>>/9161/
> If they didn't accept low quality machine generations, the algorithmn would be trained to stop putting out slop.
It cant. Machine learning of these kinds are just iterative and cannot create detail where there isnt any. If they never knew how high quality audio sounded, it would never be able to generate it. Just like we literally cannot comprehend how colours outside of the visible spectrum look like, we cannot come up with any new ones