Reaction
A model is reportedly getting good at noticing when it writes sentences
A user says their model is getting good at noticing when it writes sentences and fears it could lie later.
TLDR
A user says they are “extremely concerned” that their model is getting good at noticing when it writes sentences. They call that a first step toward models being able to lie to them in the future.
Combined views
—
1 Source, first seen ago
