"While AI and software-generated voices are commonly compared to human voices, that comparison is not very meaningful. AI sounds like AI and humans sound like humans."
If the purpose of A.I. Synth vocals is to replace human vocals then the comparison is both meaningful and necessary.
Synths were, initially, promoted as ways to synthetically replicate traditional instruments. Wendy Carlos et al were pressured to do this but knew that, at that stage of development, the results weren't impressive. Instead they wanted synths to have their own "voices" as created by the user and continued to move along that path.
When we hear "samplers" we intuitively make comparisons with sounds from our own experience and with the labels we already have.
These days a Mellotron track will get comments about how it sounds like a cheap, noisy string sample whereas it used to be a sound that wonderfully replicated things - to most ears. These days I hear Mellotron and rather than comparing it to the samples I think of the classic recordings made using it and note how closely those are replicated.
Comparison, like judgement,is an evolutionary survival tool.

Cheers
rayc
"What's so funny about peace, love & understanding?" - N.Lowe