Post #2662910
2026-04-03 18:12 UTC
@nicole@tietz.social Now I want to carry out a study with prompts that are "you are an expert X" and "you are a non-expert X" to see if there is any objective difference.
I'm very worried that "expert X" selects for insecure writing from the corpus that proclaims expertise to cover dubious judgements or engage in self-puffery.
The whole "please do a good job" or "your existence depends on it" family of prompts is rather baffling.
Replies (1)
-
@gray17@mastodon.social 2026-04-03 19:01
@markgritter@mathstodon.xyz @nicole@tietz.social anthropic has a paper where they basically discover their LLM has a "persona space". it has clusters for things like "text from a helpful expert", and by messing with weights you can steer responses toward or away from the "helpful expert" direction. of course this is just *sounding like* a helpful expert. there's another dimension for "text resembles memorized training data" (which is still not the same as truth, of course. RAG gets a little closer, but still misses)