andrew 🇺🇦

@theoneandmoely @saltwatersoph Boring stuff. Math is fun and useful

@saltwatersoph You can work in finance or financial services as an econ grad without needing to do quant stuff.

I would go back and tell 17 year old me that she should actually just try to become a quant

@flieldy Does Andrew Jackson Jihad count?

Still not as bad as the celebrity-name-pun band name thing happening to this day

@constans Tbf it’s not fully free tuition, it’s partial. But still it’s crazy

It blows my mind that the country whose capital is one of the world’s major financial centers finds an upper middle class American salary to be beyond the imagination

@stathx @mouse_math What why

@mouse_math measure theory and functional analysis

what's the most annoying field of mathematics?

@aashiq @mouse_math So so so wrong. C.f. probability theory?

@mouse_math Measure Theory. Tedious, not much benefit over everyday intuition except in exotic corner cases, limited application in other branches of math.

@watling_samuel What does bayesian micro theory look like?

Econometrics is good autism, micro theory was good autism until Bayesian persuasion then it became bad autism. IO is normal autism. Econ History will be the most autistic when the methodology catches up. Politica is still a bit too relevant. The rest are more ADHD-coded.

Due to my autistic desire to rank everything here is my list of Econ subfields ranked by autism level: 1. Econometrics 2. Pure Micro Theory 3. IO/Trade 4. Econ History 5. Political Economy 6. Applied Micro 7. Macro 8. Development 9. Behavioural

@Carboniferoys New pick up of mine. Probs dropped average IQ of this book’s readers by 15+

When ever I see one of these I do wonder what actual books would have a 130+ average IQ reader. I'd imagine the really obscure expensive and relatively new graduate level textbooks in Math/Physics might be basically it.

@tomsteyerstan We just have to move away from single representative districts and embrace proportional representation (or some other alternative system)

let’s check in on the state of the gerrymandering war

@theomitsa @MaxDiffusionRL What that’s completely false, plenty of people have attempted to create a theory of deep learning

@MaxDiffusionRL First-ever attempt to create a theory of deep learning. Nobody really understands how NNs work, because we do not have a working theory for them.

arxiv.org/abs/2605.01172 This amazing paper is from a Stanford UNDERGRAD!

@souljagoyteller I would like to know too, i’m interning in DC

DC residents: where do you guys go out for the night? Gonna be here in the summer for the first time, so I’m trying to learn the tricks of the trade

@Lib_Development Miller lite pitcher is $12 at jimmy’s, which is like 4 drinks

It’s also just outrageously expensive. Like including tax and tip, “one beer” ~= $10. If you want to get drunk, you’re looking at at least $50. Probably closer to $100 if you want to drink something that isn’t undrinkably bad. Per head! Do you know how many tokens that buys

@ferocitq That’s kinda the whole point

started watching invincible and this must be the most auraless sauceless introduction to a super group ive ever seen

Streets is saying they uncovered a holy portal in Dagestan but the Russians are covering it up

@arnie_hacker hahaha i've been using colab for a couple of weeks and it's the worst thing ever when you have a server die in the middle of hte night

Are ML researchers constantly sleep deprived because they're trying to submit all their jobs before they sleep

@kiyodia The whole comic fire and shadow is about the fire nation not rocking with zuni

oh…so the fire nation is not rocking w fire lord zuko i see

Imagine going summer time driving with buckshot and fake mink

@avt_im @RobertoDailey1 Do you think other Bayesian Nonparametric methods could fix that

@RobertoDailey1 For time series, GPs have been scalable since long before any of the works I've mentioned - via state-space methods which leverage higher-order Markovianity. Like in general, I expect the thing that limits performance to be expressive capacity, caused by Gaussianity itself.

One of my formative academic experiences: - Wrote a paper that got GPs to scale-up to 10m+ observations without sacrificing quality of uncertainty - Significant jump in GP capabilities - Almost nobody cared I should have moved on from this space much sooner than I did.

@HoodlumDoodlum @SwannMarcus89 this is posted on arxiv, not a journal

@SwannMarcus89 They can’t be wrong, though: peer review!

I hate these fucking people Neither author of this article is an economist. One is a computer and information science professor and one is an MBA It’s a coding and business major predicting the macro impact of AI on the economy decades from now

@igetsoinspired Hell yeah, with a nice runny egg and some hot sauce

These save lives

@__paleologo Pac bayes is not sufficient, unfortunately

Reading a paper on why Deep Learning works. Short answer: PAC-Bayes.