Comment by eggplantemoji69

Perhaps RSI by virtue of feeding models their own slop as training data exponentially decays the ‘quality’ of the model?