In this episode, Matt and Kevin discuss a new paper (still in early release at the time of recording) where authors use a large language model - specifically chatGPT - to make virtual AI scientists to perform a novel scientific project. They find it... some what lacking.
But it's clearly something that people are increasingly going to do, some even think it's the future of science! If it is - we're not there quite yet, but you can be sure this paper is going to get cited a bunch!