Not that trust Big AI one bit...but it occurs to me that such training could be handled correctly (meaning ethically and mostly-legally) if the AI could gatekeep its own use of the derived results:
Answering "what artist does the following sample of music most remind you of?" questions - ALLOWED
Answering "please write me a song similar to the work of $Artist" questions - DISALLOWED
Not that trust Big AI one bit...but it occurs to me that such training could be handled correctly (meaning ethically and mostly-legally) if the AI could gatekeep its own use of the derived results:
Answering "what artist does the following sample of music most remind you of?" questions - ALLOWED
Answering "please write me a song similar to the work of $Artist" questions - DISALLOWED