Forecasting: the way I think about it — LessWrong