tobychev @tobychev@social.cleverlibre.org
Not sure what this account will be about, mostly boosting things I find interesting.
Joined Nov 2022
tobychev
boosted
Vad är det DN skriver egentligen??
"Den svenska avskyn för ryssen har djupa rötter. Ibland kan man undra varför – historiskt har Ryssland trots allt inte förorsakat Sverige så djupt lidande. Lars Linder läser Dick Harrisons nya bok om den väldiga och opålitliga grannen i öster."
Byt ut "ryssen" mot "juden". Funkar bra?
https://www.dn.se/kultur/den-larde-dick-harrison-har-haft-lite-for-brattom-i-fienden/
@cygnus_ww
Yeah, visa applications can suck tremendously, and I expect this is in no small part because the entire enterprise of migration control is pretty sucky at its core: coming up with reason to deny foreigners from tainting the motherland...
As an aside, I wonder if the visa services in the EU have deteriorated after Schengen was enacted and made visa applicants more purely into "foreigners" as most neighbours now don't need to trouble with it?
tobychev
boosted
An interesting as subtle bias in our image generators: the American smile
https://medium.com/@socialcreature/ai-and-the-american-smile-76d23a0fbfaf
@brendannyhan
Alternative title "people don't know what's good for them" :(
Would be interesting to extend the test with a feed containing corrections to several media outlets to remind subjects that actual errors aren't as rare as the public corrections.
Personally I find public corrections a mark of competence; when a news weekly I subscribed to got a new editor that proudly proclaimed he would publish no trigger warnings (none has been published previously) while also cutting the modes corrections section I stopped subscribing.
@niemanlab @dangillmor
tobychev
boosted
New @niemanlab by study coauthor @dangillmor: "The corrections dilemma: Admitting your mistakes increases accuracy but reduces audience trust, a new study finds" https://www.niemanlab.org/2023/03/the-corrections-dilemma-admitting-your-mistakes-increases-accuracy-but-reduces-audience-trust-a-new-study-finds/
tobychev
boosted
one aspect of diffusion-based generative models which I find kind of funny / bittersweet is that
i) the optimisation problem is convex in terms of the { score / denoiser / ... }, and is even a least-squares problem in some formulations. this means that it would be extremely well-adapted to the sorts of statistical theory which people loved 10-15 years ago, but
ii) these models have become { popular, feasible, etc. } in the neural network era (and even as a direct result of certain NN architectures?), and the aforementioned favourable convexity properties are largely absent in weight space, and thus don't obviously yield any insights about the recent practical successes.
@AstroMikeHudson
A friend got into trouble because plagiarism software discovered he was reusing stuff he'd written in earlier work, only the software said the source was by a colleague who in reality had stolen the text from my friend.
tobychev
boosted
The 2023 edition of my long-running anti-establishment art-science-fusion code-therapy smooth-baritone causal inference & Bayesian data analysis course is complete. 20 lectures, from the basics of causal inference & Bayesian updating to mixed models & Gaussian processes.
Lecture recordings, slides, homework sets and solutions are all listed here. Take it at your own pace. First half is a solid course in regression and causal inference. Second half turns it up to 11. https://github.com/rmcelreath/stat_rethinking_2023#calendar--topical-outline
tobychev
boosted
Nonprofits delenda est.
---
RT @JakeMGrumbach
Leave it to state legislatures to make Congress look like paragons of virtue. Truly wild story of corruption in the Ohio state legislature in the post-Citizens United era.
https://grist.org/politics/ohio-bribery-scandal-larry-householder-verdict/
h/t Pierson
https://twitter.com/JakeMGrumbach/status/1635337460262764544
tobychev
boosted
Very provocative article by Animah Kosai on Corruption Perception Index.
We so easily forget that the ones who corrupt are the ones who pay bribes, not the ones who receive them.
Corruption and bribes are mechanisms at the disposal of the rich and powerful to tilt the playing field in their favour and away from the lesser rich and powerful, so who are the ones to blame?
https://www.linkedin.com/pulse/who-corrupt-really-animah-kosai
@anne_kreft
Oh, I thought you were looking for academics? ;)
tobychev
boosted
Yet another sign that the footprint of the death penalty is shrinking. I thank Gov. Shapiro of Pennsylvania.
Here’s the first part of his statement today:
Today, I'm announcing I will not issue any execution warrants during my term as Governor.
When one comes to my desk, I will sign a reprieve every time — and I’m asking the General Assembly to send me a bill abolishing the death penalty in Pennsylvania once and for all…
@anne_kreft
At first it seemed like fun branding to have the deadline on intenasjonale kvinedagen, but then I started to wonder if it isn't more likely to annoy much of the intended audience that has other plans that day...
@j_bertolotti
Oops, should have been "must *not*" but I leave it as is.
@j_bertolotti
While the importance of individual responsibility is undeniable, we must neglected addressing structural factors! :p
@j_bertolotti
But then people would fall asleep at meetings more!
tobychev
boosted
I just had whatever the opposite of 'future shock' is - I stumbled upon a telescope proposal from 2001 and I wondered if the #latex would compile - of course it did, and the proposal looked exactly as it should have.
In related news, I wrote my thesis in Word 98 and the next version refused to open one of my chapters for no obvious reason.
Plain text and openly documented image formats are definitely the way to future proof your work....
@ProfHughLewis
how many collisions would we expect by now you treat all encounters as iid an assuming no avoidance maneuver?
Not sure what this account will be about, mostly boosting things I find interesting.
Joined Nov 2022
