Dealing with infinite entropy — LessWrong