x
Learning the smooth prior — LessWrong