In my opinion, the bottom line is, auto-tune can achieve better "mechanical perfection" but that better mechanical perfection doesn't equate to better music.
Music is what? Not just a series of notes, but instead it is a communication to convey feelings and emotion via sound. People do buy the same piece of music from different artist because they communicate different emotions and communicate them differently.
The pauses, the way the note transit from one note to another, the volume of a note, and the rise and fall of a note, all convey information -- all adds to the whole of the performance.
Typical music prodigies achieved their "mechanical perfection" at an early age, but never having experienced say for example "a broken heart from being abandoned by boyfriend/girlfriend". Thus, they do not have the romantic depth in feeling that someone older with actual romantic relationship experience has. So, be it instrumental or vocal, their playing often is "wooden" as compared to older musicians. They could imitate how other artist convey the emotion, but in doing so, they are not conveying their own emotion -- they don't yet has that emotion to convey. Auto-tune will at best make an "amateur level skilled" musician sound like a good prodigy but fails at delivering the emotions that make the music come alive.
In my opinion, yeah, live performance with an audience is better - because the artist's emotion and the audience's emotion feed on each other. There is a positive feedback loop there. Recording of course will depend on the media as different media may miss the more subtle qualities differently.
Until "Commander Data" achieve his dream of having emotion, his experience in life will never equal that of a human. Same for AI today.
EDIT: typo correction, mistyped "node" which should be "note" in a few places. Musical notes is a thing, musical nodes isn't.