I’m mildly curious if those additional tokens will make AI better, or worse. We arguably trained AI on the good stuff first. Novels. Wikipedia entries. GitH

I’m mildly curious if those additional tokens will make AI better, or worse. We arguably trained AI on the good stuff first. Novels. Wikipedia entries. GitHub open-source projects with a lot of stars. What’s left but mediocrity and our “baser” internet ramblings? Some researchers already found out that AI-sourced content can affect the models, but what about content from increasingly out-of-touch people?