Skip to content
The United States Army Research Laboratory is investigating the capabilities of OpenAI’s generative AI solutions for battle planning using a simulated military video game environment.
As detailed in a report from the New Scientist, US Army researchers are employing OpenAI’s GPT-4 Turbo and GPT-4 Vision models to collect data on simulated battlefield terrain, identify friendly and enemy forces, and offer military insights on offensive and defensive strategies. They are also utilizing two other AI models based on older technology.
In their experiments, AI assistants were tasked with eliminating all enemy forces and securing a specified objective point. Upon receiving the mission, the AI assistants generated various courses of action, refined by a user acting as a commander. Although OpenAI’s GPT models performed better than the others, they also resulted in more casualties while accomplishing mission objectives.
Integrating generative AI into military strategy is a tiny part of the US Army’s broader effort to harness artificial intelligence.
Project Maven, a significant AI initiative by the US Department of Defense, has shown success in tasks like identifying rocket launchers in Yemen and surface vessels in the Red Sea and aiding in target identification for strikes in Iraq and Syria, according to a Bloomberg report.
However, the potential deployment of AI in battlefield scenarios raises notable ethical concerns. Entrusting decisions that could lead to fatalities to machines draws comparisons to the Terminator movie series rather than the optimistic vision often associated with AI. Nevertheless, the military is persisting with AI development.
The Pentagon has requested significant funding from US lawmakers to advance its artificial intelligence and networking capabilities, as reported by DefenseScoop. Additionally, positions such as Chief Digital and AI Officer have been established to support the adoption and use of this technology throughout the department.
