People are chattering are how bad it's going to be as #ChatGPT starts ingesting it's own AI-generated output as "truth".
But we are already living in AI feedback loops, like when we listen to #Spotify's recommended tracks, and it uses that new listening behavior to make the next recommendation.
It doesn't get better and better over time, it gets narrower and narrower. And to the extent it mixes in "what's popular", some flukes are getting highly recommended while some good music is missed.