Does the emergence of AI challenge the nature of warfare?
I do not think so. I think the fundamental nature of war as politically motivated, violent, and chaotic, is going to remain the same. AI will change the character of war in so much as it will allow us to do things that we cannot do without AI systems. At the same time, AI will also introduce its own unpredictable elements. Our adversaries will likely adopt AI systems against us too, compounding the chaos of war. The fundamental nature will remain the same, although the character will change.
What are the main benefits of incorporating AI into Wargaming?
We want our wargames to be able to represent the nature of combat as accurately as possible. It will be necessary to develop wargames to include AI systems in them moving forward. You might also imagine a wargame where there are human players on one side and AI players on another—that might allow an AI system to represent a whole team of individuals and extract important information that is otherwise unavailable without AI. You could expand your universe of gaming opportunities by using AI in the development or the running of wargames themselves.
As we come to learn and understand how AI systems behave, we can represent those behaviors in wargames, thus helping players understand how AI systems might interact with other systems in combat. What are the capabilities and what are the limitations of AI? Wargaming with AI can help us understand the limitations of AI systems in combat.
What are some of the most interesting observations you have gained from your involvement in Wargaming?
Wargaming allows you to gain insights into commander behavior in combat or conflict. When you have the human player making decisions on the spot and human players opposing each other, or even a human player against a computer player or an AI player, you gain insights into how real commanders might make decisions in combat, which are hard to capture ahead of time in modeling or analysis. We use Wargaming to try to gain insights into how events could unfold. We establish the conditions and the capabilities for the players in the game and then let them use them. Oftentimes, we see things that we did not expect beforehand and gain valuable insights.
Could commercial wargames provide a structured environment necessary to capture the large amount of data required to test AI applications for military decision-making?
Commercial wargames can do some very interesting things with AI systems if you can represent the AI behavior at a macro level. If you are talking about how to wargame ground combat vehicles, for example—autonomous vehicles maneuvering on a battlefield, which is what my research with RAND focused on—then a commercial game could provide a framework within which you can insert AI capabilities and see how they work. If you are going to get into the details of how AI systems work, then I think it is harder for a commercial game to do that, because of the large amount of data and the understanding of AI systems that is required. Typically, players of commercial games are not interested in drilling into the engineering details of a system. I think that there are some benefits to commercial wargaming when looking at AI, but also some limitations.
Can AI ever replicate the human judgment capacity to mitigate extreme uncertainty in conflict?
I think that AI systems will help us mitigate some of the uncertainties in combat. I do not think AI will ever replace human judgment. I think human judgment will always be required in certain things that either we cannot anticipate well enough or do not understand well enough to program into AI. There may be things that we do not want to allow AI systems to do. We still want to maintain that human involvement, although AI systems, with their processing power and handling of data, will help us understand things on the battlefield or see things on the battlefield that we cannot see today.
If you are looking at battlefield surveillance, for example, you are collecting data from systems on the battlefield. One system might not reveal all the information you need for a human to understand where an enemy formation is, or how to target it with weapons systems. But if you have a multitude of sensors, and an AI system processing all the data that you are getting from those systems, it could paint a picture for the human commander. This would allow them to understand what is going on in a way that disparate systems of sensors and human observers might not be able to do today.
Canada’s defence policy highlights the rapidly evolving nature of technology as a key trend to be considered, particularly in the regard to our alliances, such as NATO, NORAD, and Five Eyes. I’m sure many other countries have highlighted this trend as well. How can Canada and other allies approach Wargaming scenarios with each other?
Within the US we almost always look to represent allied and partner capabilities in our wargames, because where the US goes and gets involved in a conflict, we are almost always working with or fighting alongside allies and partners. So, that is very important to us. We are almost always looking to represent allied and partner capabilities in our wargames. Sometimes we represent allied and partner capabilities directly by having allied and partner military officers participate in American wargames, which is probably the most direct way of gaining that interaction. Often, when we cannot do that, we will communicate indirectly, for example, through US European command or NATO, with American personnel in direct contact with allied and partner military personnel. With that assistance, we can represent those capabilities in wargames with American players. Regarding American participation in allied and partner wargames, American military personnel often participate in operations with allied personnel. They do and are willing to participate in allied war games as well.
Sean Barnett is a senior engineer at the RAND Corporation. He is a defense analyst, nuclear engineer, and attorney, with 30 years of experience leading research and solving problems in the fields of national security, energy, and law. At RAND, most of his work has involved military operational assessments and modernization, force structure, and force management issues. Specific projects have concerned development of the DoD Joint Warfighting Concept, potential conflicts with Russia, China, and North Korea (and North Korean nuclear weapon developments), Army artillery and aviation requirements, theater-level combat modeling, the Army personnel screening system, and Department of Homeland Security requirements development. Barnett has been involved with wargaming for many years and he leads and supports RAND’s wargame development efforts across multiple theaters and potential conflicts.