Principles for the AGI Race — LessWrong