If alignment problem was unsolvable, would that avoid doom? — LessWrong