Also, how do you continue to train a model like this when everywhere you look is just output from the same model? Like, at some point, the real conversations an

Also, how do you continue to train a model like this when everywhere you look is just output from the same model? Like, at some point, the real conversations and text gets overwhelmed by the fake stuff. Where does the model pull more training data from?