Probability theory implies Occam's razor — LessWrong