I would call those "scaling issues". Let's see if deep learning advances along the lines of gpt2 can start solving them. I think that we will know the answer in 5 years.