Rafael Harth comments on Learning is (Asymptotically) Computationally Inefficient, Choose Your Exponents Wisely