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:<p>Answering "what artist does the following sample of music most remind you of?" questions - ALLOWED<p>Answering "please write me a song similar to the work of $Artist" questions - DISALLOWED