斯德哥尔摩国际和平研究所-Mapping-the-development-of-autonomy-in-weapon-systems_-A-primer-on-autonomy_40页_684kb

Summary of "Mapping the Development of Autonomy in Weapon Systems"

Core Content

This document provides an overview of the concept of autonomy in weapon systems, focusing on its technological, legal, and ethical implications. It is part of a series of working papers by the Stockholm International Peace Research Institute (SIPRI) aimed at supporting discussions on lethal autonomous weapon systems (LAWS) within the framework of the 1980 United Nations Convention on Certain Conventional Weapons (CCW).

Main Points

1. Definition of Autonomy

Autonomy in weapon systems is defined as the ability of a machine to execute tasks without direct human input, using interaction with the environment. It is a three-dimensional concept, which includes:

• Human-Machine Command-and-Control Relationship: Systems are classified as semi-autonomous, human-supervised autonomous, or fully autonomous based on the level of human involvement.
• Sophistication of Decision-Making: Systems may be automatic (mechanical response), automated (programmed to follow rules), or autonomous (capable of making decisions based on higher-level intent).
• Functional Tasks: Autonomy may be applied to different functions within weapon systems, such as mobility, health management, interoperability, battlefield intelligence, and use of force.

2. Autonomy in Weapon Systems

Autonomy can be found in various weapon systems, but the real concern lies in its application to use of force functions, such as target detection, identification, tracking, and attack. These are often referred to as "critical functions" and are the focus of legal and ethical debates.

3. Examples of Autonomous Weapon Systems

Several weapon systems already incorporate some level of autonomy, including:

• Missile and rocket defense systems (e.g., Goalkeeper, Iron Dome)
• Active vehicle protection weapons (e.g., Trophy)
• Anti-personnel sentry weapons (e.g., Samsung SGR-A1)
• Smart sensor-fused munitions (e.g., Bonus 155, SMArt 155)
• Guided missiles (e.g., Brimstone, BrahMos)
• Loitering munitions (e.g., Harpy, TARES)
• Encapsulated torpedoes and mines (e.g., CAPTOR, PMK-2, Sea Urchin)

These systems are typically semi-autonomous or human-supervised, with varying levels of decision-making autonomy.

4. How Autonomy Works

Autonomy is enabled through three core capabilities:

• Sense: Collecting and interpreting environmental data using sensors and software.
• Decide: Using control systems to determine the course of action based on the data.
• Act: Executing the determined actions.

The document highlights the importance of machine learning and artificial intelligence in enhancing autonomy, particularly in decision-making processes.

5. Feasibility and Challenges

• Achieving full autonomy in weapon systems is technically challenging and remains a topic of debate.
• Current systems rely on semi-autonomous or human-supervised models.
• Machine learning is considered a key enabler for future advances in autonomy, though its ethical and legal implications are still under discussion.
• There is no consensus on the definition of autonomy, leading to confusion in policy and legal discussions.

6. Conclusion and Takeaways

• The discussion on LAWS is still in its early stages, with many states not yet ready to commit to a ban.
• The research project aims to provide a reality check on autonomy by mapping its development across different perspectives: technical, economic, operational, and political.
• Future discussions on LAWS should focus on understanding the conceptual and technical foundations of autonomy, as well as the operational and legal implications of its application in weapon systems.

Key Information

• Autonomy is a relative and multifaceted concept.
• Three dimensions of autonomy are:
  • Human-machine command-and-control relationship
  • Sophistication of decision-making
  • Functional tasks being automated
• Current systems are mostly semi-autonomous or human-supervised.
• Technological enablers include sensors, computer vision, machine learning, and AI.
• Legal and ethical concerns center around the use of force and decision-making capabilities.
• SIPRI is conducting a comprehensive project to support structured discussions on LAWS.

Appendix Overview

• Appendix A lists existing definitions of autonomous weapon systems, highlighting the lack of consensus.
• The types of weapon systems that may be considered autonomous are discussed in section 3.

Summary of Key Questions

The working paper addresses the following fundamental questions:

1. What is autonomy? How does it work? How is it created?
2. What are the underlying technologies and where are they being developed?
3. What types of autonomous applications are found in existing and forthcoming weapon systems?
4. What are the capabilities of weapons with autonomy in the target cycle?
5. What trends are fueling or limiting the development of autonomy in weapon systems?

Conclusion

This document serves as a foundational resource for understanding the development, application, and implications of autonomy in weapon systems. It emphasizes the need for clear definitions, technical understanding, and structured dialogue to address the legal, ethical, and strategic challenges posed by increasing autonomy in military technologies.