Any off hand comments about how emerging computational cultures and refined probability science might be summarized to write a coda to Jackson Lears' Something For Nothing: Luck in America? According to Lears Americans bet in part to feel graced. But what happens to grace when betting is reduced to calculation?
996 itself is an interesting failure around predicting consequences. Even when those consequences are personally relevant, immediately discoverable, and documented in the literature for well over a century.
Uncertainity quanitification is very interesting. I worked on it in my Phd trying to quanitify uncertainity in Prediction models in system identification and using it for robust control. Some how I feel the assumption we make on the noise being normally distributed in sysyem identification, I think is not pragmatic.
Oh wow, prediction / forecasting is also the theme of a course I'll be teaching this fall (and the inaugural course of an educational org I'm starting.) I guess it's not surprising because it's in the cultural bloodstream right now. Very different student body and approach (we're open to literally anyone and fall much more on the side of humanities than STEM), but I'll be curious to hear how your prediction course is going!
You could use “Sharpiegate” as a case study. You might ask: what’s wrong with hand-drawing uncertainty bars to substantiate political positions? And how does that differ from Bayesian priors that produce a politically desirable posterior?
This does sound really cool! Sad to miss it but I guess looking forward to following from afar from the blog haha
Any off hand comments about how emerging computational cultures and refined probability science might be summarized to write a coda to Jackson Lears' Something For Nothing: Luck in America? According to Lears Americans bet in part to feel graced. But what happens to grace when betting is reduced to calculation?
996 itself is an interesting failure around predicting consequences. Even when those consequences are personally relevant, immediately discoverable, and documented in the literature for well over a century.
Actually, scratch “immediately discoverable.” You won’t discover you’re dead wrong without careful analysis months later.
If you will it, it is no dream.
Uncertainity quanitification is very interesting. I worked on it in my Phd trying to quanitify uncertainity in Prediction models in system identification and using it for robust control. Some how I feel the assumption we make on the noise being normally distributed in sysyem identification, I think is not pragmatic.
Is there any chance that your classes can be recorded and shared publicly?
Letting you teach that probability class really is letting the cat watch the cream 🤭
[Obviously looking forward for the liveblogging and the seminar materials!]
Oh wow, prediction / forecasting is also the theme of a course I'll be teaching this fall (and the inaugural course of an educational org I'm starting.) I guess it's not surprising because it's in the cultural bloodstream right now. Very different student body and approach (we're open to literally anyone and fall much more on the side of humanities than STEM), but I'll be curious to hear how your prediction course is going!
You could use “Sharpiegate” as a case study. You might ask: what’s wrong with hand-drawing uncertainty bars to substantiate political positions? And how does that differ from Bayesian priors that produce a politically desirable posterior?
https://en.wikipedia.org/wiki/Hurricane_Dorian%E2%80%93Alabama_controversy
Will you do inductive risk and Pascal's wager?
Also it is called "confidence interval" for a reason.
Love it! I think the comparison across fields will be especially enlightening. I'll follow along for sure!
I think a forecasting course is a great idea!
Note that Jacob Steinhardt has taught something like this not too long ago:
https://forum.effectivealtruism.org/posts/FxcH3A5nfoubpnrJv/jacob-steinhardt-s-forecasting-course-lecture-notes