I think they may have provided fewer citations because it felt like a less controversial claim. I think the choice of words was just a bit awkward. To me, it seems like they were asserting that deep learning requires lots of computational resources which is common knowledge. In general, this translates to higher energy requirements.