x
Does iterated amplification tackle the inner alignment problem? — LessWrong