What do you think of Alibaba Cloud

"We should put an end to the myth that there is no discussion of AI ethics in China"

The "arms race" between the USA and China for "supremacy" in artificial intelligence is often reported. But what AI technology is being developed in China - apart from surveillance - is surprisingly rare in Europe. The Oxford scientist Jeffrey Ding wants to change that with his newsletter ChinAI. In the 1E9 interview, he does away with some of the clichés about China.

An interview by Wolfgang Kerler

The largest German television station is now asking whether we should be “afraid of China” and mentions the People's Republic in the same breath as the superpower USA. So it has long been clear in this country that China will play a decisive role in shaping the future - and also that the authoritarian giant state is relying on one technology in particular: artificial intelligence. The ambition with which the Chinese leadership is promoting AI development with investments of billions is described as exemplary. At the same time, it is critically observed how AI is used to establish a surveillance state.

China, the upcoming AI superpower, the 1984 a reality - so much seems clear. But beyond that? I don't know about you, but I often had the feeling that I only suddenly saw what was happening in China. And mostly under the same headings. Although what is happening in China is of global importance.

The then senior scientist of the Chinese Baidu group made this impression three years ago in an article by Atlantic in a nutshell: "China has a fairly deep awareness of what is happening in the English-speaking world, but the opposite is not the case," said Andrew Ng at the time about the development of Artificial Intelligence. While breakthroughs from the USA - and probably also from Europe - land in Chinese tech media within a day or two, the English-speaking world often only learns about what is going on in China with a delay. If any. Even less is reported in German.

That is more likely to be due to the comparatively small number of reporters in the huge country than to the lack of interest in AI from China. Because that is definitely there, as Jeffrey Ding observes. He has been publishing the ChinAI newsletter since the beginning of 2018, with which he wants to break down language, knowledge and prejudice barriers. It now has thousands of subscribers worldwide. The highlight of every issue: the English translation of an article from a Chinese medium or a scientific publication that Jeffrey produces himself and makes available as a Google Doc. With the current issue of March 1st, for example, Ding delivers a white paper on biometric recognition. His sources include publications such as xinzhiyuan, Leiphone, or jiqizhixinthat I had never heard of before.

Jeffrey Ding is from Iowa, USA and works as a researcher at the Future of Humanity Institute at Oxford University in Great Britain. In the Center for the Governance of AI there, he is responsible for the topic of China. He is fluent in Mandarin and has also worked in China. Since he's quite busy, he didn't have time for a phone or video interview. To do this, he answered some questions from 1E9 by e-mail - and did away with some of the clichés about China.

1E9: Let's start with the current topic that you recently covered in your newsletter: the coronavirus COVID-19, which first broke out in China. What role can artificial intelligence play in the fight against disease?

Jeffrey Ding: AI is used in a number of ways. To diagnose the coronavirus, computer vision can be used to evaluate early CT scans. However, it is unclear whether the available data are good enough to train or validate a model that can be used to differentiate the coronavirus from pneumonia.

AI can also help drug discovery, such as finding out whether existing treatments [for other diseases] are therapeutic. AI can also be helpful in developing a vaccine against the coronavirus. AI can also be used to track the spread: spatiotemporal data can help build infection models and track the path of the epidemic. Incidentally, AI companies like Baidu are already making cloud computing resources available.

When it comes to artificial intelligence, Germany primarily looks to Silicon Valley. China comes into focus when it comes to dystopian surveillance technology. Does this perspective do justice to China? What do you think?

Jeffrey Ding: No it won't. There are many innovative startups and established tech giants in China that are making impressive achievements in AI.

Which AI innovation from China particularly impressed or surprised you?

Jeffrey Ding:Residual Neural Networks, or ResNets for short, were first used by researchers in the Microsoft laboratory in Beijing, i.e. at Microsoft Research Asia. This crucial innovation from 2015 enabled researchers to build extremely deep layers of neural networks.

Which Chinese companies should we keep an eye on when it comes to AI?

Jeffrey Ding: For starters, the 15 members of the National New Generation AI Open Innovation Platform would be a good place to start.

If you want to know what is happening in the development of artificial intelligence in China, you should keep an eye on these 15 companies - some of which are well known in this country, others hardly.

What advantages do AI developers in China have over their competitors in the USA, Europe or Japan? Is it the huge domestic market? The weak protection of private data? The big tech companies? The billions in support from the state? Alltogether? Or am I completely wrong?

Jeffrey Ding: China has been hyped as an AI superpower capable of overtaking the US in the strategic AI technology area. But much of the research in support of this claim suffers from the "AI abstraction problem". This concept, according to which AI is nothing from anything with mathematics to swarms of drones, is now so vague that it is no longer analytically coherent or useful.

Therefore, a comprehensive assessment of a nation's AI capabilities requires clear distinctions with regard to the subject of assessment. We can compare the current AI capabilities of China and the US by dividing the fuzzy concept of “national AI capabilities” into three cross-sectional areas:

  1. The inputs and outputs in science and technology.

  2. The different layers of the AI ​​value chain - from the basics to technology and application.

  3. The various sub-areas of AI, for example computer vision, predictive intelligence or natural language processing.

If you follow this approach, it shows that China is not on the verge of overtaking the USA in AI technology. Rather, the US has structural advantages in the quality of science and technology, the basic layers of the AI ​​value chain and in important sub-areas of AI.

Of course, China's AI capabilities are backed by tech giants like Baidu, Alibaba, and Tencent. And they also benefit from government support in certain application areas such as intelligent surveillance.

When it comes to the “European way” in the development of AI, it is often emphasized that ethical issues are only considered here - and do not play a role in China at all. Is everything really allowed there?

Jeffrey Ding: People in China - from normal internet users to data protection officers to philosophy professors - are concerned with the ethical questions of artificial intelligence, also with regard to privacy. We should finally put an end to the myth that there is no discussion of AI ethics in China. It is reasonable to highlight differences in Chinese ideas about AI ethics, or the degree to which privacy is important to Chinese consumers. But it's really dehumanizing to say that the Chinese don't care about privacy.

Chinese tech giants are even arguing about invading users' privacy. One example of this is Tencent. The group called on the Ministry of Industry and Information Technology to intervene in a dispute between Tencent and Huawei over possible privacy violations by users of the Honor Magic smartphone.

[Another example:] After a year-long investigation, China's Shandong Province brought a major personal data abuse case against 57 individuals and 11 large data companies to court in July 2018, which sparked a debate over the interpretation of a new national specification to protect personal data .

[And a think thank], the Nandu Personal Information Protection Research Center, examined 1,550 websites and apps to see how transparent their privacy policies are.

Finally, let's talk about surveillance. Where is AI actually already being used to monitor the Chinese population or to evaluate it using scores? And is there no criticism at all?

Jeffrey Ding: AI plays a role in online censorship as well as intelligent surveillance through cameras designed for face recognition. But according to a survey by the Nandu Personal Information Protection Research Center in December 2019, more than 73 percent of Chinese respondents would prefer alternatives to sharing their facial data. 83 percent of Chinese respondents would like to access and delete their data. A professor from Tsinghua University made it particularly clear about her rejection [of face recognition].

To add a bit of European dystopia: Are there any secret factories in China that are already building autonomous killer robots?

Jeffrey Ding: Not that I know.

Cover picture: Getty Images