How will advances in artificial intelligence (AI) shape the future of war? There is a growing belief among some policymakers and analysts that AI will transform the future of war, but researchers are still in the early stages of understanding how AI will actually change warfighting.In this report, the authors offer a conceptual framework and preliminary assessment to help set the terms for a more systematic debate about AI’s military implications. The authors of a new RAND research use the framework to evaluate how AI could affect four “building block” competitions in military affairs: (1) quantity versus quality, (2) hiding versus finding, (3) centralized versus decentralized command and control (C2), and (4) cyber offense versus cyber defense. Their findings suggest that the U.S. military might need to change important aspects of how it traditionally operates in order to exploit AI’s potential.
Key Findings
Quantity could gain a significant edge vis-à-vis quality
- The relative cost-effectiveness of quantity could improve as AI-enabled uncrewed systems become cheaper and more capable. The combination of “precise mass” and “affordable mass” could give a new generation of attritable and expendable uncrewed systems a growing cost advantage over exquisite platforms and weapons for some applications, although exquisite systems will still have important roles.
More-sophisticated hiding could help offset advances in finding, but this will require new approaches and investments in deception
- AI will improve finding by quickly fusing and analyzing intelligence from proliferated sensors. But militaries can develop countermeasures to keep the hider-finder competition contested, including leveraging AI to orchestrate sophisticated deception campaigns.
Mission command—a hybrid of centralized and decentralized C2 models—will remain desirable
- AI does not change the reasons mission command has advantages over more-centralized or more-decentralized C2 paradigms, which are rooted in having the right information at the right places to make time-sensitive decisions. Access to information, not just cognitive capacity, acts as a limiting factor on shifting toward alternative approaches to C2.
Cyber defenses will benefit from AI in ways that could make battle networks more resilient against cyberattacks in the long term
- AI could help address challenges with scale, speed, and effectiveness that currently limit the cyber defense and give the cyber offense a structural edge. However, cyberattackers will always retain some ability to penetrate networks, particularly with help from AI.
Recommendations
- The United States should invest more in research and experimentation for new capabilities that enable the greater use of mass and deception. Examples include exploring attritable runway-independent uncrewed aircraft for air combat missions and new AI tools for deception to counter adversaries’ AI-enabled sensing. Opportunities to capitalize on the trends identified in this report are available today, even if AI technology is still maturing.
- The United States needs to allocate scarce resources under the assumption that it will face sophisticated and adaptive adversaries. There might be temptations to use AI to double down on existing approaches in the hope that the United States will have an unassailable AI first-mover advantage, allowing it to operate in ways that defy the trends mentioned in this report. The United States should instead allocate scarce resources toward approaches that will provide durable long-term advantages.
- The United States needs a plan to manage the transition to an AI-enabled force. Important questions remain about the implementation details of leveraging AI; for example, what, exactly, might a new balance between exquisite capabilities and robotic mass in the U.S. force structure look like, how can the United States field AI tools to units in a thoughtful way that balances speed and the need for trust and reliability, and how can the United States train both human personnel and AI tools in ways that reinforce strengths and offset weaknesses?
– Published courtesy of RAND.
