It is not too long ago since generative AI tools, such as ChatGTP, were being discussed within Arts and Humanities for their ability to generate good quality essays with seemingly no easy or reliable method of detecting their use, but quantitative subjects like physics seemed immune. AI tools did not seem to be very good at solving problems, and as problem solving in all its forms is a huge part of a sound physics education, generative AI did not seem to pose much of a threat. Those days are no more. Generative AI can now solve pretty much any problem we might set to an undergraduate and we face the same challenges that beset our colleagues in the Arts and Humanities.
Sands et al. (Fri,) studied this question.