The United States military has embarked on a series of tests to determine the effectiveness of generative AI in assisting with global conflict planning and streamlining internal information access. According to a report by Bloomberg on July 6, the U.S. Department of Defense, in collaboration with undisclosed allies, is conducting experiments involving five large language models (LLMs) in the digital and AI office at the Pentagon.
While the specific LLMs undergoing testing remain confidential, AI startup Scale AI has confirmed that its “Donovan” model is among the five selected for evaluation. Air Force Colonel Matthew Strohmeyer shared with Bloomberg that an initial test of an LLM demonstrated highly successful and rapid performance. However, Strohmeyer also emphasized that the technology is not yet ready for full deployment, stating, “We are learning that this is possible for us to do, but it’s not ready for primetime right now.”
One notable test described by Strohmeyer showcased an AI model’s ability to provide the requested information within an astonishing 10-minute timeframe. This represents a significant improvement compared to the typical days and multiple personnel involvement timeline. The LLMs have already been exposed to classified operational information, generating responses to real-world scenarios. The tests aim to determine whether these models can aid in planning responses to potential escalations, such as the tense military situation with China in the Indo-Pacific region.
While the current tests are scheduled to conclude on July 26, the U.S. military has been actively exploring AI’s potential applications in warfare for some time. In a groundbreaking joint trial conducted by the British government’s Defence Science and Technology Laboratory (Dstl) in May, the U.S. and Australia collaborated to test AI-enabled military drones for target tracking and detection. Dstl highlighted several notable achievements during the trial, including live retraining of AI models during flight and seamless interchangeability of models within the AUKUS alliance. These advancements demonstrate a concerted effort to integrate these technologies into military capabilities swiftly.
As the tests progress, the U.S. military is paving the way for more efficient and informed decision-making processes by applying generative AI. While challenges and further development lie ahead, the potential for enhanced planning and expedited access to critical information holds promise for the military’s strategic operations and responses in an increasingly complex global landscape.
