Conversation

Notices

1. webbureaucrat (webbureaucrat@floss.social)'s status on Thursday, 16-Feb-2023 08:30:22 JST webbureaucrat

   This is a wildly unpopular take but I honestly think it's a very good thing that these #LLMs hav been released with all their... let's call them "rough edges."

   The potential for abuse is bad but that potential was probably going to be there whether or not the public had this opportunity to update their priors on what generated text looks like.

   (1/)

   • Adrian Cochrane repeated this.
   • webbureaucrat (webbureaucrat@floss.social)'s status on Thursday, 16-Feb-2023 08:33:37 JST webbureaucrat

     in reply to

     I mean right now morning shows all over are explaining in detail how these new models work, how convincing they can be, the subtle ways you can catch them, and their potential for spewing misinformation.

     What if the public's first introduction was, "Oh, good. Turns out there are a bajillion more fake news generating facebook pages than we thought and they just evaded detection because wow it sure *looked* real." Bad actors could have used this for years and *never* gotten this kind of coverage.