Music of Language: Human Versus AI

A comparison between how I parse a lyric and how AI does it. I've noticed that AI frequently drops either the first eighth or quarter at the beginning of a bar on the vocals, or the vocal is very rubato and floats above the groove and ignores bar lines or accents. This is because it's squeezing the vocal stems to fit the riff stems. This is not how music is typically written or how we sing. The "I've said" line is the same in both but AI displaces it 2 beats, so it is picking up on human speech rhythms.

I've said
Never say never
I've said
Never say never
I've said
Never say never
Many times before
Many times before  

The AI version: