Critch on career advice for junior AI-x-risk-concerned researchers — LessWrong