Artificial intelligence and the future of warfare

The USA, China, and strategic stability

James Johnson

• Publisher: Manchester University Press
• Year: 2021
• ISBN:
  9781526145079
• Topic: Computer Science
• Subtopic: Artificial Intelligence (AI) & Semantics
• Index: Computer Science

This volume offers an innovative and counter-intuitive study of how and why artificial intelligence-infused weapon systems will affect the strategic stability between nuclear-armed states. Johnson demystifies the hype surrounding artificial intelligence (AI) in the context of nuclear weapons and, more broadly, future warfare. The book highlights the potential, multifaceted intersections of this and other disruptive technology – robotics and autonomy, cyber, drone swarming, big data analytics, and quantum communications – with nuclear stability. Anticipating and preparing for the consequences of the AI-empowered weapon systems are fast becoming a critical task for national security and statecraft. Johnson considers the impact of these trends on deterrence, military escalation, and strategic stability between nuclear-armed states – especially China and the United States. The book draws on a wealth of political and cognitive science, strategic studies, and technical analysis to shed light on the coalescence of developments in AI and other disruptive emerging technologies. Artificial intelligence and the future of warfare sketches a clear picture of the potential impact of AI on the digitized battlefield and broadens our understanding of critical questions for international affairs. AI will profoundly change how wars are fought, and how decision-makers think about nuclear deterrence, escalation management, and strategic stability – but not for the reasons you might think.

Part I

Destabilizing the AI renaissance

1

Military AI primer

What is AI, and how does it differ from other technologies? What are the possible development paths and linkages between these technologies and specific capabilities, both existing and under development? This chapter defines and categorizes the current state of AI and AI-enabling technologies.¹ The chapter highlights the centrality of machine learning (ML),² and autonomous systems (or ‘machine autonomy’),³ to understanding AI in the military sphere and the potential uses of these nuanced approaches in conjunction with AI at both an operational and strategic level of warfare.

The chapter proceeds in two sections. The first contextualizes the evolution of AI within the field of science and engineering to make intelligent machines. It defines AI as a universal term that can improve automated systems’ performance to solve a wide variety of complex tasks. Next, it describes some of AI’s limitations in order to clearly understand what AI is (and what it is not), and how best to implement AI in a military context. This section ends with a brief primer on ML’s critical role as an enabler (and subset) of AI, based on computational systems that can learn and teach through a variety of techniques.

The second section demystifies the military implications of AI and critical AI-enabling technology. It debunks some of the misrepresentations and hyperbole surrounding AI. Then it describes how ML and autonomy could intersect with nuclear security in a multitude of ways, with both positive and negative implications for strategic stability – chapters 2 and 8 return to this idea. Next, it conceptualizes AI-augmented applications into those that have predominately operational, tactical, and strategic consequences in future warfare. This section argues that while the potential tactical and operational impact of AI is qualitatively self-evident, its effect at a strategic level remains uncertain.

What is (and is not) AI?

AI research began as early as the 1950s as a broad concept concerned with the science and engineering of making intelligent machines.⁴ In the decades that followed, AI research went through several development phases – from early exploitations in the 1950s and 1960s, the ‘AI Summer’ during the 1970s through to the early 1980s, and to the ‘AI Winter’ from the 1980s. Each of these phases failed to live up to its initial, and often over-hyped, expectations – in particular, when intelligence has been confused with utility.⁵ Since the early 2010s, the explosion of interest in the field (or the ‘AI renaissance’) occurred due to the convergence of four critical enabling developments:⁶ the exponential growth in computing processing power and cloud computing; expanded data-sets (especially ‘big-data’ sources);⁷ advances in the implementation of ML techniques and algorithms (especially deep ‘neural networks’);⁸ and the rapid expansion of commercial interest and investment in AI technology.⁹

AI is concerned with machines that emulate capabilities which are usually associated with human intelligence, such as language, reasoning, learning, heuristics, and observation. Today, all practical (i.e. technically feasible) AI applications fall into the ‘narrow’ category, and less so, artificial general intelligence (AGI) – or ‘superintelligence.’ Narrow AI has been broadly used in a wide range of civilian and military tasks since the 1960s,¹⁰ and involves statistical algorithms (mostly based on ML techniques) that learn procedures through analysis of large training data-sets designed to approximate and replicate human cognitive tasks.¹¹ ‘Narrow AI’ is the category of AI that this book refers to when it assesses the impact of this technology in a military context.

Most experts agree that the development of AGI is at least several decades away, if feasible at all.¹² While the potential of AGI research is high, the anticipated exponential gains in the ability of AI systems to provide solutions to problems today are limited in scope. Moreover, these narrow-purpose applications do not necessarily translate well to more complex, holistic, and open-ended environments (i.e. modern battlefields), which exist simultaneously in the virtual (cyber/non-kinetic) and physical (or kinetic) plains.¹³

That is not to say, however, the conversation on AGI, and its potential impact should be entirely eschewed. If and when AGI does emerge, then ethical, legal, and normative frameworks will need to be devised to anticipate the implications for what would be a potentially pivotal moment in the course of human history. To complicate matters further, the distinction between narrow and general AI might prove less of an absolute (or binary) measure. Thus, research on narrow AI applications, such as game playing, medical diagnosis, and travel logistics, often results in incremental progress on general-purpose AI – moving researchers closer to AGI.¹⁴

AI has generally been viewed as a sub-field of computer science, focused on solving computationally hard problems through search, heuristics, and probability. More broadly, AI also draws heavily from mathematics, human psychology and biology, philosophy, linguistics, psychology, and neuroscience (see Figure 1.1).¹⁵ Because of the divergent risks involved and development timeframes in the two distinct types of AI, the discussion in this book is careful not to conflate them.¹⁶ Given the diverse approaches to research in AI, there is no universally accepted definition of AI¹⁷ – confusing when the generic term ‘Artificial Intelligence’ is used to make grandiose claims about its revolutionary impact on military affairs, or ‘revolution in military affairs.’¹⁸ Moreover, if AI is defined too narrowly or too broadly, we risk understating the potential scope of AI capabilities; or, juxtaposed, fail to specify the unique capacity that AI-powered applications might have, respectively. A recent US congressional report defines AI as follows:

Any artificial system that performs tasks under varying and unpredictable circumstances, without significant human oversight, or that can learn from their experience and improve their performance … they may solve tasks requiring human-like perception, cognition, planning, learning, communication, or physical action (emphasis added).¹⁹

Figure 1.1 Major research fields and disciplines associated with AI

In a similar vein, the US DoD recently defined AI as:

The ability of machines to perform tasks that normally require human intelligence – for example, recognizing patterns, learning from experience, drawing conclusions, making predictions, or taking action – whether digitally or as the smart software behind autonomous physical systems (emphasis added).²⁰

AI can be best understood as a universal term for improving the performance of automated systems to solve a wide variety of complex tasks including:²¹ perception (sensors, computer vision, audio, and image processing); reasoning and decision-making (problem-solving, searching, planning, and reasoning); learning and knowledge representation (ML, deep networks, and modeling); communication (language processing); automatic (or autonomous) systems and robotics (see Figure 1.2); and human–AI collaboration (humans define the systems’ purpose, goals, and context).²² As a potential enabler and force multiplier of a portfolio of capabilities, therefore, military AI is more akin to electricity, radios, radar, and intelligence, surveillance, and reconnaissance (ISR) support systems than a ‘weapon’ per se.²³

Figure 1.2 The linkages between AI & autonomy

AI suffers from several technical shortcomings that should prompt prudence and restraint in the early implementation in a military context. Today, AI systems are brittle and can only function within narrowly pre-defined problem-sets and context parameters.²⁴ Specifically, AI cannot effectively and reliably diagnose errors (e.g. sampling errors or intentional manipulation) from complex data-sets, and the esoteric mathematics underlying AI algorithms.²⁵ Further, AI systems are unable to handle novel situations reliably; AI relies on a posteriori knowledge to make inferences and inform decision-making. Failure to execute a particular task, especially if bias results are generated, would likely diminish the level of trust placed in these applications.²⁶ Therefore, in order to mitigate the potentially destabilizing effects from either poorly conceptualized an accidental-prone AI – or states’ exaggerating (or underestimating) the strategic impact of military AI capabilities²⁷ – decision-makers must better understand what AI is (and what it is not), its limitations, and how best to implement AI in a military context.²⁸

Machine learning: a critical AI-enabler but no ‘alchemy’²⁹

ML is an approach to software engineering developed during the 1980s and 1990s, based on computational systems that can ‘learn’³⁰ and teach themselves through a variety of techniques, such as neural networks, memory-based learning, case-based reasoning, decision-trees, supervised learning, reinforcement learning, unsupervised learning, and, more recently, generative adversarial networks. Consequently, the need for cumbersome human hand-coded programming has been dramatically reduced.³¹

From the fringes of AI until the 1990s, advances in ML algorithms with more sophisticated connections (i.e. statistics and control engineering) emerged as one of the most prominent AI methods (see Figure 1.3). In recent years, a subset of ML, deep learning (DL), has become the avant-garde AI software engineering approach, transforming raw data into abstract representations for a range of complex tasks, such as image recog...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright
• Contents
• List of figures
• Acknowledgments
• List of abbreviations
• Introduction: opening the AI Pandora’s box
• Part I: Destabilizing the AI renaissance
• Part II: Military AI superpowers
• Part III: Nuclear instability redux?
• Conclusion: managing an AI future
• Index