Let's talk about Impostor syndrome in AI safety — LessWrong