This is becoming a common but not enough interrogated buzzphrase for a very real problem - LLM output is verification and technical debt. One person can generate plausible enough looking results that they can pass through several hands before a close enough reading shows it's full of shit.
Mallory Moore
This! Someone sent me a plan for a research project that they'd got chatgpt to write, and it was 3x longer than what I usually get and honestly I was like "why should I spend my time reading this when you haven't spent yours writing it, you may not even have read it yourself"?
Kat Steiner🔸
I think one of the things that is probably the main factor in this is that LLM agents generate *a lot* of text and reading/parsing that much text all day if you're really serious about checking and validating things is exhausting.
