Robin Hanson Queries LLMs To Rank His Political Positions At 99.8% Weird

@robinhanson Imo LLMs have a bias against extreme values

@robinhanson Right I meant maybe the correct answer is even higher, say 99.9-100.0%

@robinhanson Not that it matters yet, if ever, but ChatGPT (Plus/5.5/High), who happens to know me the best right now by far among the LLMs, thinks I'm overall a little weirder (but that you're more "unsettling," which may be the better quality here).

@StefanFSchubert Making it all the more unusual if they overcame that when rating me.

@robinhanson I wonder if you are more weird or less weird than the version of you that would correctly say “weirder than I?”

@HyperFairies If I'm one out of 200 to 500, then there are >16M people out there weirder than I.