Logo
Picture

Hossein Alidaee @halidaee

Economics postdoc researching info frictions + ext validity in technology adoption at @HarvardHBS. PhD @NorthwesternU. All opinions are wrong, some are useful. halidaee.com Joined November 2008
  • Tweets
    726
  • Followers
    195
  • Following
    167
  • Likes
    1K
halidaee Profile Picture

Hossein Alidaee @halidaee

a week ago

@Afinetheorem Let alone importance of synthesis of how these different developments interacted to form mainstream econ understanding of the market and its trade-offs.

halidaee Profile Picture

Hossein Alidaee @halidaee

a week ago

@Afinetheorem I do think its partly on us for staying out of this literature. I tried to recc reading to humanities colleague when they were earnestly interested. There is no obvious non-technical intro to e.g. hurwicz, relationship between arrow imp theorem and central planning, etc..

halidaee Profile Picture

Hossein Alidaee @halidaee

a week ago

Increasingly confident @thsottiaux is actually just an avatar for elasticity testing Codex credits.

thsottiaux Profile Picture

Tibo @thsottiaux

a week ago

@rxmphai I'll do another performative reset on Monday

halidaee Profile Picture

Hossein Alidaee @halidaee

2 weeks ago

@itaisher Specifically, the same human was making similar decisions before AI joined the loop.

halidaee Profile Picture

Hossein Alidaee @halidaee

2 weeks ago

@itaisher Without taking a stance on all this, I think it's interesting how most of the Refine issues are ultimately issues with the human in the loop using it.

halidaee Profile Picture

Hossein Alidaee @halidaee

2 weeks ago

Sure, AI can do fancy math. But more importantly, it can monitor your Slurm queues for you.

halidaee Profile Picture

Hossein Alidaee @halidaee

2 weeks ago

@RefineDotInk @cmatthes_econ Scorecard comparing these metrics vs "open source" tools (which often use closed model weights anyways) would be really nice. If this can get tamed, I think it also makes one of @ben_golub's bigger points about AI vs human reviewers (who'd certainly not score well on this)

Hossein Alidaee retweeted
andrewjkoh Profile Picture

Andrew Koh @andrewjkoh

3 weeks ago

New paper w/ Drew on R&D races in the shadow of ruin: arxiv.org/abs/2607.27638. There've been recent calls for slowdowns/pauses. A common objection goes: "if we slow down, others won't, so we shouldn't." What to make of this? We work out the game theory of how the frontier is shaped by: - Competition: what do I gain from being ahead? lose from being behind? - Coordination: when the tech becomes really dangerous I’d like to stop, but only if my rival also stops - Transparency: how quickly are actions observed? - Trust: how confident am I that my rival is rational?

halidaee Profile Picture

Hossein Alidaee @halidaee

3 weeks ago

@LocasaleLab I'm usually really sympathetic to argument that PhD should be rarer and only people who intrinsically care about knowledge should get it. I just don't think 2% stat is a relevant diagnostic.

halidaee Profile Picture

Hossein Alidaee @halidaee

3 weeks ago

@LocasaleLab Slightly disagree here. I think it's imptnt to think about the labor model in a global context. Very small % world pop has PhD. US PhD % highly foreign. Economically, not surprising US has taken on specialization of being a 'knowledge economy' where ppl at frontier gather.

halidaee Profile Picture

Hossein Alidaee @halidaee

3 weeks ago

@thsottiaux Not codex specific but LLM overall, at least w/in academic research: in the reasoning stage, they will often come up with their own jargon that doesn't exist anywhere in the project, and then refer to it in the output, so very hard to interpret. Agents.md has limited value here.

halidaee Profile Picture

Hossein Alidaee @halidaee

3 weeks ago

Underappreciated benefit of AI: it has far more patience for testing out how long a script will take to run and manually adjusting slurm resource requests for large array jobs each time.

halidaee Profile Picture

Hossein Alidaee @halidaee

3 weeks ago

@thsottiaux Give auto-review agent a full history of actions taken in the thread. E.g. I'm ok with it manipulating files created in the same thread any way it wants, but I have much stricter guidelines on any pre-existing infrastructure. Current auto-review can't reliably tell difference.

halidaee Profile Picture

Hossein Alidaee @halidaee

3 weeks ago

@D_Langenmayr Even better is telling it to do two stage prompt for you. First giving you a prompt to have opus build out a set of writing-related subagents for your process. Then a prompt on how to use them to do the rewriting. (Subagents will also avoid muddying the context window)

halidaee Profile Picture

Hossein Alidaee @halidaee

3 weeks ago

@D_Langenmayr Feed Fable an old paper and the current one. Tell it the full scope of your concerns (i.e. that you're worried about both structural issues and claudisms, and it needs to think at section, paragraph, and individual sentence level). Tell it to write a prompt for opus.

halidaee Profile Picture

Hossein Alidaee @halidaee

3 weeks ago

@alz_zyd_ I think we already saw much of a indie videogame boom gameplay wise. But I think a more of games that leaned into the 'retro' aesthetic because it was also cheaper to make indie (eg into the breach) are now going to look AAA.

halidaee Profile Picture

Hossein Alidaee @halidaee

4 weeks ago

@alz_zyd_ Genetics has this constraint but about the amount of compute, not programming. Eg Alphafold predicts loss of function but you need a molecular dynamics simulation to see if there's a structural rearrangement problem. But even HPC often isn't enough for the latter Q.

halidaee Profile Picture

Hossein Alidaee @halidaee

4 weeks ago

@MeganTStevenson My (coarse) understanding is that actually in the 80s and 90s, picking RAS/MAPK pathway work vs other catalogued oncogenes was a matter of taste. I think taste in that field also comes to play in implementation (e.g. answering question w/ patient organoids vs cell lines)

halidaee Profile Picture

Hossein Alidaee @halidaee

4 weeks ago

@btshapir @asymmetricinfo In other words, if you don't even know how you'd communicate when you're actually right, I don't think you know how to communicate the uncertainty layer around that properly.

halidaee Profile Picture

Hossein Alidaee @halidaee

4 weeks ago

@btshapir @asymmetricinfo To clarify, I *strongly* don't actually think this. But I'm also not 100% confident it is wrong. And I think it's an important case to consider when studying approach to communication. I think learning how to do it right in this case is at least revealing of best strat elsewhere.

Trends for United States

Florida
Vindman
Mariners
Bregman
Angie Nixon
Jolly
Cubs
#mcrdc
Fishback
Sasuke
Donalds
Dan Wilson
#WWENXT
#Cory
DeSantis
Javier Assad
Qixi
Grant Taylor
#FlyTheW
Rockstar

You might like

Assistant Professor of Economics @ UCLA || Macroeconomics
Joao Guerreiro
@jptguerreiro
1K Followers
Assistant Professor of Economics, University of Michigan. Development economist, studying female entrepreneurship and digital financial services.
Emma Riley
@EmmaRiley19
3K Followers
Assistant Professor @HarvardHBS TOM Unit | PhD @KelloggSchool | Behavioral & Experimental • AI–Human Interactions • Tech Adoption
David Almog
@davidalmog25
3K Followers
Assistant professor @StanfordEcon | environment, development, IO, trade
Allan Hsiao
@AllanHsiao
1K Followers
Development and Labor Economist Assistant Professor @UChi_Economics @clbrown.bsky.social
Christina Brown
@christinalbrown
3K Followers
Asst Prof @BarnardCollege @Columbia @BarnardEcon. Labor and gender economics. Formerly @TilburgU, @SIEPR, @NUeconomics, @chicagofed, @dartmouthecon
Ashley Wong
@ashleyywong
954 Followers
AP at EIEF || Corporate Finance, Trade and Macroeconomics || PhD @KelloggPhD
Joao Monteiro
@JoaoMonMon
150 Followers
Assistant Professor of Economics @UCLA. PhD @NUEconomics. MSc, BSc @Ekonomski.
Filip Obradović
@ObradFilip
634 Followers