@jasminewsun I personally think about Fuchs' alternative conception a lot (though it is probably not a _defense_ per se): I also really like this history: And this from monthly review is interesting:

@no_earthquake Grade inflation is crazy

@FawkesJake @bayesianboy @zacharylipton A postscript: there are a lot of technical ideas under the hood here, using tricks from the causal inference literature to formulate sensitivity analysis as a non-convex optimization problem. One exciting direction is unpacking when/why the resulting problems are easy to solve.

I’m at NeurIPS! If you’re in Vancouver and want to catch up let me know! I’ve been working at the intersection of generative models and causal inference, focused on bio/healthcare. If anything from active learning to causal foundation models sound interesting I’d love to chat!

RT @tianyuf: good to know that future drone strikes on children will be done with safe, trustworthy, interpretable AI

@kevinbaker i will say this is pretty correct, people do in fact just target the polling averages in various ways

@ChelseaParlett I think the only reasonable thing to do in any applied circumstance is write down a plausible data generating process and simulate data for various n. You can figure out when it starts to look like the asymptote predicts. And then you can misspecify things and check again.

RT @arthur_spirling: Our (@cbarrie @LexiPalmer_) working paper on replication for LLMs is now available here, with a non-technical explaine…

@Jacobkupp @olzhu i am sorry to say polls are fake. the directions of the errors are unpredictable. the magnitude of the error is far too large to predict elections which depend on like 20k people in a few key swing states.

@xuanalogue We tried to get at this (among other points) in this FAccT paper:

big week for ai for science

@anshulkundaje Is the stewing usually between drafting and preprinting or between preprint and submitting to journals?

@MarkSchmidtUBC yep what i'm always interested in when I read something like this is: does this seem like a path to theory where we will need less assumptions or not. i think some nn theory is insufficiently risky! new ideas!