The weird, rare, surprising patterns that make data rich slowly get smoothed out when an AI model trains on outputs from a previous model. Read the full article in Communications of the ACM.
Model Collapse Is Already Happening, We Just Pretend It Isn’t
