China’s best AI technology is facial recognition, starting to recognize African faces

Almost a year ago, we mentioned China has played facial recognition to the hilt (see link). Till this day, many experts love discussing the AI race between the US versus China and betting on the ultimate winner.

During the McKinsey interview, tech guru Kai-Fu Lee gave a rather diplomatic answer: 50-50 chance (see link). Meanwhile, he did emphasize that

most users are willing to trade off some degree of privacy for convenience makes the data acquisition even easier…Chinese government has always been techno-utilitarian…as opposed to the Western countries, which tend to want to debate and resolve issues that may relate to privacy, security, bias, and explainable AI…”

Also according to another interview with Jung Min, the managing partner from Goldman Sachs,

“…if you think about differences between those two countries, there are different expectations, different norms around privacy in particular…in China there has been an explosion of security and surveillance applications, taking advantage of the advances around computer vision in artificial intelligence…compared to the US, we see more commercialization of AI in China…” (see link)

All these conclusions point to the trend that China should be already leading the wave of facial recognition in this particular vertical of AI. China even has nurtured the world’s highest-valued AI startup-SenseTime, again specializing in facial recognition.

Now China is also launching its first AI project in Africa. CloudWalk Technology, a Guangzhou-based startup, has signed a deal with the Zimbabwean government to provide a mass facial recognition program. The company has been powered by the public fund such as Guangzhou city municipal government.

In a report by Jeffrey Ding at Oxford’s Future of Humanity Institute, he finds (see link):

“The U.S. is ahead in all AI metrics except the volume of data to which it has access. Even there, Chinese AI has the benefit of much more data; but it is all Chinese data, collected at home, and thus narrow…”

But perhaps he is wrong. It looks like China determines to gather data from overseas and will feed facial recognition algorithm with African faces. In the US, earlier this year, MIT researchers Joy Buolamwini and Timnit Gebru highlighted that American technology of face recognition is racist, basically biased against black people: darker skinned faces are underrepresented in the datasets used to train them, leaving facial recognition more inaccurate when looking at dark faces (see link).

So the project with African countries will give China an edge to leap far ahead of the US in facial recognition.

By Cecilia Wu