What? Having a way to distinguish the simpler deep learning models to the ones boiling our oceans? Maybe we should go back to the pre Geoffrey Hinton definition of AI of automated reasoning?