To calculate just how much generative AI contributes to this problem, Tzachor and his colleagues examined the type and volume of hardware used to run large language models, the length of time that …