Does anyone have pointers for what happens if in a few years a lot of the online text is generated by ChatGPT, and this finds its way into the training set? Does it get “more ChatGPT-ish”?
Does anyone have pointers for what happens if in a few years a lot of the online text is generated by ChatGPT, and this finds its way into the training set? Does it get “more ChatGPT-ish”?