Can China and Europe Cooperate in the Field of AI?
Reposted from 东不压桥研究院
Recently, the University of Oxford hosted a seminar that attracted universities, think tanks, and AI enterprises from China, the UK, and the EU. The theme of the seminar was: "Can China and Europe Cooperate in the Field of AI?" This topic holds significant practical relevance for China's AI industry today. Currently, whether large corporations or startups, many are seriously considering "going global."
According to the February 2024 Global AIGC Industry Monthly Report by Unique Research, the number of Chinese AI products expanding overseas continues to rise. Among the top 50, three are now Chinese applications.
Traditionally, the US has been the most favored market for Chinese tech companies. However, due to geopolitical risks and intensifying competition between China and the US, the market no longer seems as attractive as before. Nonetheless, factors such as a larger base of paying users, more relaxed regulatory environment and greater talent pool continue to draw many Chinese AI startups into the US market, such as 01. AI, Moonshot AI, and MiniMax. At present, these companies primarily provide virtual chatbot services that do not require excessive computational resources or large datasets and have achieved some success. For example, according to SensorTower, MiniMax's Talkie is already the 12th most downloaded AI application globally, ranking just below its American competitor Character.ai.
"How does Europe fare as an alternative overseas market?" This question has long intrigued me, which is why I attended the Oxford seminar and engaged with industry experts.
Among European destinations, the UK appears to be the most favorable choice for Chinese AI expansion. Despite the seminar coinciding with China's National Day holiday, several Chinese companies traveled long distances to participate. Discussions with attending experts suggest several key reasons why the UK is attracting Chinese AI companies:
1. Since Brexit, the UK has maintained some distance from the EU in terms of cross-border data policies and AI regulations, creating a more business-friendly regulatory environment.
2. English remains the dominant language for pre-training AI models, giving the UK a natural edge over other European contenders like France and Germany, which seek to attract AI investments but lack linguistic advantage.
3. The current Labour government is perceived to have a relatively balanced approach toward China, with geopolitical risks appearing manageable in the short term.
However, compliance costs imposed by the GDPR and AI Act have left Chinese AI firms, especially startups, hesitant about entering the EU market. Experts from Chinese AI companies universally questioned the scientific basis of the AI Act, arguing that the so-called "Brussels Effect" is, in essence, an attempt to stifle technological progress with ambiguous legal frameworks. Some scholars emphasized that AI technology is evolving rapidly, with risks emerging gradually. They argued that risk assessments should be based on empirical research rather than rigid, soon-to-be-outdated regulations that impose excessive fines on companies, hindering healthy regulation.
Beyond regulatory challenges, entering the European market entails numerous other considerations: Is there sufficient local AI talent? Are European countries welcoming to Chinese AI professionals? Is there adequate computing power? If setting up data centers is necessary, what are the costs involved?
Although Chinese AIGC firms are still evaluating their entry into the European market, major Chinese tech players like Huawei and Alibaba Cloud have already established AI infrastructure services in Europe. In recent years, Huawei has launched AI platforms based on cloud computing and data centers in Europe, aiming to expand its AI ecosystem's influence. Alibaba Cloud currently operates five data centers in Europe, with local teams in Germany, the UK, and France providing AI and machine learning solutions for industries such as automotive, retail, and gaming.
For well-known reasons, European AIGC firms have not entered the Chinese market. However, there have been some attempts in AI hardware and AI-assisted drug discovery. In 2021, Graphcore, a UK company specializing in AI intelligent processing units (IPU), entered into a strategic partnership with Digital China. As Graphcore's exclusive distributor in China, Digital China is helping build an AI reseller and channel network within the country. French pharmaceutical giant Sanofi partnered with Chinese biotech company Insilico Medicine in a $1.2 billion collaboration, where Insilico Medicine licensed its Pharma. AI platform and expert team to assist in its drug development.
Compared to the uncertainty of Chinese AI companies going to Europe, I was quite surprised by the cooperation between China and Europe in the field of basic AI research.
A representative from the Mercator Research Center shared empirical findings indicating a continuous increase in joint AI research between China and Europe from 2017 to 2022. During this period, researchers from both regions co-authored 36,951 English AI papers, making China-Europe AI research collaboration second only to China-US collaboration. By 2021, such joint publications accounted for 10% of all AI papers produced in Europe.
Unlike past high-tech collaborations where China primarily learned from and imported European technologies, China is now leading the EU in AI research. A study by the Georgetown Center for Security and Emerging Technology (CSET) found that China had already surpassed the EU in 2016 in terms of AI research citation impact, becoming the top contributor to the most highly cited AI research papers globally. China holds a substantial lead over the EU in computer vision research and is gradually surpassing it in natural language processing and robotics.
To take the above paper on joint research on AI in China and Europe as an example, we can find that many European co-authors of China-Europe AI research papers are actually Chinese nationals studying or conducting postdoctoral research in Europe. Among papers that disclosed funding sources (about half of the sample), 80% were funded by China, with 60% from Chinese government projects and most of the remainder from Chinese universities. In particular, nearly half of these joint research papers involved collaboration between Chinese and British researchers. By 2021, Chinese researchers contributed to 21% of all AI papers published in the UK, whereas British researchers contributed to only 3% of AI papers published in China. This suggests that China is making a greater contribution to the UK's AI research than vice versa.
Since 2017, two-way investment in AI between China and Europe has increased, peaking in 2021.
Overall, the share of Chinese investment in the European AI industry is small, but European investors have a high level of participation in the Chinese AI industry. According to the French Institute of International Relations, from 2019 to 2023, European tech investments in China will mainly focus on the AI sector, with the most coming from Germany and France. As of 2023, European investment in China's AI industry accounted for about 4% of the total AI investment China received, and European companies such as Siemens and Volkswagen have also set up AI labs in China.
According to statistics, 64 Chinese AI companies raised about €11.16 billion from European investors between 2017 and 2022. For example, UK-based Aurora Private Equity was a major investor in the €12.8 million Series B round of a Chinese tech company, Marknum Technology. Beyond Capital, Bosch Group's venture capital arm, has been investing in China's AI sector since 2019, particularly in companies related to autonomous driving, with WeRide, which focuses on L4-level autonomous driving, and Xingge (wholly-owned subsidiary of Cambrian), which specializes in developing autonomous driving chips, being some of the better known.
During the same period, Chinese investors invested 1.2% in European AI companies, totaling about €1.4 billion. The two largest investments went to Exscientia, a British company focused on using AI to accelerate drug discovery, and Agile Robots, a German company that develops AI-powered industrial robots. In addition, Prophesee, a French startup in the field of neuromorphic vision, received investments from China's Innovation Works, Xiaomi, and inno-chip, a Shanghai-based venture capital firm focused on semiconductors.
On the whole, the overall momentum of China-EU cooperation in the field of AI is still positive, but many experts attending the conference also pointed out that the geopolitical competition between China and the US has significantly affected China-EU cooperation.
Take the UK as an example, which has the world's top AI research institutions such as Oxford, Cambridge, and the Turing Institute. But it is very backward in terms of the transformation of technology application thus many British universities are willing to cooperate with Chinese AI companies. However, experts from universities have reflected that universities will be extra cautious when cooperating with Chinese companies because the British government considers AI to be a sensitive technology, and it takes a long time just to assess the risk. Some British universities heavily rely on funding from the defense and security sectors and have been forced to reduce or even cancel their cooperation with Chinese universities.
A British expert who used to participate in the Sunak government AI policy development, pointed out that different departments within the British government toward China's AI cooperation attitude is not the same. The international trade sector is more encouraging, but the defense, security and foreign affairs department is very cautious and emphasizes the risk. Some venture capital firms and universities in the UK have received warnings from these departments about the risks of working with China.
The UK's foreign investment review mechanism under the National Security and Investment Act 2022 has also become increasingly stringent on investments from China in high-tech sectors such as AI. In 2022, the UK government vetoed the University of Manchester's plan to license intellectual property about a visual sensing technology to Beijing Infinite Vision Technology Co., Ltd. The UK government also reviewed and vetoed a takeover bid that year for a Hong Kong company, Super Orange HK Holding Limited, to acquire British chip design software company Pulsic Limited.
The EU is promoting its own economic security strategy and "de-risking" China, and is placing more and more emphasis on "research security". Many experts at the conference estimated that the EU is likely to scrutinize the end-users of China-EU joint AI research results more strictly in the future, thus restricting the AI research cooperation. At the level of commercial cooperation, some EU companies may also voluntarily cut off cooperation with Chinese AI investors and cloud service providers for fear of reputational and political consequences. In addition, the EU's obsession with digital and technological sovereignty cannot be ignored. Given China's technological advantages in AI and the size of the AI ecosystem, the EU is also worried that Sino-European AI cooperation may one day become unbalanced, siphoning off Europe's own AI talent and turning it into a unilateral benefit for China, or leading to a tailing off of Chinese AI companies in Europe, like US companies.
Nonetheless, the mainstream view of European experts remains that European economic security strategy and de-risking is different from the US decoupling from China. Europe is not trying to curb China's technological advances, but wants to manage risks in a responsible way and in cooperation, trying to minimize risks in areas of particular concern while maintaining overwhelming cooperation with China. But it has also been pointed out that with the escalation of AI competition between China and the US, the world is likely to see the emergence of two different AI ecosystems, and the space for European AI research institutions and companies to cooperate with China and the United States at the same time will become narrower.
European experts were generally more disappointed with the lack of dialog between China and the EU on AI governance, which is an area where China-EU AI cooperation needs to be strengthened.
From the speeches of European experts, traditional issues of privacy, data security, and consumer protection remain important concerns of European governments and civil society about AI technologies and applications, and they are very serious about AI governance, even more so than the development of the AI industry itself.
Some experts believe that China is a global AI power second only to the US, and its governance practices and propositions have attracted attention from all sides. China has indeed done a lot of things in AI governance over the years, forming its own unique practices, but unfortunately has not said much to the outside world. For example, although China has launched a global AI governance initiative, European countries always feel that the initiative is relatively empty. Indeed, the specific connotation of some words is puzzling. Europe attaches great importance to AI governance, and is very interested in exchanges with China in this area, but now China seems to be willing to talk to the US only on this issue, and does not pay much attention to Europe. Strengthening the dialogue between China and Europe in the field of AI governance will be beneficial to realize the de-politicization of bilateral AI cooperation and strengthen mutual trust between the two sides in this field.
On the connotation of the global AI governance initiative, an expert from a Chinese university at the meeting responded quite appropriately, saying that the initiative is mainly about some principles, just as the "the Silk Road Economic Belt and the 21st-Century Maritime Silk Road (the Belt and Road initiative)" is a bit like a basket, in which you can put a lot of things, and that with the further development of the practice and the participation of more and more countries, the connotation will become richer and clearer.
I actually share the confusion of the European experts. Over the years, we have put forward a number of initiatives in areas such as data security and AI governance, which are China's efforts to take the initiative to lead a new type of technological global governance worthy of respect. However, I do feel more and more that China must not stop at issuing sole documents and be satisfied with emphasizing "China's propositions", sticking to and repeating the same set of words, but rather we must seriously listen to, understand and absorb the views and propositions of other countries, or else it may always be "China's propositions", but not "Global consensus".
Strengthening dialogues and exchanges on AI governance with our major partners in the global market, and absorbing their concerns, not unacceptable to us, are in fact enriching the connotation of China's global AI governance initiative, so as to make it more and more recognizable and appealing in the international arena. China does not lack various initiatives, what it lacks is the sincere recognition and following of other countries.
