The US Army has been exploring the use of artificial intelligence (AI) models to enhance battle planning, as reported by the New Scientist. This experimentation involves utilizing Microsoft-backed OpenAI’s generative AI solutions typically used for commercial chatbots. By testing these AI assistants in a military science fiction video game, Starcraft II, the Army aims to evaluate their effectiveness in a simulated war game scenario.
During the experiment, researchers focused on a straightforward scenario with limited military units and full knowledge of the battlefield. AI chatbots, including OpenAI’s GPT-4 Turbo and GPT-4 Vision models, were tasked with providing information on the battlefield terrain, friendly and enemy forces, and strategic military advice on attacking and defending. The AI assistants were then given a mission to destroy all enemy forces and capture an objective point, responding promptly with proposed courses of action for the commander to refine before finalizing orders.
While the AI assistants based on OpenAI’s GPT models outperformed older AI agents in the experiment, they were not flawless and incurred more casualties to achieve mission objectives. This testing showcases the potential of AI technology in improving battle planning and decision-making processes within the military.
The US Army has previously explored AI technologies for battlefield image analysis and target identification for airstrikes. By incorporating AI models like OpenAI’s GPT into battle planning, the Army aims to leverage advanced technology for more efficient and effective military operations in the future.
