one reason for restricting the export of chips and semiconductors to China.
A number of the Chinese companies that attended the conference are on American export control lists, including the tech giants Huawei and Hikvision, a maker of surveillance camera systems. Several presenters acknowledged the challenges that the U.S. restrictions had created for their ambitions, but they pledged to overcome them with homegrown alternatives.
China is not alone in using advances in A.I. to heighten surveillance. The New York Police Department, for example, has drawn criticism from civil rights advocates for its widespread and often undisclosed use of tracking technology.
But the scale and sophistication of China’s surveillance apparatus are unparalleled. The state tracks its 1.4 billion people through a vast network of cameras, internet controls and requirements that people use their national IDs to buy everything from SIM cards to stand-up comedy tickets.
Now the authorities are looking to make it even more powerful, after the central Chinese government recently rolled out the “A.I.+ Action Plan,” a sweeping project to put the technology at the center of education, health care, entertainment — basically, every aspect of Chinese life.
The new political mandate was palpable at the security conference, where several speakers referred to it by name.
“Our country’s security industry has already become one of the most successful sectors in implementing A.I.+,” Gu Jianguo, a former high-ranking official at the Ministry of Public Security, said at the conference. “We don’t need to be modest.”
The conference highlighted one key advantage that Chinese companies that work with the police hold over competitors, at home and abroad, in developing cutting-edge A.I. technology: access to huge amounts of data.
The companies’ close cooperation with the government gives them an unmatched supply of information that can be used to train and refine artificial intelligence systems. An analysis published by the National Bureau of Economic Research, in the United States, found that Chinese facial recognition companies that won public security-related government contracts later went on to develop more new commercial products afterward, compared with companies that did not work with the police.
Fang Ce, a representative of iFlytek, a leading Chinese speech recognition company, said that police reports were a particularly valuable source of input for the company, because “the data volume is extremely large.” Mr. Fang, the director of iFlytek’s police affairs department, said the company’s software could help police officers log and analyze emergency calls, as well as infer whether someone was talking about a crime in a roundabout way, rather than using specific keywords.
He also said that iFlytek had developed a large-scale speech model that recognized 74 foreign languages and “for our national security, supports 202 dialects” of Chinese.
The company has been “collaborating in limited ways with local ethnic minority public security agencies” to protect ethnic languages, he added.
Human rights groups have long accused iFlytek of helping the Chinese government use voice patterns for surveillance and repression, especially of minorities in regions like Xinjiang and Tibet. The company has said that it protects users’ privacy and is interested in the preservation of dialects and languages.
Indeed, many of the presenters described the use cases for their products as benign: helping to track down missing children, for example, or combat identity fraud. Mr. Fang suggested that iFlytek’s software could be used to monitor police officers, too, for missteps or misconduct.
But companies pitched their ability to target certain types of individuals who have long been identified as potential troublemakers, including migrant workers and people who complain to the government.
One presenter from the company Quan Min Ren Zheng — which translates loosely as All-People Authentication, and makes tools to help governments and companies verify people’s identities — described how his company could help the authorities detect suspicious behavior in rental homes. His company makes products like smart door locks that could determine if people who were not authorized to live in a place had entered the premises, and electricity meters that could flag abnormal electricity usage.
Some of the technology is most likely already being used. One state-owned company at the conference, SDIC Intelligence, won a $2 million contract last year to provide a county government in southwestern China with a platform that could analyze people’s “frequented areas, activity patterns and locations,” according to public procurement documents. It won a similar amount to help Dongguan, a city in southern China, build profiles of mobile phone users based on their movements, deliveries and contact book entries.
For all the excitement that advances in A.I. have created in the security industry, though, the technical prowess of the surveillance technology is less important than its sheer pervasiveness, said Alison Sile Chen-Zhao, a researcher studying the effect of A.I. on surveillance at the University of California San Diego. As long as people believe that they are being watched, they will think twice about doing anything the government might disapprove of.
“The surveillance actually doesn’t need to achieve a very high accuracy or effectiveness to have the desired effect,” she said.
Siyi Zhao contributed research.
Vivian Wang is a China correspondent based in Beijing, where she writes about how the country’s global rise and ambitions are shaping the daily lives of its people.
The post China’s Security State Sells an A.I. Dream appeared first on New York Times.
