This image (and the text around it) is certainly already in the training data. Need to create that scenario from scratch with some grad students and then try it.