Musical notes in a musical phrase are like words. Because as language music is meaning and signification in a context.
Then music MUST be recognised by the listener , and accepted or refused, as a wrong note or the good note , not because the note is not sympathic but because the note cannot be accepted as a possibility among all notes possibles at this specific place in the melodic/harmonic tonal places of the melody playing at this moment . The listener must be surprized but he must not be shocked at each moment and derailed from any known tonal landmarks .
When we cannot recognize any note as more possible and more likely possible than any other notes then no surprize can exist... When all is imprevisible there is no more any surprize. this is the case in atonal composition.our feeling cannot guide us in the interpretation then in the meaning recognition.
In tonal history each notes is used at some specific moment with some probabilities. It all these probabilities are equal there is no surprize and no more any innovative idea. Just a soup of atonal note nobody can interpret with some specific feelings succeeding some other specific feelings...
We listen music with the potential bundle of feelings rooted in our body metabolism as well as in our musical history ...
Now suppose A. I. composing music... A.I. can imitate known styles more or less skillfully.But A. I. is not an embodied player with innate and learned feelings and emotions history, he will not be able to be creative and be able to move us as a human composer or musician can do .
We cannot fake emotions using sounds only to some limited extent . The imitation will emerge more soon than late.
Then musicians will be able to use A.I. as a tool for sure. But A.I. cannot replace human body playing an instrument and improvising emotions nor imitate human mind writing moving music which will evoke very complex set of emotions mixed harmoniously together.
Hope and despair for example united together in a single melody with a specific succession of rythms is not the result of a logical computation even if you mix all hope melodies together and all despair melody together and try to algorythmically to assemble a beautiful new melody combining the two emotions with the right rythm ... Imagine now playing all the hues of this composed colored melody with two very different emotions inside ?