Greg C comments on Embedded Interactive Predictions on LessWrong