Born probabilities vs. the universal prior — LessWrong