This report is part of “A Blueprint for the Future of AI,” a series from the Brookings Institution that analyzes the new challenges and potential policy solutions introduced by artificial intelligence and other emerging technologies.
Under President Donald Trump, great power competition has become the organizing principle of American foreign policy. This has led to near-daily invocations of the Cold War to describe the intensifying rivalry between the United States and China, and to frequent analogies to an “arms race” to describe bilateral competition in advanced technologies, including quantum computing and artificial intelligence (AI). Public statements and national plans from both governments have reinforced this zero-sum dynamic. Such framing has done more to conceal than clarify and, if taken to its logical end-point, will do more harm than good for the United States.
AI will create both immense stress on the U.S.-China relationship as well as opportunities for potential collaboration.
This paper argues that we need a different narrative to describe the role of AI in the escalating competition between the United States and China. Even as artificial intelligence is contributing to an intensifying bilateral rivalry, it also is driving both countries to race out ahead of the rest of the world in innovation, economic growth, and overall national power. Moreover, the adoption of advanced technologies is hastening the arrival of intense societal disruptions in both countries. AI applications are also exacerbating ethical questions about government’s role in protecting individual liberties, and elevating the competition between authoritarian capitalism and liberal democracy. To focus on only one of these dynamics would be to lose sight of the bigger picture: AI will create both immense stress on the U.S.-China relationship as well as opportunities for potential collaboration. The core challenge for U.S. policymakers will be to manage the stresses induced by AI in a way that preserves political space for working together when it serves American interests to do so. Along with other essays in our AI policy series, this piece offers recommendations on how best to do so.
How did we get here?
In March 2016, a Google system powered by an AI algorithm squared off against Lee Sedol, an 18-time world champion in the famously complex game of Go. In front of an audience of more than 280 million mostly Chinese viewers, the Google system triumphed, plunging China into what renowned technologist Kai-Fu Lee described as an “artificial intelligence fever” that “lit a fire under the Chinese technology community that has been burning ever since.”1
A little over one year later, in July 2017, China unveiled its national plan for seizing the spoils of AI. The “New Generation AI Development Plan” set targets and pledged national resources, calling for China to catch up on AI technology and applications by 2020, achieve major breakthroughs by 2025, and become a global leader in AI by 2030. President Xi Jinping reinforced these themes in his 19th Party Congress speech in October 2017 and in a major Politburo study session in late October.
Further stoking unease has been some of China’s official rhetoric, which promotes military-civil fusion of technological development to degrade America’s competitive edge. Such unease has been amplified by China’s ambitious Belt and Road Initiative, a massive global initiative that some in the United States fear will enable Beijing to set global technological standards. Cumulatively, China’s efforts have fed what Dean Garfield, president and CEO of the Information Technology Industry Council, has characterized as a newfound “hysteria” in Washington that America is losing its innovation edge to China.2
Americans are unaccustomed to other countries publicly projecting plans to displace them. As former Secretary of State Condoleezza Rice recently said, “When we see [China]…say, ‘We’re going to do whatever it takes to surpass the United States’… you’re going to get a response from the United States.”3 Thus far, a large part of that response has been attempting to slow China’s progress, including by tightening screening of foreign investments in core technologies, scrutinizing Chinese academic exchanges, applying targeted tariffs to reduce China’s competitiveness in key sectors, increasing prosecutions of Chinese actors involved in economic espionage, and investing greater resources in counter-intelligence operations.
To be clear, it is fair and appropriate for countries to defend their economic crown jewels from foreign exploitation or infringement. Like any other country, the United States has the right to defend itself, and should continue to do so vigorously. But in protecting itself, the United States needs to avoid inflicting self-harm. Arguments for “decoupling” the economic relationship between the United States and China—including by collapsing ICT supply chains—would do just that. America’s leading sources of innovation increasingly are found in its technology sector, which is deeply intertwined with China’s. There are high levels of collaboration between researchers and engineers in both countries, manifesting in growing numbers of jointly authored peer-reviewed academic papers and deep levels of joint investments by U.S. and Chinese venture capital firms into AI-related enterprises in both countries.
As Lorand Laskai and Samm Sacks recently argued, fencing off the U.S. technology sector from China would cede ground to Chinese competitors, slow down new breakthroughs, reduce the competitiveness of American firms, and increase costs for American consumers.4 Determining what key U.S. technologies to protect for national security purposes will require policy precision in order to avoid a brickbat approach that undermines American innovation. Former Secretary of Defense Robert Gates once dubbed this the “small yard, high fence” strategy – selectively protecting key technologies, and doing so aggressively.
Focusing on the big picture: U.S. and China separating from the pack
In the process of protecting itself, the United States must endeavor not to lose sight of the bigger picture. While competition between the United States and China is intensifying, these two powers are increasing the distance between themselves and every other country in the world in terms of economic size, pace of innovation, and overall national power. This separation of the United States and China from the rest of the pack is being fueled largely by both countries’ technology sectors. According to a widely cited study by PricewaterhouseCoopers, the United States and China are set to capture 70 percent of the $15.7 trillion windfall that AI is expected to add to the global economy by 2030.
Both countries are being propelled forward in AI by unique attributes that no other country soon will replicate. These include world-class research expertise, deep capital pools, data abundance, largely supportive policy environments, and highly competitive innovation ecosystems. Of the roughly 4,500 AI-involved companies in the world, about half operate in the United States and one-third operate in China. So, while it is fair to say the United States and China are competing against each other, the larger truth is that both countries also are navigating the frontier of innovation simultaneously.
This is where a purely competitive zero-sum framing does a disservice to both. When every step forward by one is viewed as a setback for the other, there is disincentive to coordinate on shared challenges or be open to learning from the other’s experiences.
Identifying areas of competition and cooperation
One way of overcoming the trend toward all-encompassing competition would be for the United States and China to develop a better shared understanding of where cooperation would be mutually beneficial and where inherent conflicts of interests will need to be managed. This would enable both sides to build cooperation where interests align, which in turn would give both sides greater confidence to deal with issues where they diverge. Below is an illustrative – not exhaustive – set of examples, broken down into four categories: military and security; trade; politics; and society.
Military and security
The military domain presents the greatest risk for miscalculation. It also is where the need is greatest for ongoing, direct, authoritative bilateral communication to develop a better shared understanding of ethical boundaries around AI, particularly given the potential implications for warfighting.
The military domain presents the greatest risk for miscalculation. It also is where the need is greatest for ongoing, direct, authoritative bilateral communication to develop a better shared understanding of ethical boundaries around AI.
The bilateral relationship already faces an acute security dilemma, where actions on one side make the other feel less secure and push it to develop countermeasures. As AI technologies become more integrated into weapons systems and those systems gain autonomous capabilities, this security dilemma could grow more pronounced, causing each side to nationalize innovation streams and limit transparency in order to seek an edge over the other. In other words, an existing security dilemma could quickly morph into an AI nightmare.
The stakes are high. As others have pointed out, the United States and China stand on the cusp of rapid change in the conduct of war, not unlike the employment of cavalry, the advent of the rifled musket, or the merging of fast armor with air support to achieve a blitzkreig.5 Both countries are investing heavily to merge AI-enhanced capabilities and enable machine-based decision processes with minimal human interaction.
In the event of a confrontation between U.S. and Chinese forces (e.g., in the South China Sea), robotics and AI could play a critical role. Rapid escalation is an acute risk, particularly if the pace of technological advancements in capabilities exceeds the development of protocols for maintaining human agency in decision-making loops.
The real possibility of unintended and rapid escalation should provide incentive for both sides to begin developing boundaries around uses of AI in warfighting. The normative development process around previous arms control treaties, including the Chemical Weapons Convention, could offer applicable lessons that the United States and China could draw from.
Intensifying technological competition between the United States and China also risks leading to separate technological spheres, with Europe, North America, South America, and Australia largely adopting American technology and standards, and Asia, Africa, and the Middle East adopting Chinese ones. The ongoing global competition between the United States and China over 5G standards may be an early indication of the battles to come over these boundaries.
Through the introduction of 5G networks, the United States and China will shape the development of next-generation mobile standards, spectrum allocation, and deployments in key markets and regions. Particularly as U.S.-China trade tensions persist, and as both Washington and Beijing seek to lock in overseas 5G markets, there is growing risk of a bifurcated, non-interoperable 5G ecosystem emerging. In such a scenario, one system likely would be led by the United States and supported by technology developed in Silicon Valley, and the other would be led by China and supported by its highly capable digital platform companies.6
When a less politically charged moment in the U.S.-China relationship arrives, leaders in both countries should examine – both individually and collectively – whether or not their interests are best served by hastening the bifurcation of the global technology sector into U.S. and Chinese spheres. In such a scenario, both sides would limit their expansion potential: China’s markets primarily would be in developing countries with limited resources for technology build-out, and U.S. companies would operate mostly in developed markets where competition would be fierce.
AI technologies have the potential to be highly disruptive of political relations between the United States and China. AI technologies could become a vehicle for intensifying ideological rivalry, particularly if one or both sides harness such technologies to interfere in the other’s domestic political affairs.
Russia’s interference into America’s 2016 presidential election raised awareness of U.S. vulnerabilities on this front. As Elaine Kamarck documents, Russia used misinformation and disinformation to suppress voter turnout in targeted geographic districts and among certain demographic groups. Russia also spread derogatory information to denigrate certain candidates, foremost Hillary Clinton. What if those actions were the election interference equivalent of Lindbergh’s test flight – proof of concept, but far from realization of potential?
Going forward, artificial intelligence technologies may enable even more intrusive interference into democratic elections, including by improving an adversary’s ability to target and persuade particular voting groups. In her piece “Malevolent Soft Power, AI, and the Threat to Democracy,” Kamarck envisions a future where polling or search algorithms are linked with artificial intelligence and a human voice to call swing voters and persuade them, in real-time, that a certain candidate will harm them on the issues they identify as important and that the alternative (i.e., preferred) candidate is committed to addressing their individual concerns. She described this as “high-frequency trading in political persuasion.”7 Alina Polyakova classifies tactics like this as examples of “AI-driven asymmetric warfare.” In her piece “Weapons of the Weak: Russia and AI-driven Asymmetric Warfare,” she warns of the perils that ever-improving, low-cost commercial tools present. “Whereas most Russian disinformation content has been static,” she writes, “advances in learning AI will turn disinformation dynamic” through the creation and dissemination of manipulated video and audio.8
Going forward, artificial intelligence technologies may enable even more intrusive interference into democratic elections, including by improving an adversary’s ability to target and persuade particular voting groups.
If external interference becomes more prevalent and the legitimacy of election results around the world increasingly are called into question, democracy’s appeal could dim and alternative models (e.g., China’s economically statist and politically Leninist system) could become more attractive. This could open the door for Beijing to argue in other capitals that its model delivers higher rates of economic growth and that democratic systems are brittle against manipulation and ineffective at equitably distributing benefits within society.
To be clear, no public evidence exists to indicate that China has meddled in U.S. domestic political affairs in a manner reflecting the hypothetical scenario described above. Even absent such meddling, though, there are still significant concerns in Washington about how Beijing is harnessing AI technologies to surveil its citizens and suppress domestic dissent. These concerns reside on two levels. The first is the manner in which Beijing is perfecting its ability to track its citizens’ movements, communications, spending habits, news consumption, and so on. The second concern is that China may seek to export these practices to foreign leaders who desire tighter control over their citizens. In effect, China’s AI-driven model of intrusive surveillance could challenge America’s long-running efforts to spread democratic principles around the world. Even if this occurs more by default than design, China’s exportation of its policies and technologies to other countries could intensify ideological competition between the United States and China.
In short, there is serious risk that AI-powered technologies could inflame political and ideological tensions between the United States and China. But this outcome is not a foregone conclusion. To mitigate the possibility, there needs to be serious, sober, and sustained bilateral engagement to identify boundaries around what constitutes state interference in election processes and political systems. In 2015, for example, President Obama and President Xi agreed that government-sponsored, cyber-enabled economic espionage for commercial gain is out of bounds, and both leaders subsequently attracted other bodies such as the G-20 and the Gulf Cooperation Council to embrace a similar understanding. Reports have surfaced in recent months, however, suggesting that China has resumed government-sponsored, cyber-enabled economic espionage for commercial gain. Given the ill will within the U.S. government following reports that China may not be abiding by the 2015 cyber agreement, discussions on the boundaries of acceptable government involvement in other countries’ political systems may need to begin at the Track II level and mature over time into official channels.
As the world leaders in AI, the United States and China will be among the first to confront intense social disruptions from this new technology. AI could prove to be every bit as revolutionary as the introduction of electricity or the steam engine – innovations that hastened a shift from agrarian to industrial modes of production and accelerated urbanization.
Already, a cottage industry has formed to predict the range of job losses that could result from the adoption of AI and the widening use of robotics. While there is nothing yet approaching a consensus around the scale of dislocation, the low-end projections are sobering, and the high-end estimates are frightening. At the low end, a team of researchers at the Organisation for Economic Cooperation and Development found that 10 percent of jobs in the United States are at high risk of being automated.9 Former Treasury Secretary Larry Summers has predicted that the rise of AI could bring about unemployment for about a third of American men ages 25-54 by mid-century. Kai-Fu Lee predicts that within ten to twenty years, the United States technically will be capable of automating between 40 and 50 percent of jobs.10 A similar story also applies to China.
In the face of such disruptions, both the United States and China will face hard choices, such as how to reform their educational systems, cope with widening wealth inequality, determine whether some form of a universal basic income is needed to preserve social cohesion, reform social safety nets, develop new concepts around privacy, and find productive ways for displaced workers to feel connected to society.
Each country also will contend with how to seize opportunities presented by AI to improve the national condition. A prime example is health care. The U.S. and China are the world’s two largest health care markets, and both countries are projected to experience a surge in spending over the coming decades as their populations age. Both would benefit from jointly leveraging AI applications for image analysis and diagnosis, discovering cures for cancer and other diseases, and identifying the most efficient care models for treatable conditions. Beyond health care, both countries could benefit from sharing data and expertise on major challenges like weather modeling, efficient energy use, tracking the effects of climate change, increasing access to education, enhancing wildlife conservation, and identifying and responding to illegal, unreported, and unregulated fishing. Both countries also could work together on standard-setting for new technologies, which could create greater efficiencies for bringing new products, such as driverless cars, to market.
Steps to improve America’s ability to work with China on AI developments
Given the magnitude of risks and opportunities on the horizon, and the fact that the United States and China simultaneously will be navigating uncharted territory in dealing with the societal dislocations caused by AI, it is imperative that both countries have candid discussions about how to effectively manage AI developments. These discussions should be guided by the objectives of managing risks and seizing opportunities. Many of these conversations likely will begin outside of government channels, a circumstance that reflects the private sector’s key role in AI innovation. Baidu’s recent decision to join the U.S.-based Partnership on AI provides an encouraging example of how leading U.S. and Chinese actors can come together to establish best practices for AI systems.
The following are four recommendations of steps that could be taken to strengthen America’s ability to manage the impact of AI and related technologies within the context of U.S.-China relations:
- Mutual reassurance. The level of senior-level official communication on AI and related technologies lags significantly behind the potential impacts these could have on the bilateral relationship. Leaders in both countries could jointly reaffirm that their objective is (1) to address forthrightly and manage constructively areas where the introduction of AI and related technologies will elevate competition and (2) to encourage collaboration in areas where both sides would benefit from greater coordination and cooperation. Such parallel messaging would create a demand signal for leading thinkers on both sides of the Pacific to prioritize these topics in upcoming exchanges.
- Maintain perspective. Chinese AI researchers and developers are not ten-foot-tall giants. China’s plan to overtake the United States and dominate the future of AI innovation is more ambition than achievement to date. To be sure, Chinese researchers and developers enjoy significant relative advantages, including extensive government support, data abundance, and a hyper-competitive entrepreneurial environment, all of which accelerates the pace of innovation.11 But China’s path to unalloyed dominance is far from assured. China also confronts some serious challenges, including an uncertain data regime and lack of global players in key data sectors; government pressure on technology companies to ensure some advances serve the needs of the Communist Party; a more restrictive global environment for acquiring cutting-edge technologies abroad; the possibility that the government could, in the future, pressure Chinese firms to procure components domestically rather than from globally integrated supply chains; the risk of government-directed investment leading to speculative boom and bust cycles; and the potential for AI to become socially divisive as it is employed domestically for intrusive surveillance or targeted repression. Additionally, while much is made of China’s data abundance, it is worth bearing in mind that Google and Facebook each have more users globally on their platforms than the entire population of China, with further growth potential still. America’s confidence in its ability to compete with China matters greatly. If the United States feels back-footed by the pace of China’s technological advances, it naturally will be less comfortable exchanging lessons learned with China on shared AI-related challenges.
- Invest in strengths. The United States holds three core advantages when it comes to innovation – education, immigration, and investment. With its ability to draw the best minds from around the world, its world-class university system, and its deep and efficient capital pools, the United States should welcome healthy and fair competition with China. The United States needs to advance a proactive strategy to build on these strengths – for example, by easing the immigration process for leading innovators; resurrecting efforts to strengthen the research triangle between academia, government labs, and the private sector; and maintaining policies that attract capital to American shores.
- Find our friends. Through coordination with allies and partners, the United States will be better able to harmonize national export controls, defense trade controls, and investment review mechanisms to limit transfer of dual-use technologies to China. The United States will have greater impact in pressing China to curb actions it finds unacceptable if allies and partners join the chorus. The United States also would benefit from increasing collaboration with allies to accelerate AI-enabled advances in defense innovation.
In the AI age, no other country likely will catch up to the United States or China in terms of technological development or national power, and the United States and China each will not be able to dominate or impose its will on the other, at least in peacetime conditions. Both stand to gain if they find ways to learn from each other’s experiences navigating the information frontier, or to lose if they descend into unvarnished confrontation or conflict.
Instead of viewing AI in zero-sum, Cold War–like terms, the United States and China need to make deliberate efforts to adopt a more balanced narrative.
In order for both sides to manage tensions and maintain open channels for cooperating in areas of overlapping interest, the paradigm of AI’s role in the bilateral relationship will need to change. Instead of viewing AI in zero-sum, Cold War–like terms, the United States and China need to make deliberate efforts to adopt a more balanced narrative. At a time of intensifying rivalry, this shift will not come easily or naturally, but the costs and consequences of the alternative should awaken both sides to the value in doing so.
- Lee, Kai-Fu, AI Superpowers: China, Silicon Valley, and the New World Order (New York: Houghton Mifflin Harcourt, 2018), 3.
- Scola, Nancy, “China Rattles Washington’s Tech Debates,” POLITICO, October 27, 2018, https://www.politico.com/story/2018/10/27/china-looms-over-washingtons-tech-debates-830905.
- Rice, Condoleezza, “China Town Hall with National Committee on U.S.-China Relations,” October 9, 2018, https://www.ncuscr.org/content/video-2018-china-town-hall-secretary-condoleezza-rice.
- Laskai, Lorand, and Samm Sacks, “The Right Way to Protect America’s Innovation Advantage,” Foreign Affairs, October 23, 2018, https://www.foreignaffairs.com/articles/2018-10-23/right-way-protect-americas-innovation-advantage.
- Allen, John and Amir Hussain, “On Hyperwar,” Proceedings 143/7/1373 (July 2017), accessible at https://fortunascorner.com/2017/07/10/on-hyper-war-by-gen-ret-john-allenusmc-amir-hussain.
- Triolo, Paul, Kevin Allison, and Clarise Brown, “Eurasia Group White Paper: The Geopolitics of 5G,” Eurasia Group, November 15, 2018, https://www.eurasiagroup.net/siteFiles/Media/files/1811-14%205G%20special%20report%20public(1).pdf
- Kamarck, Elaine, “Malevolent Soft Power, AI, and the Threat to Democracy,” Brookings, November 29, 2018, https://www.brookings.edu/research/malevolent-soft-power-ai-and-the-threat-to-democracy.
- Polyakova, Alina, “Weapons of the Weak: Russia and AI-driven Asymmetric Warfare,” Brookings, November 15, 2018, https://www.brookings.edu/research/weapons-of-the-weak-russia-and-ai-driven-asymmetric-warfare.
- Nedelkoska, L. and G. Quintini (2018), “Automation, skills use and training,” OECD Social, Employment and Migration Working Papers, No. 202, OECD Publishing, Paris, https://doi.org/10.1787/2e2f4eea-en.
- Lee, 164.
- Lee, Kai-Fu and Paul Triolo, “China’s Artificial Intelligence Revolution: Understanding Beijing’s Structural Advantages,” Eurasia Group, December 2017, https://www.eurasiagroup.net/files/upload/China_Embraces_AI.pdf.