I’ve used it to show that PLMs are meant for a certain kind of target: discriminating or classifying things that look the same to the untrained eye but are not. To me it’s absolutely worth the extra time if there’s one specific perceptual knot that’s giving you a migraine.