Morality as Fixed Computation — LessWrong