AI generated audio would be weird sure, it would be amazing to see AI struggle with making audio themselves or For that example Text to speech or Voice to speech that i thought of. So Basically the AI would acquire a audio of yours or generally any audio of a youtube video, music, etc, then that ai would try to make the music, video, etc, by trying to guess what their saying, what sort of genre of music or generally anything.
I doubt that Audacity will be developing AI generated audio. “Only” making a full featured audio editor is a huge job in itself.
You may find this site interesting: https://magenta.tensorflow.org/