"Matthew Strohmeyer, the colonel, said a large language model, or LLM, zipped through a task in 10 minutes that usually would have taken humans hours or days to complete.
"It was highly successful," he told Bloomberg. "It was very fast."
The story said the Pentagon wouldn't say which LLMs it's testing, but it did say the models are being fed some "secret-level" data to see how they function in real-time.
The Defense Department could eventually use AI to help make decisions. The current eight-week exercise runs until July 26, Bloomberg reported. Representatives for the Department of Defense did not immediately respond to Insider's emailed request for comment ahead of publication."
"In February, the State Department outlined guardrails that countries should adopt around testing and implementing AI tools, with the stated goal of making sure they would "minimize unintended bias and accidents."
That includes making sure that trained humans experts have decision-making control and oversight of AI tasks.
And it also warns of a robot's finger on the nuclear trigger. The guidelines say humans should be involved in "all actions critical to informing and executing sovereign decisions concerning nuclear weapons employment."
"Military use of AI capabilities needs to be accountable, including through such use during military operations within a responsible human chain of command and control," the department said in a declaration."
https://www.yahoo.com/finance/news/air-force-colonel-says-took-090400445.html