We seem to be crawling out of the "uncanny valley" on the other side. Give it a couple more years, we've pretty much nailed the interpolation of human output (across conventional modalities).