Naomi Clark (@metasynthie) 暗悪直美 @halfrobot.com · May 22 There's a certain kind of "creative" whose ambition lures them to efficiency traps that create more work, which is then often outsourced to... an intern? A LLM? Look, I favor harm-reduction at this point. But you shouldn't use LLMs if you don't understand why they will foul up text extraction!
17 likes 3 replies
Replies
the Ars Technica debacle where a techie beat reporter used AI to pull quotes from a blog entry (which it fabricated instead) about AI was really instructive. Even people who allegedly understand this shit _as a job_ and write about it every day still fall into the trap. futurism.com/artificial-i...
Don't argue with a LLM! The LLM didn't make transcription errors "because it wanted to make your language stronger!" That's a retroactive explanation of an unpredictable token-probability process! You are not conversing with a state-based mind, only a pile of context. Also, THIS takes seconds
Nobel laureate and esteemed journalist Maria Resse wrote the forward to the book and the following: "He asks us to think about how truth can be protected and rebuilt at a time when lies are rewarded." I wonder if she'll reward Rosenbaum with allowing her forward in future editions!