it's less "generating more information" and more "presenting the same information in new ways". A more ideal model wouldn't need augmented data, but this is wha

it's less "generating more information" and more "presenting the same information in new ways". A more ideal model wouldn't need augmented data, but this is what works well with current architectures. It may be that practical constraints mean we never move away from augmentation, just as we'll never move towards single-layer neural nets, even though theoretically they can fit any model.