Looking at a variety of armament sectors, the book examines how Artificial Intelligence (AI) impacts the fields of armament and arms control, how existing arms control measures will be affected by AI, and what new approaches based on AI have been or are currently developed.
The significant increase in computing power, the increasing reliance on software, and the advent of (narrow) AI and deep-learning algorithms all have the potential to lead to disruptive changes for military operations and warfare, rendering many classical arms control instruments less effective, or even useless. On the other hand, AI might lead to completely new arms control approaches, raising the effectiveness and reliability of new verification measures. To provide a common understanding, the book starts by presenting a general introduction to the state of the art in artificial intelligence and arms control, and how the two topics are interrelated. The second part of the book looks at examples from various fields of weapon technology, including weapons of mass destruction (WMD), conventional armament, and emerging technologies. The final section offers a cross-cutting perspective based on the examples presented in the second part.
This volume will appeal to students and scholars of international relations, as well as policy-makers and practitioners interested in a better understanding of peace and security studies in general, and armament and arms control in particular with a strong focus on AI.
Author(s): Thomas Reinhold, Niklas Schörnig
Series: Studies in Peace and Security
Publisher: Springer
Year: 2022
Language: English
Pages: 236
City: Cham
Acknowledgements
Contents
Editors and Contributors
Abbreviations
Introduction
1 The Use of AI as a Revolution in Military Affairs
2 The Purpose of the Book
3 The Structure of the Book
4 Conclusion
References
Introduction into Artificial Intelligence and Machine Learning
1 Introduction
2 The History of Artificial Intelligence and Machine Learning
3 Machine Learning Algorithms
4 Measuring the Quality of Machine Learning Algorithms
5 Conclusions
Appendix
Artificial Neural Networks
Machine Learning Metrics
Classification Metrics
Classification Accuracy
Detection Rate
Precision
Recall
Specificity
Balanced Accuracy
F1 Score
Regression Metrics
Mean Absolute Error
Mean Squared Error
References
The Military Rationale for AI
1 Introduction
2 AI and the Military
3 Weapon Autonomy and ``Fighting at Machine Speed´´
4 Conclusion: The Hype Is Real-And So Are the Risks
Further Reading
References
Military AI Applications: A Cross-Country Comparison of Emerging Capabilities
1 Introduction
2 Assessing Emerging Military AI Capabilities: A Framework for Analysis
2.1 The Sources of Military AI
2.2 Areas of Military AI Applications
2.3 Risks Posed by Military AI Applications
3 Assessing Military AI Capabilities: A Review of Four Countries
3.1 The United States
3.1.1 Sources of Military AI
3.1.2 Areas of Military AI Application
3.1.3 Risks of Military AI Applications
3.1.4 Summary
3.2 China
3.2.1 Sources of Military AI
3.2.2 Military AI Applications
3.2.3 Risks of Military AI Applications
3.2.4 Summary
3.3 France
3.3.1 Sources of Military AI
3.3.2 Areas of Military AI Applications
3.3.3 Risks of Military AI Applications
3.3.4 Summary
3.4 Israel
3.4.1 Sources of Military AI
3.4.2 Military AI Applications
3.4.3 Risks of Military AI Applications
3.4.4 Summary
4 Comparison and Discussion
References
Artificial Intelligence as an Arms Control Tool: Opportunities and Challenges
1 Introduction
2 Theory of Arms Control, Disarmament and Non-proliferation
2.1 Arms Control, Disarmament and Non-proliferation: Basic Concepts
2.2 Verification
2.3 AI-Enhanced Verification from a Theoretical Perspective
3 AI for Enhancing Arms Control and Verification
3.1 Translation and Analysis of Text
3.2 Analysis of Graphical Data
3.3 Analysis of Other Sensor Data
3.4 Multimodal Data and Other Uses
4 Limitations and Challenges
5 Conclusion
References
Verifying the Prohibition of Chemical Weapons in a Digitalized World
1 Introduction
2 Verification Under the Chemical Weapons Convention
2.1 Routine Verification
2.2 Non-routine Verification
2.3 Beyond Traditional Verification
3 Science and Technology Under the CWC
3.1 CWC Provisions
3.2 The SAB Report to the 4th CWC Review Conference: Toward a Holistic Approach to Verification
4 CWC Verification in the Age of Digitalization
4.1 Digitalization, Artificial Intelligence and Security
4.2 What Is Artificial Intelligence?
4.3 Digitalization in Routine Chemical Weapons Convention Verification
4.4 Digitalization in Non-routine Chemical Weapons Convention Verification
4.5 Will Digitalization Disrupt Verification Procedures?
5 Conclusions
References
AI and Biological Weapons
1 Introduction
2 Adding Computing Power to Bioinformatics
3 Mounting Security Concerns
4 How Should the BWC Respond?
5 Evolving Biological Arms Control
References
Doomsday Machines? Nukes, Nuclear Verification and Artificial Intelligence
1 Introduction
2 AI in Nuclear Weapon Systems
2.1 Qualitative Improvements in Nuclear Weapon Systems
2.2 The Other (Dark) Side of the Coin
2.3 The Nuclear-Conventional Nexus and Nuclear Psychology
3 Opportunities for Nuclear Verification
4 Advanced Wonder Weapon or Doomsday Machine?
References
AI, WMD and Arms Control: The Case of Nuclear Testing
1 The Nuclear Testing Status Quo
2 Application of AI in Nuclear Testing
2.1 AI and Virtual Testing
3 Detecting Nuclear Tests: Present State of the Art
3.1 Detecting a Nuclear Test with Seismic Waveform Analysis
3.1.1 Creating a High-Quality Event Bulletin
3.1.2 Yield Estimation
3.2 Radionuclides
3.3 Threads for AI in Nuclear Test Detection
4 Conclusion
References
Artificial Intelligence in Conventional Arms Control and Military Confidence-Building
1 Introduction
2 Theoretical Approaches to Arms Control and Military Confidence-Building
2.1 The Structural Realist Approach to Arms Control and Military Confidence-Building
2.2 The Neoliberal Institutionalist Approach to Arms Control and Military Confidence-Building
2.3 The Constructivist Approach to Arms Control and Military Confidence-Building
2.4 Summary
3 Conventional Arms Control and Military Confidence-Building in Europe: A Brief Overview
3.1 The Vienna Document (VDoc)
3.2 The Treaty on Conventional Armed Forces in Europe (CFE Treaty)
3.3 The Treaty on Open Skies (OS)
3.4 Major Challenges
4 Artificial Intelligence and Machine Learning in Conventional Arms Control: Opportunities, Challenges and Risks
4.1 Changes in the Balance of Power
4.2 Analysis, Planning, Coordination and Evaluation
4.3 More Thorough and Comprehensive Verification
4.4 The Importance of Maintaining the ``Human Factor´´ in Arms Control
5 Concluding Remarks
References
Cyber Weapons and Artificial Intelligence: Impact, Influence and the Challenges for Arms Control
1 Introduction
2 Cyber Weapons and the Militarization of Cyberspace
2.1 The Current Situation of State-Driven Cyberattacks
3 How the Technology of Cyber Weapons and Its Application Will Evolve
4 How Artificial Intelligence and Machine Learning Could Influence Cyber Weapons
4.1 Explainability and Responsibility of AI-Enabled Cyber Weapons
4.2 AI and the Pitfalls of the Attribution of Cyberattacks
5 The Negative Impact on Arms Control of Artificial Intelligence in Cyber Weapons
6 How Can Artificial Intelligence Support Cyber Arms Control?
7 Conclusion
References
Drones and Lethal Autonomous Weapon Systems
1 Introduction
2 Drones and LAWS: Technology and Functions
2.1 Remotely Piloted Drones: Steps Toward Autonomy
2.2 LAWS: A New Quality of Autonomous Functions Enabled by AI and ML
3 LAWS Necessitate a New Perspective on Arms Control
3.1 The Arms Control Debate on LAWS
3.2 Human Control: The Decision-Making Process as a New Subject of Regulation
3.3 Specific Challenges Arising from Enabling Technologies
3.4 Positive Impact of AI on Arms Control for LAWS
4 Conclusion
References
No, Not That Verification: Challenges Posed by Testing, Evaluation, Validation and Verification of Artificial Intelligence in ...
1 Introduction
2 TEV&V: The Traditional Way
2.1 Military TVVE
2.2 Verification and Validation
2.3 Methods
3 What Is AI?
4 Problems with the Technology
4.1 The Logics of AI
4.2 Developing ML
4.3 Integrating AI
5 Problems with the Process
5.1 Testing AI
5.2 Measuring AI
6 Implications for Arms Control
7 The Search for Solutions
8 Conclusion
References
Further Reading
Applying Export Controls to AI: Current Coverage and Potential Future Controls
1 Introduction
2 Current Export Controls on AI and Related Hardware, Software and Technology
2.1 Key Types of Export Controls on AI
2.1.1 Hardware Controls
2.1.2 Software Controls
2.1.3 Technology Controls
2.1.4 Catch-All Controls
3 National and Multilateral Approaches to Expanding Export Controls on AI
3.1 US Emerging Technologies Review Process
3.2 EU Emerging Technologies Review and Dual-Use Regulation Recast
3.3 Wassenaar Arrangement Discussions on Export Control on AI
4 Challenges and Opportunities of Applying Export Controls to AI
4.1 Challenges and Adverse Consequences
4.1.1 Implementation Challenges
4.1.2 Potential Adverse Consequences
4.1.3 Conflicting Aims of Export Controls on AI
4.2 Opportunities and Benefits
5 Conclusion
References
Arms Control for Artificial Intelligence
1 Introduction: Or why Hard Arms Control for Artificial Intelligence Should Be Considered
2 The Rise of Artificial Intelligence and Its Militarization
3 Best Practices and Lessons Learned from Other Technologies
4 The AI Life Cycle: The Components of Artificial Intelligence Applications
5 The Components of AI Development: Applying Tailored Arms Control Measures
5.1 The Training Data
5.2 The Classifiers
5.3 The Model
5.4 The Effectors
6 Verification
7 Pre-conditions and Pitfalls for Arms Control
8 Confidence-Building Measures for Military AI Applications: An Alternative?
9 Conclusion: Or How AI May Develop and What Arms Control Can Do About It
References
