Robert Morris has a very unusual quality: he’s never wrong. It might seem
this would require you to be omniscient, but actually it’s surprisingly easy.
Don’t say anything unless you’re fairly sure of it. If you’re not omniscient,
you just don’t end up saying much.
[....] He’s not just generally correct, but also correct about how correct
he is.
Being well-calibrated is great, but it sounds like rtm isn’t even wrong in retrospect. I much prefer to say wrong things very loudly so that I will discover when I am in error.
-- Paul Graham
Being well-calibrated is great, but it sounds like rtm isn’t even wrong in retrospect. I much prefer to say wrong things very loudly so that I will discover when I am in error.
Calibration is awesome. However, note that without an audience like the NSA or Paul Graham, this is probably sub-optimal signaling.
I have to say, I haven’t found calibration hugely useful. It’s certainly nice, but for the most part people ignore you.
Does it give you better answers, though?
Sure, but I find that most of what I do is not dependent on small probability increments.