4 pointsby theopsimist5 hours ago3 comments

theopsimist5 hours ago
The only snippet with a statistically significant majority preference went to the AI. Shocking to me as AI writing always seems palpably inferior.
- numeri5 hours ago
  AI's learned their style through RLHF, with semi-focused, partially motivated humans giving feedback on short(ish)-form content.
  For the most part, Modelese is the revealed preference of the average, not-heavily-invested human. This doesn't fully explain the Claudish coming from Opus 5 and Fable (this seems like it may be due to excessive RLVR or RLAI), but yeah.
  It's got a lot of cheap writing tricks that make people think it's smart and helpful.
Kim_Bruning5 hours ago
One crazy part of this is how Opus 4.5 is pretty old by now, for an LLM!
codevark5 hours ago
[dead]