Opinion | Machines are struggling to recognize people in China

China has managed to create a “Great Internet Wall”. Its policies forced out giants such as Google and Uber, and insulated its economy from US Big Tech firms. This allowed the creation of home-grown rivals such as Baidu instead. The Chinese internet market has exploded over the years. It has grown to about a billion users, most of them on mobile phones, from just over 300 million in early 2010, according to the China Internet Network Information Center, a branch of the country’s ministry of industry and information.

The Chinese have also leapfrogged the US in artificial intelligence research in the intervening decade. This is especially true in the fields of neural networks and deep learning, as well as facial recognition, a technology that is now taking the world by storm. China has been using its prowess in this field to market the technology far and wide and some of the country’s increasing “aid” to other countries, which seek to use this technology, goes into buying surveillance cameras and facial recognition technology from Chinese companies such as Hikvision and Zhejiang Dahua. Importantly, the intellectual property produced actually belongs to China. It is not a faithful duplicate of someone else’s product or technology.

Facial recognition technology has also been used to great effect in China. Apart from more mundane uses such as the unlocking of cell phones or the opening of automatic doors, the Chinese government has created a vast citizen surveillance complex. Using facial recognition, it assigns social scores to citizens based on their behaviour in public places. China has gone so far as to restrict the dispensing of toilet paper in public bathrooms based on facial recognition technology.

The technology is now near ubiquitous in China. It is used to access bank accounts, unlock the front doors of apartments, and of course to access and unlock newer smartphones. Last December, the government passed a law that forces anyone registering a new mobile phone’s SIM card to undergo a face scan, in the stated interest of protecting “the legitimate rights and interest of citizens in cyberspace”, according to China’s ministry of industry and information.

This runs counter to what is happening among lawmakers in the US. The National Institute of Standards and Technology (Nist), a federal laboratory that develops standards for new technology, found “empirical evidence” that most facial-recognition algorithms exhibit “demographic differentials”. They lose accuracy based on a person’s age, gender or race. Nist’s analysis examined most of the industry’s leading systems, including 189 algorithms that were voluntarily submitted by 99 companies, academic institutions, and other developers.

Nist found that facial recognition algorithms sold in the US were 10-100 times more likely to misidentify Asians, Africans, and native Americans than they were to misidentify Caucasian people. This racial bias has caused lawmakers in the US to slow down the use of this technology. I have written in this column before about how Californian cities, including San Francisco, have come down hard on the use of this technology by their police forces.

After the outbreak of the novel coronavirus, Chinese authorities have struggled with their self-contained internet ecosystem. Both censorship and facial recognition face challenges. The authorities came down hard on eight people who first raised the alarm in Wuhan over the disease. Mostly doctors who put out warnings on various (censored and monitored) chat groups on 30 December 2019, they were called rumour mongers by the authorities.

Among them was Dr Li Wenliang, an ophthalmologist in Wuhan, who said on an instant messaging group chat that his hospital’s ophthalmology department had put in isolation wards seven patients from a local seafood market who were diagnosed with severe acute respiratory syndrome (SARS). The punishment of the eight “rumour spreaders” was widely reported by the police, local media and Wuhan health authorities. It silenced many doctors, Dr Li said, and though more of them sensed the danger of the disease, they were afraid of “being called in by the police”.

Dr Li has since succumbed to Covid-19 himself and the Chinese government has since had to relent on its hardball censorship stance on the disease. The government has even hailed Dr Li as a “hero doctor” for having raised the alarm and vigils have been organized in his honour.

Recently, Chinese authorities in some provinces have made medical face masks mandatory in public and the use and popularity of these is going up across the country. However, interestingly, as millions of masks are now worn by Chinese people, there has been an unintended consequence. Not only have the country’s near ubiquitous facial-recognition surveillance cameras been stymied, life is reported to have become difficult for ordinary citizens who use their faces for everyday things such as accessing their homes and bank accounts.

Companies such as Apple have confirmed that the facial recognition software on their phones need a view of the person’s full face, including the nose, lips and jawline, for them to work accurately. That said, a race for the next generation of facial-recognition technology is on, with algorithms that can go beyond masks. Time will tell whether they work. I bet they will.

Siddharth Pai is founder of Siana Capital, a venture fund management company focused on tech