"I basically told them, in no uncertain terms, that this was not the way that I thought physics research should be going. And they agreed with me, actually, about the off-the-shelf use of machine learning."