“[ChatGPT’s] role and impact on the military domain cannot be ignored.”
Chinese regulators have acted swiftly since ChatGPT’s launch in November 2022 to manage perceived political risks by laying out draft rules on the use of generative artificial intelligence (AI).[i] However, this has not stopped the Chinese military’s PLA Daily from inviting security scholars to consider the potential applications of generative AI in the military.
In the three excerpted PLA Daily pieces, Hu Xiaofeng, Zhao Jingxuan, and Hu Yushan explore the risks and opportunities associated with integrating AI technologies and large language models (LLMs) in military planning, operational, and combat systems. Hu Xiaofeng, a professor at China’s National Defense University, argues that the integration of ChatGPT with future information-based and intelligent warfare systems would bring a “qualitative leap to the commander’s decision-making ability.”[ii] Hu Xiaofeng and Hu Yushan argue ChatGPT will likely trigger an expansion of China’s intelligence collection capabilities due to its ability to summarize large swathes of scientific and academic information, providing commanders with near real-time information perception and allowing them to make an accurate threat assessment. Hu Xiaofeng and Hu Yushan also discuss the integration of ChatGPT with other military systems to complete more complex tasks, including training scenario preparation, combat plan generation, combat plan arrangement, action plan drafting, exercise result commentary, as well as higher-level planning tasks like war plan analysis, supply chain risk analysis, and crisis response plan evaluation. At an operational level, Hu Yushan identifies ChatGPT as a useful tool to produce fake news, fake emails, and imitate human language for information deception in cognitive domain operations and cyberattacks. PLA researchers are also wary of the potential negative implications of overreliance on AI technologies in military decision-making. Zhao Jingxuan draws upon the Roman mythology of Janus. Janus is the two-faced god of gates, transitions, doorways, and duality, with the front facing the future and the back facing the past. Zhao likens the use of AI technologies in military decision-making as a doorway presenting serious security, legal, and ethical dilemmas, echoing sentiments expressed at the call to action set forth in February 2023 at the first summit on Responsible AI in the Military Domain (REAIM).[iii]
Hu Xiaofeng (胡晓峰), “ChatGPT, 我们该怎么看” (How Should We View ChatGPT),” PLA Daily (official newspaper of the Chinese People’s Liberation Army), 21 March 2023. http://www.81.cn/szb_223187/szbxq/index.html?paperName=jfjb&paperDate=2023-03-21&paperNumber=07&articleid=901476&utm_source=substack&utm_medium=email.
“In future information and intelligent warfare, ChatGPT can be used for basic data analysis and decision support, natural language processing, and the processing of massive volumes of battlefield information to facilitate a qualitative leap in commanders’ decision-making abilities. On this basis, it may lead to transformational mutations in some traditional operational positions, thus triggering changes in established military systems. If it is professionally trained to work with other systems, it can be used for more complex work such as training idea preparation, combat program generation, combat plan arrangement, operational planning and preparation, exercise result evaluation, etc. This will induce many adjustments to command and decision-making institutions, it may even reshape the command and decision-making process. If it continues to iterate and mutate in the future, it may also be involved in completing other higher-order work involved in war plan analysis and crisis management program evaluation, among others. ChatGPT technology may also be used to produce fake news, fake emails, or even imitate human language to implement information decoys, or be used in cyber-attacks.
Zhao Jingxuan (赵静轩), “雅努斯的两副面孔 (The Two Faces of Janus),” PLA Daily (Official newspaper of the Chinese People’s Liberation Army), 21 March 2023. http://www.81.cn/szb_223187/szbxq/index.html?paperName=jfjb&paperDate=2023-03-21&paperNumber=07&articleid=901476&utm_source=substack&utm_medium=email
When a large number of intelligent combat systems are applied to the battlefield, the cost of war will be greatly reduced, and the ‘zero casualties’ of combatants is expected to become a reality. However, in a complex battlefield environment, the intelligent combat system is very likely to have problems such as indiscriminate killing of innocent people due to identification errors, which will bring infinite hidden dangers to human beings.
Hu Yushan (胡玉山), “作战+ChatGPT，会撞出来什么样火花 (Combat + ChatGPT, What Kind of Sparks Will be Struck )” PLA Daily (Official newspaper of the Chinese People’s Liberation Army), 21 March 2023. http://www.81.cn/szb_223187/szbxq/index.html?paperName=jfjb&paperDate=2023-03-21&paperNumber=07&articleid=901477&utm_source=substack&utm_medium=email. Based on ChatGPT’s powerful analysis and judgement capability, it can act as a super scheduler and quickly solve issues arising in the logistics supply chain with the support of Internet of Things (IoT) devices, cloud computing, and big data. In the procurement link, through analysis of past material procurement data, it can automatically predict future material demand, and at the same time, according to market fluctuations, independently set funding budgets to improve the military’s economic efficiency. In warehousing processing, it can analyze a series of data such as the number of materials in storage, reserve time, and maintenance to carry out intelligence allocation between personnel and people and materials and equipment. In the transportation link, it can determine the best means of delivery by intelligently analyzing demand, resources, and means of delivery to optimize transportation plans.
[i] The Cyberspace Administration of China (网信办) released draft measures to lay out the ground rules that generative AI services must follow, including the type of content these products are allowed to generate. Importantly, Article 4 stipulates that all AI generated content must reflect the core values of socialism and should not subvert state power. For the full Chinese text, see: “《生成式人工智能服务管理办法（征求意见稿)” (Generative AI Service Management (Draft for Comments)),” Cyberspace Administration of China, 11 April 2023. http://www.cac.gov.cn/2023-04/11/c_1682854275475410.htm
[ii] Hu Xiaofeng explicitly mentions integration with AlphaGo (阿尔法狗). AlphaGo is a computer program that was designed to play the strategy board game Go. AlphaGo uses a combination of deep neural networks and tree search techniques and was trained using a combination of supervised and reinforcement learning. In 2016, AlphaGo made headlines when it was the first program to publicly defeat a professional Go player. PLA researchers have discussed military applications of AlphaGO for over half a decade, but Hu admits it has made little progress in effectively applying the program to PLA decision-making processes.
[iii] For more on REAIM and the call to action that 60 countries including the United States and China signed, see: “REAIM 2023 Program,” Government of the Netherlands, 20 April 2023. https://www.reaim2023.org. While China has yet to publish its own vision for governance of AI in military systems, the United States’ “Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy” can be seen here: “Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy,” U.S. Department of State, 16 February 2023. https://www.state.gov/political-declaration-on-responsible-military-use-of-artificial-intelligence-and-autonomy/
Image: ChatGPT Logo.
Attribution: Public Domain