In a twist that sounds straight out of a sci-fi novel, the US Army is test-driving OpenAI's GPT-4, in war game simulations. Dubbed "COA-GPT," this system plays the role of a digital general, whipping up military maneuvers with a speed and efficiency that traditional processes can't match.
Is pushing the envelope on AI's capabilities worth the risk, especially when the outcome could be written in casualty reports rather than code?
Think this is an exaggeration? Take a look at the COA-GPT research out of the DEVCOM Army Research Laboratory - https://arxiv.org/pdf/2402.01786.pdf
