The only snippet with a statistically significant majority preference went to the AI.
Shocking to me as AI writing always seems palpably inferior.
numeri•16m 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_Bruning•24m ago
One crazy part of this is how Opus 4.5 is pretty old by now, for an LLM!
theopsimist•1h ago
numeri•16m ago
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.