Follow
LLM AIs are basically always trying to flimflam you.
You should model them as pathological liars, who will tell you lies even when the truth would be more convincing.
LLM AIs are basically always trying to flimflam you.
You should model them as pathological liars, who will tell you lies even when the truth would be more convincing.
Note that the LLM is not actually trying to flimflam you. It doesn't have that kind of interior planning.
But if you're going to anthropomorphise it, you should picture it as Don Draper crossed with Walter White.