Only tenuously related to your point (which I agree with), but I think we'll have better-than-human deep learning based vocals in less than a decade.