Sat, Mar 14, 2020

The dangers of decoupling

Smart Partnerships Series by Julian Mueller-Kaler

Related Experts: Mathew J. Burrows,

Americas China Civil Society Economy & Business Europe & Eurasia Indo-Pacific Politics & Diplomacy Security & Defense Technology & Innovation Trade

This report recap is part of the AI & China: Smart Partnerships for Global Challenges project, focused on data and AI. The Atlantic Council GeoTech Center produces events, pioneers efforts, and promotes educational activities on the Future of Work, Data, Trust, Space, and Health to inform leaders across sectors. We do this by:

  • Identifying public and private sector choices affecting the use of new technologies and data that explicitly benefit people, prosperity, and peace.
  • Recommending positive paths forward to help markets and societies adapt in light of technology- and data-induced changes.
  • Determining priorities for future investment and cooperation between public and private sector entities seeking to develop new technologies and data initiatives specifically for global benefit.

@acgeotech: Championing positive paths forward that nations, economies, and societies can pursue to ensure new technologies and data empower people, prosperity, and peace.

Summary of the Roundtables in China on AI and Sino-US Relations January 2020

  1. Roundtable I in Beijing
  2. Roundtable II in Beijing
  3. Roundtable III in Shanghai

Artificial intelligence will most likely influence every aspect of modern life and decision makers are under pressure to prepare for a new world in the digital age. In order to establish forums, enable discussions about opportunities and challenges of modern technologies, and evaluate their implications for society as well as international relations, the Atlantic Council’s Foresight, Strategy, and Risks Initiative was awarded a Rockefeller Foundation grant that helped lay the groundwork for a new GeoTech Center at the Council. The tasks at hand are the publication of reports and the organization of conferences in Europe, China, Africa, and India. At these gatherings, international participants evaluate how AI will influence their societies and how countries can successfully collaborate on emerging technologies, while putting a special emphasis on China, the country’s use of AI as a development tool, and its role as a global citizen in an ever changing world.

After the successful completion of meetings in Paris, Brussels, and Berlin, a delegation from the Atlantic Council went to China in early January 2020 to discuss aspects of the following questions: “What are China’s views on the role of AI and other emerging technologies in societal and economic development, inside the People’s Republic and within other countries?”; “How do Chinese leader envisage using AI and related technologies to transform their society?”; “Can China, the US, and Europe find common ground in a world where AI and other emerging technologies are shaping global relations?”; and finally, “Will a more economically and technological competitive China be a better global citizen?” Two roundtables in Beijing and a third one in Shanghai brought together leaders from government, civil society, academia, and think tanks. In order to ensure freer exchanges, all discussions were held under strict Chatham House Rules. That is why the subsequent summaries cover key ideas, trends, and policy recommendations, however, they do not make specific attributions.

Roundtable I in Beijing

The first roundtable in Beijing featured speakers from China’s national security as well as government related think tank community. According to them, the People’s Republic (PRC) puts great emphasis and importance on the development of AI and other emerging technologies, as they are regarded as new ways for increasing productivity. For China, emerging technologies serve as a cornerstone for further economic development, taking into consideration that the digital economy accounts already for 1/3 of the country’s GDP today. Hopes are that modern technologies can help increase living standards, smarten the production industry, and transform the agricultural sector as well as the finance industry. With that in mind, Chinese participants highlighted that the PRC is already in the process of rapidly digitalizing its economy, and decision makers know about the importance of developing the next generation of digital infrastructure.

Societal Challenges and Fatalism

While many in the United States look at the emergence of modern technologies with concern, worrying about economic disruptions, it was remarkable to hear that China seems to emphasize more the opportunities that come with artificial intelligence. One participant used online shopping as an example — its rise might be responsible for the closure of local shops, but also developed a lot of new jobs in the delivery business. He thought that the development of modern technologies is inevitable, anyway, so the task of governments is to steer and guide such progress, “so that it moves into the right direction.” There was a general notion that China has already seen many economic disruptions, but has always been able to adopt successfully, given the governments mission and ability to create millions of jobs each year. Today, policy makers are changing the education system, beginning with primary schooling to better prepare students for a future technological world.  

Chinese participants admitted that compared with the United States, the PRC still lags in terms of needed skill levels and innovative capacities. Many saw China’s strength rather in AI applications, not as much in “theoretical innovation,” which they thought was the US’s strong suit. One Chinese participant stated that “till now, we have been far behind when it comes to human talent. Look at the tech giants here in China: all of their important employees or even CEOs came back from the United States after receiving their education abroad. China’s strength lies in the application of modern technologies, such as AliPay, and the respective pulling factors. Modern technologies might disrupt jobs here, too, but they will first and foremost increase productivity and guarantee social progress.”

Potential for Cooperation

Cooperation with the US and other foreign countries was, of course, important to China’s economic development. It was said that “the most successful papers on AI often come from international research collaborations, particularly between the United States and Chinese academics. The PRC’s eight AI principles, for example, emphasize cooperation, sharing knowledge, and reliance on open source methods.”  

The following discussion about potential cooperation on the international stage kicked off with an acknowledgement of the recent issuance of guidelines by the United States, the European Union, and China — regulations for respective authorities on how to govern the development of artificial intelligence. There was general agreement that last year’s G20 meeting was a big milestone in this regard, as it facilitated the agreement on five guiding principles regarding AI. Chinese participants highlighted that despite current bilateral tensions, there are indeed shared views between the US and China that could enable cooperation: both governments put emphasis on talent, research, and a fair application of artificial intelligence. However, speakers also highlighted respective differences: “While China seems to take a micro approach, the US uses a macro approach when it comes to modern technologies.” Furthermore, China focuses more on agriculture and finance, while the US, in their mind, puts its emphasis on healthcare and education. Many Chinese speakers expressed their surprise and discomfort with the US seeing China primarily as a rival. The general notion on their side was, “yes, it is important to recognize differences, but the focus should be on finding opportunities for common ground.”  

Experts thought that both governments could do joint investments in digital infrastructure that have global impact or develop political guidelines for the use of AI in order to ensure better applications for the people. “It will be difficult to cooperate on [AI] security, but we have a responsibility to preserve a strategic balance in this regard”, one participant explained. Many of the Chinese speakers were worried that the United States is closing itself to China and asked numerous questions of the American delegation about the political winds in the US. “The current American government seems to think that China is a threat. Protectionism might come with short term political benefits, but will hurt both countries in the long term. An open approach will be the most successful way forward. AI and other modern technologies rely on the open source community and those in the tech world don’t think like the people in the political realm. They still emphasize cooperation, instead of protectionism. Let us therefore overcome the nationalist tendencies of politics.” Another Chinese participant interjected: “Recent developments of US-China relations are a great concern for me. Starting last year, some people speak of a new Cold War on tech. The same people are now advocating for economic decoupling between the two countries. So I am happy to see this discussion’s emphasis on cooperation, but I am really worried about the future given the current political climate.”

Reversing the Trend Towards Decoupling

Atlantic Council participants asked what the Chinese side thought would be good joint projects that could reverse the current trends away from cooperation. One Chinese participant suggested jointly working on protecting the core global cyber infrastructure. “The internet is universal, everybody is using it — and so is artificial intelligence.” Another one thought that because AI and other modern technologies have already been applied to military units, a solely civilian regulatory approach would be “almost impossible” at this stage. Rather, arms control mechanisms should serve as models for technological cooperation.

Several participants indicated that the Chinese organizations they represented had cooperation agreements with multiple international and American organizations, for example, US think tanks, Harvard’s Belfer Center, or semi-government entities. In recent years, that cooperation had focused on emerging technologies with AI guidelines being at the center of attention. One Chinese speaker indicated that he agrees whole-heartedly that international guidelines should be developed, in order to assure that AI and other modern technologies serve as a source for good. “Artificial intelligence, encryption, and decryption techniques may advance offensive as well as defensive capacities, and modern technologies will have significant effects on society, given the emergence of deep fakes. Examples like this illustrate the dire need to develop good governance for the regulation of artificial intelligence.” He emphasized that for the successful development of such guidelines, it is not just policy people that are needed, but tech experts, too. “Given public perception as well as the chaotic state of cyber governance, the road forward for regulation, however, will be very difficult.” Another head of a Chinese think tank that participated in the discussion emphasized the necessity of communication, which, according to him, is the only way to foster collaboration. “Dialogue of civil organizations can enable government cooperation in the long run. Together, think tanks in United States and China can bring a wide range of military, government, and technological experts to the table and facilitate cooperation.” 

More concrete was the proposal of a Chinese professor who said that her “research teams have developed ideas that could combine the multi-stakeholder and multilateral approach. They don’t necessarily need to be mutually exclusive. It’s key not to assume that it is a single entity that governs AI or modern technologies, but supervision is dependent on the level and importance. Think tanks can play an essential role, because the emergence of modern technologies has already surpassed the government capacity of authorities and international entities anyway. Technologies develop faster than respective regulations.”  

Worry About Political Factors Undermining Stability

Despite emphasizing the need for cooperation, Chinese participants were well aware that current political forces are not in favor of collaboration, as evidenced by the difficult trade talks. The Chinese do not want economic decoupling to happen, though many were worried that such political tendencies are here to stay. Chinese participants asked about potential time frames and highlighted that even if there’s political will, there are also technical obstacles for cooperation as the lack of a shared definition of AI exemplifies. By answering a question from a US participant about ways to ensure strategic stability, a Chinese speaker said that “the major powers have indeed the responsibility for strategic stability. The United States, China, Russia, and the EU must ensure that modern weapon systems, emerging technologies, and artificial intelligence do not lead to mutual destruction, but enable cooperation in the world. A very practical approach would be the United Nations Group of Governmental Experts on Information Security (UNGGE), where not only state actors can put up proposals, but also think tanks or other civil entities.”

Others were more pessimistic: “AI as a new technology is producing far reaching implications for societies and nations around the world. In fact, it might be the playing field for new international competition. Different societies have diverging approaches on how to regulate. China, which is still a developing country, focuses on the practical approaches of AI for society and people’s everyday life. Even though we recognize the international dimension of emerging technologies, we don’t overestimate its competitive nature. China wants a peaceful rise and is most worried about decoupling.” However, and that was one of the most important points of the meeting, it was stated that some see decoupling as an economic opportunity for China. According to them, “decoupling will point up China’s weaknesses and need for urgent remedial actions.” The same participant continued by saying that “AI’s impact on international relations is significant as it can lower the threshold of using force. I am pessimistic about the future. The increasing human dependency on modern technologies makes us more vulnerable, and traditional means of cooperation are no longer sufficient as modern technologies are a completely new area. Traditional regulatory approaches focused on defining, identifying, and organizing, for example, the use of missiles. Such an approach, however, is no longer useful, given the different nature of modern technologies.”

The roundtable concluded by participants agreeing that there is significant risk of non-cooperation. Without regulation, the development and use of emerging technologies might spiral out of control. Some suggested that countries should refrain from using AI in military aspects until international guidelines on AI’s military uses are developed. Of course, a rather unrealistic scenario, but only active communication and negotiations can build needed trust among countries, the United States and China in particular. Trust that is essential for a successful cooperation on emerging technologies.

Roundtable II in Beijing

The second roundtable in Beijing was hosted by an independent think tank. However, the panel contained experts closely linked to the PRC’s Ministry of Foreign Affairs, many with a background and experience in arms control. Our host explained that his think tank has extensive experience in organizing and partnering on conferences about disarmament and arms control, but increasingly expands its efforts to discuss digital questions. His introduction listed some of the major difficulties, when dealing with AI: 1) What norms and principles do we want to see agreed upon at the international level? 2) What moral guidelines should be set? 3) What are the military implications of artificial intelligence? 4) What are peaceful uses of AI? 5) How can we emphasize the good elements of modern technologies, instead of being overwhelmed by the bad ones? Even though there might be expert-level proposals on these questions out there, there’s no agreement at the international policy level yet. According to him, it is therefore important to keep the political dimension in mind, because AI will significantly impact bilateral as well as multilateral relations of the future. 

Social and Economic Impacts from Artificial Intelligence

While everyone agreed that the technology revolution started in the United States, Chinese experts that were present at the roundtable thought that their country had been able to catch up “pretty quickly.” While a generational divide remains, the Chinese government has tried to push modern technologies in order to improve people’s lives, according to participants. “China is very good at AI application, less so at innovation.” It became clear that modern technologies gained significant importance over the last couple of years, but the question of how a real AI “breakthrough” would look like remains. Would it be a tipping point when robots become too smart, superseding human capabilities? Or is this a scenario still too far away and unrealistic? Generally speaking, Chinese participants thought that we should be cautious when it comes to the hype surrounding artificial intelligence. According to them, AI is still weak and not yet used widely in society. However, one of the discussants thought that AI is going to have similar effects that the internet had, creating huge markets and disrupting society. He continued by saying that taking different cultural heritages into consideration will be important when it comes to AI applications. 

Increasing unemployment and inequality are concerns for China, too, and politicians need to address them. One participant suggested it might be necessary to revisit socialist elements that could counter capitalist notions of increasing efficiency and maximizing profit. “Every year, the Chinese government needs to create 16 million new jobs already. This endeavor is going to be even more critical, once social implications of AI start to occur.”

Implications of AI for International Relations

When the discussion came to international politics, Chinese roundtable participants thought that the United States misinterprets Chinese goals significantly. “The PRC never said that it wants to be number one with regard to artificial intelligence — it only wants to be among the leading group.” Subsequently, decoupling is, according to them, a “really stupid policy.” When economies are decoupled, the conditions won’t be right for the rapid development of AI and other modern technologies. “Decoupling will stop science and progress. Do decision makers in the United States really want to see a fragmentation of the internet?” Participants voiced incredulity as to why the US wants to decouple. Such a policy will hurt everybody, although participants thought that China maybe suffers more.

According to Chinese speakers, the PRC will continue to try to engage. If political tensions remain, track two- or one-and-a-half-diplomacy should replace government-to-government consultations. Participants further thought that China is in the position to make more concessions on market-opening in order to prevent a further deterioration of the Sino-US relationship. “The United States should be more confident, when it comes to innovation, and decision makers should stop thinking about a new Cold War. We all are living in a world of advancing globalization. If China and US cooperate, it will be beneficial for the two countries and the world. AI winter is coming, but we don’t know where it will get cold first.” For the benefit of humankind, participants highlighted the need for cooperation, especially between China and the US. 

Impact of AI on Global Stability

Non-cooperation, on the other hand, could lead to fatal consequences, as militaries are already developing robotic systems and drones. For China, the use of AI in the military sense is still very limited, according to several Chinese participants. “We are still thinking through how AI will change military strategy.” Furthermore, the PRC seems to be very concerned about putting technologies in the wrong hands. Drones, for example, can be widely used in a variety of civilian fields, but Chinese authorities favor more control, rather than less and, in this sense, do not want markets to determine the parameters. According to Chinese speakers, the main challenge of modern technologies is setting reliable limitations. One expert worried “what happens when robots start to make decisions by themselves?”

Chinese participants furthermore disputed US interpretations of Chinese goals to use AI in order to obtain a military advantage over the United States. “With regard to the word ‘strategy’, its translation and US interpretation is wrong. The PRC doesn’t mean strategic advantage, but rather global stability.” Chinese experts were emphatic that confidence-building mechanisms are very important, although they admitted an arms control effort on AI would be difficult, given the perceived need for secrecy about one’s capabilities by both, the United States and China.

Roundtable III in Shanghai

The third roundtable, held in Shanghai, featured academics from local universities, although some of them work closely with the Chinese government, including law enforcement entities. Conversations started with emphasizing the “black box” nature of artificial intelligence. “In traditional AI, people know why the machine reaches a certain judgement, while in modern AI, we don’t always understand how programs reach their conclusions,” one Chinese academic explained. Unlike in the other roundtables, academics drew attention to the negative aspects of artificial intelligence as well as its uses in protecting and controlling society. “We shouldn’t just develop AI and sacrifice human intelligence on the way.” Roundtable participants highlighted the importance of a common definition of artificial intelligence, if the US and China were to cooperate on its uses. Another Chinese academic weighed in and said that AI is a very complex issue, given its wide-ranging field. Considerations that make it a very difficult endeavor to find a common definition. “Most of my colleagues and students don’t really understand how to approach the subject matter. Maybe we need to cut down the scope of the definition and focus on what is sensitive, and what is not sensitive instead, allowing China and the US to strike agreements in some uncontroversial areas.”

Humanity Should Not Be a Prisoner of AI

Some of the discussants focused on the usefulness of AI for policymaking. “Human intelligence remains important, but artificial intelligence can help improve rational decision making,” one academic argued. He also noted that “data is not really an issue anymore, because it has become a lot more available.” With this, AI is seen by some as even being able to predict future events, although challenges to such practices remain. “We used to rely on information from the intelligence community and experts etc., but now that AI can learn from bigger pools of data”, there is a temptation to use it to ignore qualitative judgment and just use Big Data to forecast the future. The same person questioned the general assumption that learning from history enables us to predict the future. “Big Data still has dark spots where it is hard to find real patterns.” Too heavy of a reliance on AI could lead to false conclusions due to human bias, which he worried about. “Believing in patterns may create a dangerous self-fulfilling prophecy. As humans, however, we should have the ability to start anew and break with patterns — thereby generating a source of innovation.”

Participants agreed that AI might make that harder, given its reliance on learning from past data. “Do we really want to live in a post-human world? I believe building human confidence is key for success,” the academic concluded. Other voices at the meeting thought that the “collection of Big Data always undermines democracy.” One academic continued to say that “in the United States, there is at least a democracy that can be undermined. The internet is leading to the extinction of local languages, cultures, and sub-groups in China, with the emphasis by Big Data on extrapolating common patterns across all society. Such ‘mainstreaming’ might have more negative effects in China than in the US.”

Large-Scale Social Disruptions Leading Back to the Ideas of Communism

Much more than our Beijing interlocutors, the Shanghai academics thought that both the United States and China should prepare for a “post-work society”, even though there was disagreement on what that actually meant. What will happen with, what one participant termed, the “useless class” — those who will see their jobs being eliminated and cannot find employment anymore? Other Chinese participants suggested that socialist ideas might need to be revisited in a completely automated society. “Universal basic income or similar measures might be the right way forward to solve emerging social problems with modern technologies.” Facing such a future world, discussants agreed that there’s an absolute necessity to work together, tackling the negative ramifications of a “non-work” or even a “post-human world.” 

Significant Differences Between US and China

In China “not all data is equal,” another participant said. There’s “open data,” which is publicly available and “secret data,” mostly collected by the government. “So, the privacy question is defined by the different nature of data in China.” Privacy, understood in the US sense, is really not applicable to the PRC, given the fact that data is already collected everywhere. “I do not think that AI will take over the world, though it can help to make decision-making smarter. Modern technologies will not change the status quo of power politics, however, they might improve decision-makers’ abilities to develop smart policy.” The same academic concluded by explaining that in the West, AI might be trained by millions of users, while in China, it already is trained when applied. Many were worried about the use of modern technologies for bad purposes, such as assassinations, killings, etc. as well as its implications for international politics: “If you listen to Peter Navarro, he thinks every AI program being developed in China is ultimately aiming to undermine US dominance, but that’s wrong. We could and should use modern technologies to improve international, as well as bilateral relations, fostering more collaborative approaches on economic and security questions.” 

There was no disagreement among participants that new technologies have always changed societies. And so will AI, which exemplifies the importance of cooperating and finding common solutions on the international level. One Chinese academic cited the recent US call for Chinese cooperation on Afghanistan. “Such regional cooperation can serve as an example and need to be expended to the tech realm.” Another cited the problems resulting from differing meanings of “military” in China and the United States. According to him, working with the military in China is more widespread than in the United States. “Chinese civilian medical researchers normally work with the military. But because of their collaboration, their counterparts in the US always have a hard time partnering with them.” 

The discussants recognized that intellectual property theft is indeed a serious issue for the United States, but medical research, whether within military or civilian facilities, should not be considered sensitive by Western authorities, some suggested. “Cancer research, for example, is a human endeavor and should not be impeded by global power competition [between US and China].” One Chinese participant thought it might be best to initially base the Sino-US cooperation on civilian aspects of modern technologies and avoid the more military uses. “Military arms control for China is very sensitive and politically driven, which will make the entire process extremely difficult.” Another academic said that it was important to recognize Europe’s role. “Its policymakers developed a strong sense for privacy protection, while China and the US both lag behind in that area.” Cooperation on AI should therefore always be triangular. Additionally, he thought it is wise not to start with the most difficult issue. “Internet security is one area of potential cooperation. The internet does not know any boarders, so regulations shouldn’t either.”

AI and Terrorism

According to the academic with law enforcement expertise and experience, “being truly unknown to authorities has basically become an impossibility in China. For that reason, it’s easier for the state to fight ‘lone wolves.'” And while security forces in the past often relied on human intelligence, they can now use big data to locate, contain, and potentially fight “terrorists.” Counterterrorism efforts in China therefore have become more efficient, according to the same speaker. “Trends that are supported by the fact that the Chinese public feels more comfortable with AI and other surveillance methods than people in other countries.” Allegedly, China has learned a lot from Western countries when it comes to security and policing. Facing a common enemy often unites forces and successful collaboration has occurred on a variety of levels, for example at the UN, other multilateral bodies, and on a bilateral basis. 

Broadly speaking, Chinese participants highlighted that big events have always been an opportunity for cooperation. “When China worked together with other countries to secure the 2008 Olympics, the country looked outside and followed international standards.” While such sport events present a wonderful opportunity for security cooperation, they also pose a significant challenge, given the potential for an international crisis to develop over a security incident. “Terrorism remains a big threat, in China, too — particularly when terrorists hold legal passports, join terrorist organizations, and then return to their home countries. Often, there are warnings, but not sufficient evidence. So what do you do? You can’t really arrest them without proof.” It was said that with regards to surveillance mechanisms, facial recognition is used most by Chinese authorities. However, guaranteeing security and stability is often a matter of using different technologies that can create an holistic picture of a human, in case all the different data are amalgamated.

Controversial Discussion about Xinjiang

Subsequent discussions of what is happening in Xinjiang sparked significant differences between Chinese and US participants. One of the non-Chinese discussants posed the question as to whether “China is culturally much more comfortable with the notion that order/stability trumps liberty, such as the world is witnessing in Xinjiang province.” While recognizing that the topic is a very sensitive issue, Chinese participants emphasized that it is not “really a question of AI, but rather one of policy.” One academic suggested that if the United States is unhappy with Chinese policy in Xinjiang, then it should be declared a sensitive issue by both sides, but it “should not impede cooperation on emerging technologies.” Not being satisfied with such an answer, the US side pressed the issue, saying that the Chinese treatment of Uyghurs is of great importance for Western policymakers as well as for public perceptions. Without being surprised by the US participants’ reaction, Chinese academics contended that “reeducation camps”, as portrayed in the Western media, were more “schools” than camps or prisons. One of the experts who allegedly had firsthand knowledge of the internment facilities asserted that the real problem for the government in Beijing is the increasing political separatism on site. “There are many minorities in Xinjiang province — Uyghurs are only one group. The region has a very complex history, but the main concern is about the people’s desire for independence, rather than terrorism, religion, or ethnicity. The West might focus on the ethnic or religious aspects, but from a Chinese perspective it’s a broader societal and state question.” 

In conclusion, participants emphasized that AI application is different in China because of Beijing’s policies. That, however, does not mean that AI itself is different. Smart AI might even become more transparent, showing how it reaches conclusions despite its current “black box nature.” Geopolitically, the different ways that AI are likely to be implemented in China and the US pose great dangers — making international collaboration ever more relevant as it is key for lowering risks. Chinese participants highlighted that “if the world wants China to change, cooperation is much more effective than confrontation.”

Interested in learning more about the events in China? The Economist wrote a Chaguan column about the roundtables in Beijing and Shanghai: “Different views of AI fuel distrust between China and America”

The meetings in China were part of the Atlantic Council’s ongoing endeavor to establish forums, enable discussions about opportunities and challenges of modern technologies, and evaluate their implications for society as well as international relations — efforts that are championed by the newly established GeoTech Center. Prior to its formation and to help lay the groundwork for the launch of the Center in March 2020, the Atlantic Council’s Foresight, Strategy, and Risks Initiative was awarded a Rockefeller Foundation grant to evaluate China’s role as a global citizen and the country’s use of AI as a development tool. The work that the grant commissioned the Atlantic Council to do focused on data and AI efforts by China around the world, included the publication of reports, and the organization of conferences in Europe, China, Africa, and India. At these gatherings, international participants evaluate how AI and the collection of data will influence their societies, and how countries can successfully collaborate on emerging technologies, while putting a special emphasis on the People’s Republic in an ever-changing world. Meetings in Africa and India are scheduled to take place later this year and summaries of the roundtables in Paris, Brussels, and Berlin have been published.

More about the GeoTech Center and Commission

Championing new technologies and data to benefit people, prosperity, and peace.