The application of artificial intelligence (AI) technology to military purposes has raised many concerns. In particular, many observers predict a new arms race between states, which might result in profound ethical, legal, and security implications (Maas 2). In his article Artificial Intelligence, International Competition, and the Balance of Power, professor Michael C. Horowitz, discusses the questions around AI that contribute to the emergence of such concerns. Despite not being a weapon itself, AI provides a lucrative opportunity for gaining strategic advantage both for democratic and autocratic regimes. In this regard, AI diffusion has the potential for disrupting the existing arrangements in regional and international security. This disruptive effect will be more severe if one of the first-mover powers, such as China or the United States, fails to adapt military AI while its competitors succeed in doing so. Professor Horowitz provided five key arguments, that make military AI a powerful, yet unpredictable and barely controllable variable for international security.
AI – “Enabling” Technology Rather Than a Weapon
The current state of AI development is associated with considerable uncertainties, especially in terms of potential military applications. To a certain extent, this problem stems from the nature of AI technology. According to Horowitz, AI is much closer to the so-called enablers, or general-purpose technologies with multiple applications (39). In this regard, AI can be compared to electricity or an internal combustion engine rather than a weapon. Therefore, AI should be considered a substantially broader concept than a missile, tank, or submarine (Horowitz 39). Due to this fact, imposing international legal restrictions on AI development is rather challenging. In theory, AI-enabled weapons pose an additional threat due to the prospect of losing human control (Gill 171). However, one should understand that even the narrow military AI will not necessarily be designed for operating weapons.
In general, a military AI may be utilized in three distinctive dimensions, and only one of them is directly related to weapon systems. Firstly, AI can reduce the need to use humans in the operation of tanks, planes, ships, or any other armament types. Secondly, AI can assist in processing and interpreting information, which is highly valuable for military intelligence. Third, AI systems could be used in new forms of command and control on the battlefield (Horowitz 41). Overall, emerging technologies can shape the balance of power through military and economic means and directly influence countries’ ability to win wars. However, it is still unclear how to translate the capabilities of a military AI into actual power (Horowitz 42). Therefore, even the specialized military AI can hardly be considered a weapon, and its true value lies in the potential for enabling subsequent innovations in warfare.
The Universal Appeal of Military AI
Incentives for the development of military AI exist for a wide spectrum of contemporary political regimes. On the one hand, democracies are interested in military AI which can decrease the burden of warfare on the population and reduce the risk to soldiers. On the other hand, autocracies that usually lack trust in people will benefit from outsourcing military decision-making to algorithms (Horowitz 48). Whereas this Horowitz’s argument seems plausible, it can be augmented with two other advantages of military AI — geopolitical and strategic. Both of these potential advantages are lucrative to any country, regardless of its political regime.
From the geopolitical perspective, the emergence of new technology offers countries a chance to close the gap with the major players. Gill provides an example of how China and Israel utilized the enabling technology of the Internet and developed formidable cyber warfare capabilities, which gave them a competitive edge (172). In this regard, any country, whether democratic or autocratic, may use the enabling AI technology to strengthen its military power. Consequently, the balance of power in the world may shift depending on which player finds the best application for the enabling potential of military AI.
Regarding the strategic dimension, military AI offers considerable flexibility in terms of options available for the country. In particular, military AI technologies may provide better access to the so-called asymmetric warfare tools. For instance, it would become possible to disrupt superior adversaries with cyber attacks, as machine learning would likely expand their scope and scale (Johnson 14). Such application of the military AI holds value both for democracies and autocracies limited in resources for conventional war. Therefore, Horowitz’s point about the universal appeal of military AI can be considered valid.
AI Diffusion and Large-Scale Consequences in Balance of Power
The lucrative nature of the military AI application has led to a situation that is increasingly described as a new arms race. Various countries, ranging from the world to regional powers, have made substantial investments in AI technology. For instance, in 2017, China published a national strategy on artificial intelligence that proposed a coordinated program to build a first-mover advantage (Horowitz 45). The USA responded in 2019 when President Trump signed an executive order entitled Maintaining American Leadership in Artificial Intelligence that deemed AI paramount for U.S. national security (Gill 173). Russia had developed remotely piloted tanks; regional powers, such as South Korea, Australia, Canada, Israel, and France also invested in military AI or incorporated AI in their defense strategies (Horowitz 46). Given such evidence, Horowitz claims that the implications of the narrow AI for the global balance of power could be quite significant.
Horowitz’s point of view finds support in other scholarly sources. For instance, Johnson claims that the emerging China-U.S. race to harness the military potential of AI will have potentially destabilizing implications for international security. In particular, Chinese and Russian geopolitical ambitions may cause the USA to abandon the “keep humans in the loop” approach to AI (Johnson 14). As a result, the unpredictable, unchecked arms race between the world powers and their geopolitical satellites will receive a new incentive. Garcia compared military AI diffusion to the nuclear arms race, with a notion that international actors will develop and replicate AI technologies much faster than nuclear weapons (6). Given the current political climate and the complicated nature of AI technology, these concerns will not likely be addressed shortly.
Impact of AI on International Security — The Matter of Diffusion and Use
The eventual impact of AI technology on international security will largely depend on diffusion channels and a particular state’s capacity for military innovations. Horowitz described two scenarios for the diffusion of military-relevant AI. Firstly, AI technologies may diffuse through commercial drivers after the militaries adopt solutions widely available on the market (Horowitz 51). Secondly, AI technologies may be militarily-exclusive, driven by the research applicable only to militaries (Horowitz 52). These scenarios lead to the drastically different spread of AI technologies and, consequently, produce different implications for the global balance of power and international security.
The commercially-driven AI diffusion results in a situation where competitors can relatively quickly acquire or copy others’ advances. As a result, it becomes difficult for a single country to stay ahead in AI technology quality. Consequently, the competition evolves into a multilateral state, where advantage in quantity becomes paramount. This scenario is similar to the Second Industrial Revolution when many countries transferred advancements in steel production, chemistry, and other areas to the military-industrial complex (Horowitz 51). On the contrary, the military-driven diffusion limits the competition to a handful of countries, similar to a space race between the Soviet Union and the USA. This diffusion scenario holds substantial benefits for the first mover since militarily-exclusive research is difficult to copy (Horowitz 53). However, limited information may result in the first-mover’s inability to adapt AI technology, even if the technical base is adequate. In this case, the global balance of power and international security disruption may occur when competitors realize that the first-mover lost their advantage.
In addition, the impact of AI on international security depends on technology owners’ ability to develop military innovation based on enabling technology and use it in practice. Horowitz illustrates this point in the example of the 1940 Battle of France (51). The Germans, the British, and the French had equally advanced equipment. However, the Germans developed and successfully implemented a military innovation of blitzkrieg and overwhelmed an equal adversary.
In this regard, the negative implications of AI development for international security may be mitigated via a comprehensive control system that monitors the use of AI technologies. According to Maas, such a system must consider the military AI technologies beyond automated weapon systems and incorporate ethical, legal, stability, and global safety rationales (19). Unfortunately, the states that achieved considerable progress in AI development will unlikely surrender potential advantages unless the military application of AI appears to be ineffective, unreliable, or financially inefficient. At this point, technology owners consider military AI a promising enabling technology, which may strengthen their positions in international relations rather than threaten global security.
Mixed Strategy — Optimal Choice for Military AI Development
Given the current situation around military AI, many countries disregard negative yet vague implications for international security and pursue tangible benefits of AI development. The perspective of falling behind the competition and losing the influence in the balance of power incentivizes national states to draw the military AI development strategies. Horowitz suggests a mixed approach between market- and state-driven AI development (56). If non-proliferation is not feasible in the future, a state has to find the balance between directed initiative and innovation. The mixed strategy offers the necessary flexibility by combining government-funded basic research and market incentives with the innovative thought of private companies.
Artificial Intelligence, International Competition, and the Balance of Power by M. C. Horowitz explores multiple problems associated with military AI development. Most importantly, controlling advancements in AI is extremely difficult because this technology is an enabler, not a weapon. As a result, any attempt to ban military applications would severely harm the private sector of the world economy. Secondly, both democracies and autocracies can see tangible benefits of military AI, which undermines organized non-proliferation efforts. Lastly, the exact implications of military AI development largely depend on diffusion scenarios and innovative use rather than the technology itself. As such, the search for military AI applications reminds a new arms race, which may potentially pose a significant danger to international security. However, the uncertainties surrounding AI technologies push the states to military AI development and incorporation in national defense strategies. Overall, experts in international relations perceive the possible issues of military AI highlighted by Horowitz, but the sense of benefit loss supersedes national states’ concerns about potential dangers.
Garcia, Denise. “Lethal Artificial Intelligence and Change: The Future of International Peace and Security.” International Studies Review, vol. 20, no. 2, 2018, pp 1-8.
Gill, Amandeep Singh. “Artificial Intelligence and International Security: The Long View.” Ethics & International Affairs, vol. 33, no. 2, 2019, pp. 169-179.
Horowitz, Michael C. “Artificial Intelligence, International Competition, and the Balance of Power.” Texas National Security Review, vol. 1, no. 3, 2018, pp. 38-57.
Johnson, James. “Artificial Intelligence & Future Warfare: Implications for International Security.” Defense & Security Analysis, vol. 35, no. 2, 2019, pp. 1-23.
Maas, Matthijs M. “How Viable is International Arms Control for Military Artificial Intelligence? Three Lessons from Nuclear Weapons.” Contemporary Security Policy, vol. 40, no. 3, 2019, pp. 1-27.