Researchers from the University of Maryland created a list of 25 questions related to breast cancer screening guidelines — then submitted them to ChatGPT to gauge their accuracy. Here's what happened.