AI could influence 'the whole future of war,' starting with gaming out the next fight, experts say

• Experts suggest Generative AI could improve decision-making in wargaming.
• Wargaming faces issues in cost, transparency, and accessibility, which AI could help address.
• Despite the potential benefits of AI, there are also limitations that users have to keep in mind.

The way militaries prepare for and fight the wars of the future may be radically changed as machines increasingly learn to think through complex scenarios at both the tactical and strategic levels.

Introducing generative artificial intelligence into wargaming, extensive simulations used by experts, policymakers, and the military to develop strategy in the event of war, has the potential to improve the decision-making process for participants.

It's something that has appeared in fiction writing on imagined future wars but is also being looked at right now.

AI "can shape the wargames and actually the whole future of war," Yasir Atalan, an associate data fellow at the Center for Strategic and International Studies, told Business Insider. "It will allow the policymakers, planners to see what are the weaknesses, what are the opportunities, what are the challenges."

In a piece Atalan co-wrote with CSIS's Benjamin Jensen and Dan Tadross, some of the greatest issues plaguing wargaming include its cost, transparency, and accessibility.

Wargames can cost thousands to millions of dollars and are somewhat exclusive in that they don't always take into account the perspectives of qualified players.

"Games pivot on the quality of the players, but the best players are often overbooked and on the move," the CSIS report said. "Rather than directly relying on human players sitting around a table to play a game, the twenty-first-century analyst can use generative AI and LLMs to create game agents."

Talking with BI about the opportunities available with artificial intelligence, Atalan said that with the right prompting and data curation, AI could assist players by contributing "different perspectives from different allies" and provide probability distribution and other data points in a wargame.

Wargaming is often misunderstood as a tool to predict outcomes, but Atalan, along with other experts emphasize that wargaming is not about that.

"It's about for the players, for the different teams, for the stakeholders, to see what are their range of policy options, what are the possible opportunities and challenges," he said.

The US military began putting increased emphasis on the need for wargaming in 2015 and over the years, in concert with the hype, there has been growing interest in the possibilities of AI. In February 2023, for instance, the US military let AI successfully pilot a fighter jet and engage in simulated air-to-air combat.

There's a growing realization the two interests could be combined to leverage the potential of AI.

"Wargaming and simulation continue to be crucial tools for decision-makers in Defence," a 2023 study from the London-based Alan Turing Institute concluded. "They can be used to train personnel for future conflict, and offer insights on critical decisions in warfighting, peace negotiations, arms control, and emergency response."

The study also found that using AI for wargaming could potentially cut down the number of players needed, speed up games, produce new strategies, better immerse players in games, and speed up game completion.

AI also has its limits in this space. A recent RAND study found that AI may not work as well in a wargame if the digital infrastructure of a game is restricted or if AI can't directly interact with the computational models and simulations needed.

Wargaming expert Ivanka Barzashka has also raised concerns that AI may obscure explanations for actions, potentially leading to faulty conclusions.

"The current landscape of human-centric wargaming, combined with AI algorithms, faces a notable "black box" challenge, where the reasoning behind certain outcomes remains unclear," she wrote in a post for The Bulletin of the Atomic Scientists.

Barzashka argued that "this obscurity, alongside potential biases in AI training data and wargame design, highlights the urgent need for ethical governance and accountability in this evolving domain."

Right now, Atalan told BI, it is difficult for AI to participate on its own in wargaming because it can't perform "strategic reasoning," which might take into account different schools of thought.

Organizations like OpenAI have been gearing up to create "autonomous agents," that would be able to perform tasks on their own, but Atalan said the current version of wargaming agents are only able to offer the most likely responses based on an average of data compositions and the internet.

Atalan said he would want to be "cautious" with autonomous agents as AI can inherit bias and prejudice depending on the Large Language Models or LLMs it's trained on.

"When people are using these LLMs in their approach, they need to be transparent, they need to show their prompting," Atalan said. "With new experiences and studies we will see what kinds of normative prejudices of these models have for certain actors and also in different languages."

"It's impossible to replace humans," Atalan said.

"Humans are still the core of the whole process," he said, "but the AI assistance can be really useful about the whole thought process and decision-making processes for the experts."