Using Machine Learning To Rule – The Chinese Communist Party

In 2018 the Chinese Communist Party rules a country of 1.4 billion people, some of whom have grievances over quality of services, corruption, and public policy. To alleviate discontent, past Chinese leaders such as Hu Jintao allowed Chinese citizens to use the internet, especially blogging platforms, to air their grievances [1]. Under current President, Xi Jinping, however, China has decided on a different policy, harnessing advances in machine learning to preclude dissent and crime.

From robotic birds outfitted with cameras in western Xinjiang to facial scanners that determine how much toilet paper someone receives in eastern Shanghai, no government has shown as much enthusiasm for using artificial intelligence in governing than China’s. By 2050, China has set a goal of spending $150 billion to springboard Chinese AI capabilities past those of the US. Unlike the US, where companies have resisted cooperating with government surveillance, Chinese companies are part and parcel of the government’s surveillance efforts [2]. In 2017 China’s technology giants – Baidu, Alibaba, and Tencent – spent more than $17 billion on AI acquisition, dwarfing the $1.7 billion spent by their US counterparts [3]. Compounding the gap is the fact that China’s has the world’s largest online population, 750 million people, providing a wealth of data to continually refine its algorithms [4].

China’s vision for AI is as expansive as its population. It has 170 million surveillance cameras in its Face++ system. Face++ feeds facial data to Skynet which contains citizen’s national IDs data, allowing the state to monitor citizens’ movements [5]. In cases when a citizen jaywalks, screens later display his/her picture to shame him/her. But China’s ambitions do not stop with current offenders. According to Li Meng, Vice-Minister of Science and Technology, its eventual goal is to identify criminals before they commit crime. Skynet will identify citizens who frequent “suspicious” locations, such knife stores, update their crime risk, and flag them to authorities [6].

The piece-de-resistance of the China’s AI regime is the “social credit” system. Like a private credit score, the government hopes to track a person’s behaviors, label them as desirable or undesirable, and then train an algorithm to produce scores. The scores determine a person’s access to everything from train tickets to mortgages [7]. According to China’s supreme court, over six million Chinese have already been banned from trains for “social misdeeds” [8]. A more advanced version of the system was piloted in Jiangsu Province, where citizens received 1,000 points and then saw deductions for misdeeds like public intoxication. Citizens with scores above 950 were A-rated; those below 599 received Ds. Those with lower scores faced difficulty accessing services, like heating subsidies during winter [9].

While China’s plans for AI may be revolutionary, they face technological and political hurdles. Though the state has trumpeted its AI capabilities, in practice they are limited. For example, the Face++ system can only search a limited number of faces at a time and is localized [5]. If police are searching, for example, for someone at an airport, they have to download the subject’s face from Skynet and then re-upload it to the airport’s server. Connecting the two databases would expose Skynet to unacceptable risk. Even Face++’s ballyhooed ability to catch jaywalkers is suspect [5]. According to the New York Times, the images of jaywalkers on screen are usually a week old. The police sift through thousands of faces manually, matching them to offenders. This gap between China’s claims and its capabilities raises the question – how much of its investments in AI are about changing people’s behaviors rather than monitoring them?

Pathways to Just Digital Future

Watch this tech inequality series featuring scholars, practitioners, & activists

There is also evidence of push-back from citizenry. In the aforementioned Jiangsu pilot, residents attacked the scoring system as an overreach, comparing it to “good citizen” cards Japanese forces gave to collaborators during World War II [9]. Eventually, popular backlash forced the local government to dismantle scoring system. Still, this has not prevented other localities from marching forward. As many 40 other schemes have sprung up, covering millions of people. Backlash has also not prevented abuse either [9]. China aggressively tracks and monitors Uyghur Muslims in Xinjiang, using algorithms to identify candidates for internment in “re-education” centers. According to the UN, up to one million Uyghurs may have been caught in the dragnet [10].

To increase the legitimacy of AI-based governance, China can use machine learning to complement its anti-corruption campaign. Researchers at the University of Valladoid recently used neural networks to predict the risk of corruption in Spanish provinces [11]. A similar approach could be applied to Chinese provincial officials.  Similarly, machines learning could detect fraud in state contracts [12]. Finally, to prevent abuse of AI, the China can practice transparency. Before a person is blacklisted, the state could notify the individual. By setting up a specialized data appeals court and allowing the individual to appeal his/her blacklisting, China could be a pioneer in both AI and criminal justice. (798 words)

Bibliography

1. Christina Larson, “Who Needs Democracy When You Have Data?,” MIT Technology Review, August 20, 2018, [https://www.technologyreview.com/s/611815/who-needs-democracy-when-you-have-data], accessed November 2018.
2. Arthur Herman, “China’s Brave New World of AI,” Forbes, August 30, 2018, [https://www.forbes.com/sites/arthurherman/2018/08/30/chinas-brave-new-world-of-ai/#18fae4c728e9], accessed November 2018.
3. Hassan Chowdhury, “China’s Tech Spending More on AI Than Silicon Valley,” The Daily Telegraph Technology Intelligence, October 7, 2018, [https://www.telegraph.co.uk/technology/2018/10/07/chinas-tech-giants-spending-ai-silicon-valley], accessed November 2018.
4. Christina Larson, “China’s Massive Investment in AI Has an Insidious Downside,” Science Mag, February 8, 2018, [https://www.sciencemag.org/news/2018/02/china-s-massive-investment-artificial-intelligence-has-insidious-downside], accessed November 2018.
5. Paul Mozur, “Inside China’s Dystopian Dreams: A.I., Shame and Lots of Cameras,” The New York Times July 8, 2018, [https://www.nytimes.com/2018/07/08/business/china-surveillance-technology.html], accessed November 2018.
6. Yuan Yang, “China Seeks Glimpse of Citizens’ Future with Crime-Predicting AI,” The Financial Times July 23, 2017, [https://www.ft.com/content/5ec7093c-6e06-11e7-b9c7-15af748b60d0], accessed November 2018.
7. Tara Francis Chan, “Debtors in China Are Placed on a Blacklist That Prohibits Them from Flying, Buying Train Tickets, and Staying at Luxury Hotels,” Business Insider December 19, 2017, [https://www.businessinsider.com/chinas-tax-blacklist-shames-debtors-2017-12], accessed November 2018.
8. “China to Bar People with Bad ‘Social Credit’ from Planes, Trains,” Reuters March 16, 2018, [https://www.reuters.com/article/us-china-credit/china-to-bar-people-with-bad-social-credit-from-planes-trains-idUSKCN1GS10S], accessed November 2018.
9. Christopher Udemans, “Blacklists and redlists: How China’s Social Credit System actually works,” Business Insider October 23, 2018, [https://technode.com/2018/10/23/china-social-credit/], accessed November 2018.
10. “Detentions of Uighurs Must End, UN Tells China Amid Claims of Mass Prison Camps,” The Guardian August 30, 2018, [https://www.theguardian.com/world/2018/aug/31/detention-of-uighurs-must-end-un-tells-china-amid-claims-of-mass-prison-camps], accessed November 2018
11. Lopez-Iturriaga, Felix Javier and Pastor-Sanz, Iván, Predicting Public Corruption with Neural Networks: An Analysis of Spanish Provinces (November 22, 2017). Social Indicators Research, Forthcoming . Available at SSRN: https://ssrn.com/abstract=3075828or http://dx.doi.org/10.2139/ssrn.3075828
12. “Benefits and Risks Involved in AI-Powered Fraud Detection,” CIO Review September 12, 2018, [https://www.cioreview.com/news/benefits-and-risk-involved-in-aipowered-fraud-detection-nid-27120-cid-175.html], accessed November 2018