The accomplishment is yet another sign of China's overall AI ambitions.
Analysis: China’s nationwide pursuit to become the world leader in artificial intelligence (AI) is an attempt to not only match U.S. economic power, but to bypass it geo-strategically.
While Beijing’s involvement is spurred by economic ambitions, it has made it clear that the development of AI will simultaneously be for military applications that could change the character of warfare and place the U.S. at a geopolitical disadvantage.
Background: China has spurred its innovation engines into action, seeking to leapfrog U.S. military and technological supremacy through advanced AI and machine learning. Their unique brand of capitalism and government control has enabled bottom-up innovation that is broadly guided by the hand of the Chinese Community Party. China’s whole-of-nation approach means the U.S. has found itself in a race against a strategic competitor.
Ambassador Joseph DeTrani, former Director for East Asia Operations, CIA
“The future of AI, for commercial and security application, appears unlimited; China will pursue this new technology with speed and significant state and private resources.”
- China’s national push toward development in AI can be seen just in the sheer number of recent Chinese academic publications on the subject. Since 2014, China has surpassed the U.S. in number of published papers on deep learning and continued to increase its output by nearly 20 percent in 2016. While increases in the quantity of AI publications do not necessarily correspond with advances in quality, it is clear that China is seeking to advance its AI development significantly.
- In March 2017, China established its National Engineering Laboratory of Deep Learning Technology under the leadership of Baidu. The purpose of the new research center involved exploring image and voice recognition, biometric identification, and new forms of human-machine interaction.
- In July 2017, Beijing’s State Council released the New Generation AI Development Plan, laying out the nation’s strategy to lead the world in AI development by 2030. Broadly modeled off of the Obama administration's 2016 push for a revolution in artificial intelligence, Beijing’s initiative includes building out indigenous capacity to create a $150 billion Chinese AI industry – led by tech giants Baidu, Alibaba and Tencent – as well as through foreign mergers and acquisitions, equity investments, venture capital and the establishment of research and development centers abroad.
- There are a number of societal and organization characteristics that China possesses that might put it at an advantage in the development of advanced AI. The first is China’s particular brand of socialist market economy affords the government a significant amount of control and involvement in market forces. Not only does this mean Beijing can push for over-the-horizon innovation where market forces would typically fail, but it also has created a unique level of cooperation between Chinese tech companies and the government. This gives Beijing significant economic leverage to expand its political clout around the world, including through its One Belt, One Road initiative.
- Perhaps most notably is the Chinese government’s nearly complete access to consumer data – the lifeblood of machine learning and artificial intelligence – with little to no privacy protections. The Chinese startup Yitu Tech, which maintains a close relationship with state security, shares access to the biometric data of 1.8 billion Chinese that it feeds into its facial recognition software. But while the Chinese government continues to maintain complete access to the data of its citizens – and shares it with industry partners – there does seem to be a recent effort to create data privacy protections among China's industry similar to the EU’S General Data Protection Regulation (GDPR), which could have long-term implications for Chinese AI development.
Ambassador Joseph DeTrani, former Director for East Asia Operations, CIA
“China is dedicating significant resources to research and development on all types of artificial intelligence, with commercial, security and military application. China's recent five-year plan reportedly committed well-over one hundred billion dollars to AI, indicative of the leadership's keen interest in eventually becoming the world's leader in AI. As China moves forward with its One Belt One Road Initiative and the Asia Infrastructure Investment Bank, projects that reach out to more than eighty countries, AI will become an integral part of these international infrastructure projects. It will permit China to apply and further develop their AI capabilities, eventually securing their leadership role with this new multifunctional technology.”
William Carter, Deputy Director, Technology Policy Program, CSIS
“From China’s perspective, AI will be like mobile and desktop computing before that. It will be an economic revolution that creates and entire new generation of digital capabilities in physical systems that they can sell and embed around the world. So they see the development of AI as an opportunity to develop a presence and set the baseline for how other countries around the world, particularly in the developing world, interact with technology and data.”
Issue: China’s pursuit of AI extends to the People’s Liberation Army (PLA) through a national strategy of military-civil fusion. This militarization of AI has far-reaching implications for how China will hold political sway abroad and conduct itself militarily – a strategy referred to as “intelligentized” warfare, according to a November 2017 report by the Center for New American Security. Broadly speaking, advancements in Chinese AI have the capacity to support military command and control, intelligence deduction, advance combat training and military readiness, tailor and scale cyber and information operations against opponents and create counterintelligence vulnerabilities.
Elsa Kania, Adjunct Fellow, Technology and National Security Program, CNAS
“Notably, China’s New Generation AI Development Plan explicitly highlights an approach of military-civil fusion to ensure that advances in AI can be readily leveraged for national defense. To actualize this objective, China will continue to establish and normalize mechanisms for coordination and collaboration among scientific research institutes, universities, enterprises and military industry units, while seeking to ensure that military and civilian innovation resources will be ‘constructed together and shared.’ This strategy is advanced through CCP’s Military-Civil Fusion Development Commission, established in early 2017 under the leadership of Xi Jinping himself. Consequently, the boundaries between military and civilian advances will remain highly blurred in AI.”
Ambassador Joseph DeTrani, former Director for East Asia Operations, CIA
“Given China's ambitious military modernization program, and its focus on the space and cyber domains, it's fair to assume that AI will be incorporated into China's C4ISR – Command, Control, Communications, Computers, Intelligence, Surveillance and Reconnaissance. Its immediate application to China's activities in the South China Sea and pursuit of an anti-access/area denial is obvious.”
- Much like the Pentagon’s Third Offset strategy, Beijing is seeking to embed increasingly sophisticated AI into robotics for autonomous operating guidance and control systems. Automation is already being incorporated in the China’s unmanned aerial, ground and maritime vehicles. Manned-unmanned teaming operations, such as those involving the Caihong-5 (CH-5) aerial drone or the D3000 stealthy maritime combat drone, could be particularly useful in controlling the airspace and waters beyond the Chinese mainland.
- Should Chinese drones be linked through neural networks to create swarms as part of China’s anti-access area denial (A2/AD) strategy in the South China Sea, Beijing could hinder U.S. freedom of navigation in the region. In December 2017, at the Guangzhau Airshow, the Chinese demonstrated the largest swarm to date, flying over 1,000 small drones in formation. This was following a display of 119 fixed-wing drones in June.
- Aside from pursing autonomy within military robotics, China seeks to leverage AI as an enabler of enhanced decision-making and intelligence analysis. Advances in the automation for aggregating different sensors and processing that information – such as satellite imagery – shorten the decision-making time for strategic advantage, perhaps even predicting military maneuvers of opponents in advance. China already plans on equipping its nuclear submarines with AI capable of filtering through large quantities of data to support naval commanders’ decision-making.
- There is a discussion however that with the advent of advanced AI and machine learning, the tempo of decision-making could surpass human cognitive ability; meaning humans could eventually be removed from the decision loop over military command and control.
- The more data fed into AI systems to assist in decision-making, the more likely these systems will also be useful in virtual wargaming. Given the PLA’s lack of tangible combat experience, such data-informed wargaming could help better train the Chinese military for a confrontation with a peer opponent such as the United States.
- Perhaps the most likely medium where there is strong utility for advances in AI is in psychological warfare and cyber operations. Automation allows scale, while machine learning facilitates tailored messaging and attacks. China could leverage AI to profile targets through their social media and customize attacks to shape and amplify narratives. Chinese companies such as iFlyTek are already capable of spoofing video and audio recordings, a potentially disruptive tool for psychological operations. Chinese hackers equipped with advanced AI could similarly customize attacks or overwhelm network defenders with hoards of autonomous hacking bots engaged in highly tailored spear-phishing campaigns at scale.
- China’s development of AI to better enable its internal surveillance and censorship regime also has implications for counterintelligence. The ubiquity of closed-circuit cameras and invasive monitoring of online communications in China – along with the advent of biometric data-crunching AI – mean that maintaining secrecy of U.S. intelligence operatives in the country will become more and more difficult. China will also be able to filter through U.S. data – such as the over 20 million profiles of U.S. federal employees stolen from the Office of Personnel Management (OPM) discovered in June 2015 – and create a profile of potential targets for their own intelligence collection.
Ambassador Joseph DeTrani, former Director for East Asia Operations, CIA
“Currently, China's success in the utilization of AI for internal security purposes (like facial recognition), as a complement to the work they're doing on bio-metrics, are tools available to the Ministry of Public Security for monitoring and surveillance purposes.”
William Carter, Deputy Director, Technology Policy Program, CSIS
“Where I am really concerned, and where I think they are making a lot of progress, is in counterintelligence. The Chinese are collecting a huge amount of data on their own people. They are collecting as much data as they can on foreigners. They collect the fingerprints and huge amount of data on every individual that enters China and they are feeding all of this into databases that are supported by machine learning and artificial intelligence to identify potential security threats to the Chinese government – potential foreign operatives who are actually coming to China to recruit people to try to extract classified information. They are also using AI to identify Chinese who could be coopted by foreign governments and foreign intelligence services. That is a huge issue for us.”
Elsa Kania, Adjunct Fellow, Technology and National Security Program, CNAS
“Consistent with its asymmetric approach to military modernization, the PLA could leverage AI to target perceived weaknesses in U.S. ways of warfare. For instance, the PLA has concentrated on advancing integrated network-electronic warfare to target U.S. battle networks, and the capability to leverage AI, whether in enabling cognitive electronic warfare or autonomous cyber operations, could further enhance these capabilities. The PLA recognizes the potential advantages of swarms to saturate the defenses of high-value U.S. weapons platforms, such as fighter jets or aircraft carriers, even depicting such a scenario in its Military Museum in Beijing. To offset current U.S. dominance in the undersea domain, the PLA is also developing autonomous underwater vehicles and reportedly planning to introduce ‘AI-augmented brainpower’ into its nuclear submarines to achieve an advantage. At present, the PLA’s capabilities for information support remain a limitation on its capacity to project power. However, the introduction of AI to enhance intelligence, surveillance and reconnaissance (ISR) could increase the efficiency of these capabilities. In addition, the PLA appears to be incorporating AI technologies into its next-generation missiles and likely also missile defenses, seeking to enhance their precision and lethality. It remains to be seen whether the PLA may incorporate AI in support of its nuclear systems, which would raise questions about the potential impact on nuclear and strategic stability.”
Response: Since 2014, the U.S. military has already begun its Third Offset strategy by seeking to work more closely with U.S. tech giants in Silicon Valley and elsewhere, and to quickly acquire cutting-edge technology for military and intelligence purposes. And while China may have certain societal and organization advantages to enable their quick development of AI, it is possible these could leave them open to countermeasures in the long-term.
- One of the potential organizational inhibitions of the U.S. incorporating advanced AI into its own military and intelligence systems surround requirements of justifying actions – particularly lethal action – within a democracy. For most AI and machine learning systems, how they come to the conclusions that they do remains a “black box” and therefore developing “explainable” AI so that decision-making can be trusted is important. China is not necessarily held back by the same constraints to justify their actions externally.
- In the long-term, however, this could open China up to countermeasures against their AI such as data manipulation or corruption or the modification of its protocols. Without the societal incentive to peer into AI decision-making and examine the path it took to its conclusions, it would be difficult to detect when the U.S. might be employing these countermeasures that could lead Chinese military commanders and systems astray.
Elsa Kania, Adjunct Fellow, Technology and National Security Program, CNAS
“The U.S. military must recognize the PLA’s emergence as a true peer competitor and reevaluate the nature of U.S.-China military and technological competition. As China seeks to become a scientific superpower to rival the U.S., this race to innovate is emerging as a new frontier of strategic competition. In recent history, the U.S. has possessed clear, often uncontested military-technical advantage, but it may not be feasible to achieve a similar edge in AI, given China's rise and the rapid diffusion of these technologies. Consequently, U.S. military advantage might be best assured through leveraging perhaps more enduring advances in the human and organizational dimensions of innovation in which the Chinese military may struggle.”
Anticipation: The strategic race for AI dominance between the U.S. and China is only beginning and it remains uncertain who will lead the world in this new technology. U.S. tech giants such as Apple, Google, Amazon, Facebook, and Microsoft have remained at the forefront AI development, despite Chinese advancements, and will continue to innovate using their unique multinational positions. But China will look expand its industry to facilitate AI innovation along its One Belt, One Road trade initiative.
William Carter, Deputy Director, Technology Policy Program, CSIS
“If you are talking about access to data, U.S. companies are in a better position than China’s because of their global footprint and also as a result of that global footprint, they have information on a wider and more diverse array of people. And ultimately it is studying the similiarities and differences across heterogenous groups in a large dataset where you can get the most value for AI. Chinese companies have data on Chinese people. They do not have a large international footprint, they do not have a lot of foreign customers. U.S. companies have massive datasets on billions of users around the world. That is a strength.”
Elsa Kania, Adjunct Fellow, Technology and National Security Program, CNAS
“As AI catalyzes a fourth industrial revolution, China intends to lead it, leveraging AI to enhance its economic dynamism and military capabilities alike. As China builds a vibrant digital economy, the commercial applications of AI and big data could transform the Chinese economy and society, from healthcare to self-driving cars. Pursuant to the One Belt, One Road strategy, there is a new focus on the digital Silk Road, through which China will seek to leverage the advantages of sharing big data and enhancing digital connectivity, while advancing international scientific cooperation in AI. As an AI power and first mover, China also intends to lead in the formulation of technical standards and mechanisms for global governance of AI, perhaps reinforcing its own interests and advantages in the process.”
This Cipher Brief Analysis by Levi Maxey was first published in February, 2018.
