Deterring Artificial Intelligence
Armed conflict in the Internet age of today compared to the Cold War? Can deterrence be expected to work now, like then?
“The twentieth century… has seen the development of technologies which have allowed military power to be projected over much greater distances than in the past. These technological developments have also led to involvement of whole populations in many of the conflicts which have occurred. As a result, a broader definition of strategy has emerged.”
- John Baylis writing on Nuclear Deterrence, and the concept of ambiguity (Baylis, J., 1995)
Already when defining deterrence in the previous century John Baylis one of the foremost scholars in strategy and arms control was struggling with this concept. Even in one of the latest editions of strategic studies (Baylis, 2016) there is still a struggle to define what strategy is supposed to be, yet we can agree that it is changing — and still increasingly ambiguous in the 21st century. One important aspect of strategy has been deterrence. One recent grand strategy relates to AI: “Early this year, the US announced a grand strategy for harnessing artificial intelligence in many areas of the military, including intelligence analysis, decision-making, vehicle autonomy, logistics, and weaponry. The Department of Defense’s proposed $718 billion budget for 2020 allocates $927 million for AI and machine learning.” (MIT Technology Review, 2019)
“Deterrence is a strategy intended to dissuade an adversary from taking an action not yet started by means of threat of reprisal, or to prevent them from doing something that another state desires. The strategy is based on the psychological concept of the same name.”
Deterrence has been the subject of research for those studying international security policy for 200 years, some claiming it started with Clausewitz. This can be split into different types of deterrence theory, yet most depend on taking actions. Especially in the recent century deterrence has been focused on nuclear war, with first a clear unipolarity held by the United States with its nuclear capacities. Then later a growing bipolarity with Russia against the US, and increasingly smaller states and other actors getting increased nuclear capacities. These have led into the pursuit of nuclear triads (land, air, sea) which several countries have achieved recently China has increased its military spending.
Graham Allison did as early as the 60s argue that states had different reasons to deter, and that there was a stigma surrounding the use of nuclear weapons (Allison, 1969). These constructivist theories are becoming increasingly relevant as there has been increasing use of violence, but little direct use of nuclear capacities in any type of warfare, as of yet. Therefore it is important to consider deterrence in relation to other type of activities within warfare and the use of force.
There is a strange relationship with being there in war and being away as countries seek to remove themselves from the battlefield while still maintaining a vested involvement. One such case example would be that of Afghanistan with training of soldier, but another one could be the drone strikes in Iraq. There is an increasing focus on proxy warfare and new questions arising in terms of strategy relating to proxy warfare. Beyond this focus countries are escalating their capacities in other areas such as cyber warfare, indeed this combines with existing worries.
“For realists, the state is the main actor and sovereignty is its distinguishing trait. The meaning of the sovereign state is inextricably bound up with the use of force.” Baylis,
“Foreign policy has often been compared to moves, sequences of moves, and games of chess. If one were limited to observations on a screen upon which moves in the chess game were projected without information as to how the pieces came to be moved…” (Allison, 1969)
What if you could disable the capacity of an opponent or ensure that a nuclear weapon was being used against another state etc. One such example of delay can be said to be Stuxnet, a hack done presumably by the United States to delay the development of nuclear capacities. Stuxnet is a malicious computer worm, first uncovered in 2010, thought to have been in development since at least 2005. Stuxnet targets SCADA systems and is believed to be responsible for causing substantial damage to Iran’s nuclear program. This type of capacities has been said to be the entry into a new era of warfare by some. Beyond this there is currently mentioned to be a race towards increased capabilities within artificial intelligence.
As such if we combine complexity with nuclear weapons proliferating or spreading; the advent of cyber warfare; and the race to computing capacities and deployment we are looking at a sketchy picture for deterrence. Not to say that deterrence is possible, but with a bipolar world with China and the US how will conflict look like? Chaos necessitates an increased focused on disciplines working together — computer scientists and a broad range of scholars in the field of social science.
Use of force is still physical because the Internet is a series of physical properties with consequences in society. Deterrence is therefore highly relevant in this regard to prevent the use of force by state actors or non-state actors. With sovereign states online anarchy is redefined, but the battles are now distributed amongst servers with traditional or conventional warfare and force in this regard still very much present.
Allison, Graham T. (1969):”Conceptual Models and the Cuban Missile Crisis”, American Political Science Review. №3. (p. 689–718)
Baylis, J., Wirtz, James J. and Colin S. Gray (eds.) (2016). Strategy in the Contemporary World: An Introduction to Strategic Studies. Oxford: Oxford University Press. Part I and part II (303p)
Baylis, J. (1995). Ambiguity and Deterrence: British Nuclear Strategy, 1945–1964 (№4). Oxford University Press.
This is #500daysofAI and you are reading article 143. I write one new article about or related to artificial intelligence every day for 500 days.
What is your favourite solution or way to adress different problems assisted by the field of AI? Please let me know by responding to this text.