Comments on “Classifying images: massive parallelism and surface features”