Arpit Gupta

66K posts

Arpit Gupta banner
Arpit Gupta

Arpit Gupta

@arpitrage

Associate Professor of Finance, @NYUStern. Finance/Real Estate/Urban newsletter: https://t.co/lWVJvq6vKg.

Manhattan, NY Katılım Eylül 2015
5K Takip Edilen47.3K Takipçiler
Arpit Gupta
Arpit Gupta@arpitrage·
@weakinstrument *Odysseus taunts Polyphemus after blinding him, sacks Ismarus, kills the serving women* Wow, no one cares about the principle of Xenia anymore; whatever happened to taking care to strangers
English
1
0
1
280
Arpit Gupta
Arpit Gupta@arpitrage·
@weakinstrument Greeks have a whole lot of nerve complaining about the “Sea Peoples” when THEY are taking to the sea and pillaging
English
1
0
5
854
Ryan Cummings
Ryan Cummings@weakinstrument·
Folks PLEASE no spoilers on Nolan's new original screenplay about the ancient greeks 🤫🤫🤫
English
1
0
14
1.3K
Arpit Gupta retweetledi
Simon Jäger
Simon Jäger@simon_jaeger·
New @nberpubs paper w/ @PawelAdrjan, @BalgovaMaria, @jonasjessen & @JSockin: Workers infer unlisted job characteristics from what ads do list. Using a priced amenity—the 🇩🇪Deutschlandticket—we show this contaminates discrete choice estimates of WTP for amenities. 1/N
NBER@nberpubs

Job ads can signal hidden job traits. Evidence from German workers shows this distorts discrete choice estimates of workers’ willingness to pay, from Pawel Adrjan, Mária Balgova, @simon_jaeger, Jonas Jessen, and @JSockin nber.org/papers/w35457

English
3
9
30
6.1K
Daniel Stephen
Daniel Stephen@danstevejon1·
@arpitrage @IrvingSwisher Yeah but data centers are more akin to an extraction business. it’s so hard to mine in the first world precisely because mining companies were famously bad at true stakeholder management. For most random factories the worst externality is like traffic backups bc of the workers
English
1
0
0
14
Skanda Amarnath
Skanda Amarnath@IrvingSwisher·
I thought Hochul did a pretty admirable job of explaining her policy on Odd Lots, and yet… I also think it’s hard to buy into the logic of a moratorium if youre not also throwing everything at the wall to improve the grid. She is doing some very impressive things on nuclear (🫡 ), but is also leaving the highest ROI project on the table b/c of the (real) power of Westchester NIMBYs. Pragmatic incrementalism is good; make progress where you can. But at some point, powerful legacy interests should be confronted
Emmet Penney@nukebarbarian

.@GovKathyHochul cited skyrocketing bills as an inspiration for NY's new data center ban. Perhaps she was relying on @Bloomberg's analysis that data centers spiked wholesale prices up to 267%. But Bloomberg's wrong--closing Indian Point spiked prices. 🔗⏬

English
8
9
82
24.9K
Arpit Gupta
Arpit Gupta@arpitrage·
@danstevejon1 @IrvingSwisher A quibble, but neither are really externalities. They are just paying for goods which pushes up the price. In most sectors, we don’t do anything about that. If a car manufacturer or office building uses energy or chips, that’s not a broader social concern.
English
1
0
1
21
Daniel Stephen
Daniel Stephen@danstevejon1·
@arpitrage @IrvingSwisher We’ve been insuring data center financings for 5 years. This is most definitely not how it has played out. This industry took far too long to realize (more like to care about) the externalities which are unique to it, as far as power and chips
English
1
0
1
16
Arpit Gupta retweetledi
Matthew E. Kahn
Matthew E. Kahn@mattkahn1966·
I agree with the point that Chris is making but this huge forest offers the opportunity for A/B testing. Use Spatial Big Data methods to target Canadian Forests for different randomized treatments and let's learn how to cost effectively reduce PM2.5 production. a thread.
Chris Martz@ChrisMartzWX

I am seeing a lot of posts from people claiming that Canadians should adopt better forest management practices to reduce forest fire incidents and prevent blazes like the ones in Ontario from getting out of control and blowing smoke downwind, causing transient reductions in air quality. While it is true that regular mechanical thinning and prescribed burning act to reduce forest overgrowth and clear out highly flammable small-diameter fuels like leaves and twigs, the vast majority of Canadian forests are isolated from the more urbanized Quebec City–Windsor Corridor, meaning they are very difficult to access to have forest management on the regular. Now, you can strongly argue that better forest management practices would help (and I also did the other day), but due to the geography of Canada and the population distribution, it’s not an easily feasible solution like it is in the U.S.

English
1
1
10
2.5K
Arpit Gupta
Arpit Gupta@arpitrage·
@danstevejon1 @IrvingSwisher I mean, this is how most firms just work. You buy some land, pay for inputs as necessary, the usual taxes, and do what you want. It’s the lack of “good jobs” that makes data centers seem unusual
English
1
0
0
40
Daniel Stephen
Daniel Stephen@danstevejon1·
@arpitrage @IrvingSwisher Unfortunately this is what happens when private industry runs roughshod over people and loses all the goodwill. Coming from a mining background, 2nd and 3rd world, I always found that the S was the most important element of esg. You treat the S right, the e and g follow
English
1
0
2
38
Max Kagan
Max Kagan@max_kagan·
@arpitrage Changing to a large lab STEM-type model of research because of fixed cost overhead would be a complete shift social science If you need a GPU cluster that can run a trillion+ param open weight model the days of a grad student being able to solo author a paper are done
English
1
0
1
163
Arpit Gupta
Arpit Gupta@arpitrage·
I think we're already at the point where academics should default to using open source models for research Good enough for many tasks, cheaper, and we get replicability (instead of relying on models that will get deprecated)
English
6
9
104
13.4K
Paul Novosad
Paul Novosad@paulnovosad·
@arpitrage The whole universe has been converted to computronium, there's nothing interesting left to do but simulate experiences from the time when experiences were worth having
English
1
0
1
108
Arpit Gupta
Arpit Gupta@arpitrage·
If you believe the Copernican Principle, it's weird to be conscious right now unless you are a doomer With a dramatic increase in machine intelligence, extended human lifetimes, or humanity through the stars; the odds of being intelligent *now* become very low
English
5
2
16
4.1K
Arpit Gupta
Arpit Gupta@arpitrage·
@max_kagan Research will get concentrated in AI labs, US R1s, and anyone else with $
English
1
0
0
183
Max Kagan
Max Kagan@max_kagan·
@arpitrage I support that plan. But what about people who aren’t professors at a top US R1?
English
1
0
0
160
Max Kagan
Max Kagan@max_kagan·
@arpitrage How do you propose academics get access to hardware to run those, though? I can scrape together $200/mo for closed-source models that work whenever I need, I cannot afford to buy GPUs, and the slurm queue for shared GPUs runs at least several weeks.
English
3
0
3
739
Arpit Gupta
Arpit Gupta@arpitrage·
@UESModerate Run your task through an expensive model, then test a bunch of free models and pick the one which correlates best
English
0
0
1
591

Keşfet

@Susan_Athey @guido_imbens @weakinstrument @nberpubs @PawelAdrjan @BalgovaMaria @jonasjessen @JSockin