The AI Cyclone and Warfare

Sub Title : AI and its impact on the nature and conduct of war

Issues Details : Vol 13 Issue 6 Jan/Feb 2020

Author : Lt Gen Sanjiv Langer PVSM, AVSM, (Retd)

Page No. : 26

Category : Military Technology

: January 28, 2020

From perception management to operational outcomes AI is influencing the way Nations and peoples have gone to war. The article addresses issues related to the significant convergences between future war and the exponential profusion of  AI and  the impact of the phenomenon on the nature and conduct of war

It is an elementary fact that warfare has been around as long as organised society. AI, on the other hand commencing from the imaginative  conception of computer scientists, has till a couple of decades ago developed, silently. This unsung branch of computer science has however swiftly pervaded significant aspects of individual and organised life. While war and conflict are in the public eye, AI has worked assiduously and stealthily.

The central issue is that are they, significant convergences, between, future war and the exponential profusion of AI. How do these interact and impact on the nature and conduct of war? These unfortunately are the easier set of questions. Prediction of the nature and manifestation of future wars is a perilous pursuit, seldom successful. How AI will churn, twist and manipulate the Battle Space is largely speculative. None of this should deter us as professionals, since we have to meet the speculative and indistinct zones of strategy head on. A view on war as we proceed would be helpful.

At the outset it would be appropriate to talk of war and conflict in the same breath. This is especially so since we have seen a profusion of conflicts that defy all definitions of war as we knew it. Not being classically termed wars, we tend to be occluded, from their huge destructive and disruptive effects that linger tenaciously.

As an exemplar, the conflict against ISIS better termed Daish, waged in the territories of Iraq and Syria, is educative. The conflict was fought by three coalitions against the State claimed by Daish. US led, Russian led and one of 34 Islamic countries. Central to the struggle were Iran, Iraq and Syria. This multifront war, has resulted in approximately 6 million displaced personnel, 5 million refugees and close to a million casualties out of a base population of 22 million. This in no way is a comprehensive view on suffering, since data of those executed, tortured, raped, gassed etc are inadequate. A largely undeclared war with few clearly defined adversaries, involving some of the most powerful nations of the world. Fought against a non-state actor, and in large part against the populace of the Nation itself, along indeterminate frontiers. This has provided an extremely creative manner of waging war with little interference and great impunity, irrespective of the costs.

We are unfortunately tuned to look at warfare in generations and were comfortable in that knowledge. Arguably the last conflict that could be typecast was the Gulf War 2, of March 2003. It was a fine example of 21st century war, and its DNA could be traced to evolution from the earlier generations of warfare. In an overlap but distinctly different genre was the war in Afghanistan,  the Global War on Terror (GWOT) bridgehead. Commencing in 2001, this war took a completely different trajectory. It was an intelligence fired war, with far reaching changes in the US intelligence community, and critical Departments of State and Defence. The hunt for Al Qaida harnessed the available AI and provided the necessary catalyst for powerful converged AI. Let’s pick up the threads on AI.

In 1956 John McCarthy of MIT stated, that, “Artificial Intelligence is the branch of computer science concerned with making computers behave like humans”. Encyclopaedia Britannica states, “Artificial Intelligence (AI ), the ability of a digital computer or computer controlled robot to perform tasks commonly associated with intelligent beings”. This branch of science has over the decades set out to achieve exactly that. Some of the astounding examples of AI will serve to state the case far move eloquently. A view of these will also reveal the thinking, where it is perceived that computers will have human intelligence by 2029, and by 2045 machine intelligence will be infinitely more powerful than human.

Most of us would remember the Chess games between Gary Kasparov and the computer Deep Blue. The world was stunned by Deep Blue defeating Kasparov. Thereafter human, computer collaboration in Chess flourished, till a stage when the humans were of little value. On 07 December 2017, the computer chess champion Stockfish 8 (champion of 2016), was defeated by Google’s Alpha Zero programme. Stockfish 8 had decades of computer experience and access to centuries of human chess experience. It could calculate70 million chess positions per second. Alpha Zero in contrast could do only 80,000 calculations per second and had been taught no Chess strategies or even opening moves. It used the latest machine learning techniques to self-learn, by playing against itself. It took Alpha Zero 4 hours to prepare for the match and of the 100 games played it won 28 and tied 72 not losing even one! Alpha Zero went from ignorance to mastery in 4 hours.

A more recent illustrative example is in the realm of multiplayer poker. Carnegie Mellon university developed Pluribus to be fielded for the game. In the 6 player  competitions, Of the 5000 hands played Pluribus won all. Next Pluribus played 12 professionals, 5 at a time and won all the 10,000 hands played over several days. Pluribus displayed ability to use mixed strategies, unpredictability, and of course exceptional computational capabilities. It is no surprise that this galloping trend in AI has been exploited for surveillance and intelligence.

The use of facial recognition AI, in the German railway stations, by Welsh police for soccer matches, and the Chinese for their population surveillance programmes are some examples. Even more compelling is the US Total Intelligence Awareness programme, launched in 2003 later named Terrorist Awareness Programme, which was fired by advanced AI. Its core was true precognition, since it chooses to identify and predict crime/terrorist acts and apprehend the individual(s) prior to the event. Each of these initiatives have raised significant questions on morality, legality, discretion, and invasion of privacy. None of these concerns have however in any way effected the meteoric trajectory of the rise and spread of AI. Artificial Intelligence a domain that was perceived as esoteric, today pervades everyday life. Occluded from view, it fires diverse aspects from travel to medicine and imposes its will on our choices.

Traditionally we have sought comfort in the fact, that AI cannot control Humans since it is a child of human intellect and cannot exercise human thought and discretion. Probably this view is oblivious of the meteoric and focussed developments in AI, that seek to place it exactly in that domain.  Consider the following. Facebook’s DeepFace facial recognition algorithm achieved an accuracy rate of 97 % challenging human capability. Microsoft speech recognition AI has an error rate of 0.4 % compared to the human rate of 5.9. North-western AI system beat 75 % Americans at a visual intelligence test. Alibaba’s AI has outscored humans in a Stanford University reading comprehension test.

Undoubtedly the enlarged envelope of conflict and war is also a great arena for this intelligence. As professionals a lot we are aware of but when AI sets about to alter war as we know it, we need to sit back and reflect. In the domain of conflict AI has an enviable spread from space to the deep seas. Computer computational abilities, accuracies, and enormous task handling capacities are highly attractive for military applications. This approach has led to the development of Autonomous Weapon Systems, robots etc. Undoubtedly a large spectrum of which  is desirable. The discussion now however needs to transcend from the world of Drones, Autonomous weapons, to the nature and genre of new wars and their intersection with AI.

AI has demonstrated that it has a commanding position in the wide space of C4ISR. If conflict has to be viewed as the outcome of Intelligence and operational decision making, AI will prevail. In 2003 the US launched TIA, Total Information Awareness, later called Terrorist Information Awareness, which was a mass surveillance programme. The programme was designed for predictive policing, based on precognition, and aimed at prevention of terrorist actions and crime. A very close match to the Hollywood Film, The Minority report. It involved the merger of 12 extant programmes, integrated the work of 9 Government entities, 11 corporations and 10 universities. Facilitator and execution platforms were AI. The GWOT fired, demand for a National Counter Terrorism Centre, was met in 2004. This centre was set up in close supervision by Walt Disney Imagineering engineers, who also produced ARGUS, an AI based monitoring programme. It has the capability of monitoring one million web pages in 28 languages over the world. Both examples illuminate the enormous role of AI in making possible capabilities earlier not achievable. We need to focus on the impact on decision making and outcomes. ARGUS has undoubtedly contributed to elimination of several key operatives in Al Qaida, Taliban and Daish. It together with a family of Intelligence based AI have a direct bearing on operational outcomes.

Commencing with the GWOT, and transiting through the Syrian conflict, a new paradigm in conflict has come to be. Public opinion and informed decision making at all levels is deeply influenced by perceptions, largely based on equivocation. To paraphrase Aeschylus “in war, truth is the first casualty”, in media today truth is a shifting goalpost. Populace and citizenry are led to thought and decision, imperfectly. It is no small wonder that AI algorithms are a dominant part of media. General staff operations and intelligence, love data, scenarios, evaluation of factors, probabilities and possibilities and assessed outcomes. The huge seductive capabilities of AI are inexorably attractive. The use of these however causes overloads that reduce time for human dispassionate evaluation. Further imposed windows of response, reduce time for decision making to a fraction of what they were, impacting discretion. In the race to shorten the OODA loop, which AI has made possible, discrimination and discretion are challenged.

The type of conflict extant in Syria, Afghanistan and Yemen, have demonstrated the unique power of no contact or minimum contact war. The complexity and range of the conflict envelope, grants impunity to the dominant powers. There is hardly any voice of the populace and citizenry directly affected. Unlike the earlier generations of war where soldiers bore the brunt of casualties, it is citizenry that bears the larger agony of war. Politico Military guidance and war aims constantly change, conditioned more by the cascade of violence, than by the expression of National will through the Armed Forces.

From perception management to operational outcomes AI is influencing outcomes that alter the way Nations and peoples have gone to war. The role of media and several genres of media is vaguely known but seldom analysed. In the context of a state going to war, war was the ultimate expression of sovereignty. But what off the new paradigm of Half-Wars, or Non-Wars or better still events that involve minimum force, with far reaching outcomes. Killings of individuals, strike on oil facilities, strike on nuclear facilities etc.

This is a preferred playground for AI. Since AI, speed and computational capabilities cannot be humanly matched, once allowed into the Intelligence and Operational space they will never be equalled. The development in AI is, to be lighter, quicker, and smarter. Also, it is taught to learn on the job. It has already displayed the ability to use complex mixed strategies at great speeds. Human inadequacy, laziness, and incompetence will be constantly pitted against efficient AI. The tendency to follow AI recommendations and options may become the norm. This tendency is furthered by the fact that in many cases AI would be correct. Unfortunately, in war and conflict each decision has to be correct.  Further, intuition, feel, reading of the battle space, are all deep cognitive functions, which if allowed to AI, will change the nature of war.

One wonders if AI was developed in the First World War, would it have permitted the decision-making that repeatedly sent millions to their death? Would the Nuclear Attacks in the Second World War have happened on an AI based analysis? Would the Russian Fleet have been sent to do battle with Japan? Would Napoleon have attacked Russia? All in the realm of conjecture.

AI is firmly in the Intelligence space and definitely influencing decision making. In the operational space it manifests the lower end of the spectrum as Lethal Autonomous Weapons Systems (LAWS) which themselves raise substantial uncomfortable questions. Whether as a recoil from LAWS, or from fire on the top (intelligence), the operational space will definitely be invaded by AI. They will impact unpredictably. In war as in conflict, an overall strategic comprehension and understanding of actions in context is essential. More significantly contexts change and the domain of conflict is not rules based. No operational plan survives once war or conflict occurs. Anticipation, dealing with unforeseen or unintended outcomes, empathy, gut feeling, adaptability, are all evolved human cognitive functions. AI will be severely challenged in these domains. As it will surely seek to stray into these areas, General Staff will be stretched to limit them, and retain human discrimination.

The roll of the 21st century, promises to manifest unpredictable and unrestricted. Structures and architecture post the world wars are inadequate. Definitions and contexts of conflict and war are rapidly mutating. The huge efficiencies and consistencies of AI are striking. Its siren song proven in several human enterprises, is compelling. Like brave Ulysses, General Staffs must tie themselves to their ‘ship masts’, so that they do not crash on the seductive rocks of AI. It will take a tremendous effort of discrimination, to marry the promise of AI and retain the largely cognitive and human domain in conflict decision making. At any rate, ‘war as we knew it is gone’. In its place is a range of conflictual options, distinctly different to the spectrum of conflict we spoke of three decades ago.