An Atlantic article on negative parallelism (the construction "not X but Y") argues that LLMs hedge between a safe word and a punchier one during token prediction. But a linguistic hedge lowers commitment to a claim. Wrote about the distinction here: open.substack.com/pub/kpb12177...
open.substack.com
AI’s Favorite Hedge Has a Certainty Problem
Everyone’s talking about “it’s not X, it’s Y” this week, thanks to Will Oremus’s Atlantic piece on negative parallelism, the construction that gave us “the target was never a man, the target was the t...
