The flip side is it's very unlikely that AI won't become that good any time soon, so it'll always remain a means to hold out. Especially since nobody has explicitly defined what "good enough" entails.