Another view of quantilizers: avoiding Goodhart's Law — LessWrong