I think knowing what you don't know is still a pretty big win. It can help people trust the models in cases they do work, and it can serve as a diagnostic for w

I think knowing what you don't know is still a pretty big win. It can help people trust the models in cases they do work, and it can serve as a diagnostic for why it fails in certain circumstances.