Comment by dcrazy

You cannot rely on the model to never make mistakes. You need deterministic controls around the model. I believe that’s what the article’s authors are selling.