I believe this is different. People are calling it AI in this instance, but all they really did was isolate the vocals and removed the TV playing in the background from the original recording. Basically, using machine learning and 'stems' which is found in many DJ and other programs around, where you can now isolate instruments, vocals and split apart a track. So they didn't use AI in the same way many people think of when they say AI. More machine learning, as it was just splitting the elements of the track, not creating anything or modifying anything.