The problem isn't the generation, it's the taste of the generators.

An earnest young lady with a guitar can already sing a light jazz version of 'Highway to Hell' or whatever. Just go to your local cafe to hear it. The objective quality is terrific.

In the past, this wouldn't have been made because the end result is subjectively banal. But now people with no taste can churn it out by the thousands of hours for free.

When they say "worse" they do mean the AI will get better which will be worse because they are ideologically opposed to AI.