AlphaGo and Beyond: The Chinese Military Looks to Future “Intelligentized” Warfare

AlphaGo’s historic defeat of Lee Sedol in their 2016 match and its latest successes against world champion Go player Ke Jie during last month’s Future of Go Summit in Wuzhen, China have demonstrated the power and potential of artificial intelligence. Although the summit was presented as an opportunity for AlphaGo to “explore the mysteries of Go” with leading Go players, the contest was seen as a ‘battle of man and machine’ and the result a triumph of artificial over human intelligence. Ke Jie was “shocked” and “deeply impressed” by his opponent, including because certain of the AI’s moves “would never happen in a human to human match.”

AlphaGo, designed by a team of researchers at Alphabet Inc.’s Google DeepMind, utilizes deep neural networks trained through both supervised learning from the games of human experts and reinforcement learning from self-played games. While its ‘value networks’ evaluate the board positions, ‘policy networks’ select its moves. In the course of its career in Go, AlphaGo has displayed the capability to formulate unique, creative tactics undiscovered and unanticipated by human players. Although AlphaGo itself represents only “narrow AI” or “weak AI”—tailored to a specific task, rather than capable of generalized intelligence—the techniques used for its development will be applied to new contexts and challenges after its ‘retirement’ from the game. Ultimately, the strategic ramifications of AlphaGo’s successes extend far beyond Go itself.

The definitive defeat of China’s best human Go players by foreign AI could have been seen as awkward or problematic by the Chinese leadership, demonstrating the limitations of not only human intellect but also Chinese technology. (Reportedly, at one point, scientists from the China Computer Go team planned to challenge AlphaGo with their own AI, but apparently did not manage to do so.) Despite extensive coverage of the event in official media, Chinese censorship instructions forbade websites from live streaming the match, emphasizing it “may not be broadcast live in any form and without exception.” This decision could reflect the sensitivity of Google itself, which had withdrawn from the market in 2010 due to concerns over censorship and cyber espionage and remains blocked. Google’s hosting of the summit has been characterized as a charm offensive, in the context of overtures towards reentry. However, its prospects for future market access remain questionable, particularly given China’s intensified drive for indigenous innovation and the success of national champions.

Indeed, the rapidity of recent Chinese advances in AI has indicated its ability to keep pace with—or perhaps even overtake—the U.S. in this critical emerging technology. The dynamism of private sector efforts in China is clearly demonstrated by the successes of major Chinese companies, including Baidu, Alibaba, and Tencent, and even start-ups such as Iflytek, Uisee Technology, or Turing Robot. From speech recognition to self-driving cars, Chinese public and private efforts in AI are cutting-edge. The magnitude of research, as reflected by the number of papers published and cited, has already surpassed that of the U.S. However, for the time being, AlphaGo represents a high-profile demonstration of the sophistication of U.S. AI.

Looking forward, the Chinese leadership aspires to achieve a dominant position in AI, surpassing the U.S. in the process, in order to take advantage of the unique advantages that AI could confer to China’s economic competitiveness and military capabilities. To date, China has released several national science and technology plans involving AI and established a national deep learning lab, headed by Baidu. In particular, China’s new “AI 2.0” mega-project will advance an ambitious agenda for research and development, including economic and national security applications.

At the highest levels, the Chinese People’s Liberation Army (PLA) also recognizes and intends to take advantage of the transformation of today’s informatized (信息化) ways of warfare into future “intelligentized” (智能化) warfare. According to Lieutenant General Liu Guozhi, director of the Central Military Commission’s (CMC) Science and Technology Commission, the world is “on the eve of a new scientific and technological revolution,” and we are “entering the era of intelligentization” due to rapid advances in AI and its impactful military applications. He believes that AI will result in fundamental changes to military units’ programming, operational styles, equipment systems, and models of combat power generation, even leading to a profound military revolution. The PLA might have a unique opportunity to take advantage of these trends through leveraging the dynamism of Chinese advances in AI through a national agenda of military-civil integration (军民融合).

While the lessons learned by the PLA from other nations’ wars have traditionally informed its approach to military modernization, its current thinking on the military implications of AI has, in fact, been influenced not by a war but by a game. AlphaGo’s initial defeat of Lee Sedol appears to have captured the PLA’s imagination at the highest levels, resulting in the convening of high-level seminars and symposiums on the topic. PLA thinkers’ apparent fascination with AlphaGo presents early indications of its initial thinking on and potential future employment of AI in warfare, with applications ranging from autonomous unmanned systems and swarm intelligence to command decision-making. The PLA appears to see AlphaGo’s mastery of the complex tactics associated with the game as an apt demonstration of its future military utility.

From the perspective of PLA strategists, this “great war of man and machine” (人机大战) decisively displayed AI’s potential to take on an integral role in command decision-making in future “intelligentized” warfare. The successes of AlphaGo are considered a turning point that demonstrated the potential of AI to engage in complex analyses and strategizing comparable to that required in warfare—not only equaling human cognitive capabilities but even enabling a distinctive advantage that may surpass the human mind.

Certain PLA thinkers anticipate that the intelligentization of warfare will result in a trend towards a battlefield “singularity,” such that human intelligence may prove unable to keep pace with the new operational tempo of machine-age warfare. At that point, as warfare occurs at machine speed, keeping humans “in the loop” for the employment of weapons systems or even certain aspects of decision-making could become a liability, rather than an asset. Consequently, AI could necessarily take on a greater role in command and control.

The PLA’s response to AlphaGo thus raises critical—and, for the time being, unanswerable—questions about how its future approach to autonomous weapons and other military applications of AI might differ from that of U.S, including on issues of meaningful human control. Indeed, as the U.S. and Chinese militaries compete to advance their capabilities in this critical technological domain, the resulting legal and normative issues—as well as potential adverse implications for strategic stability—will merit further analysis and bilateral engagement.