Electronics360: "AI algorithm can create music that sounds like a live performance"

A probably ridiculous question comes to mind... Is it possible the visual cues include, if not rely upon the body language of the performing artist, exhibited during the piece?

It seems a minimal analysis of keystrokes of any short, even incongruent sequence of keys seeing just 1 or 2 fingers would easily identify the composition being performed.

I must be missing or misunderstanding something.