Petropolitan comments on Alignment Faking in Large Language Models

[Petropolitan]

I think LGS proposed a much simpler explanation in terms of an assistant simulacrum inside a token-predicting shoggoth