Basil Halperin

2.3K posts

Basil Halperin

Basil Halperin

@BasilHalperin

Assistant professor of economics @UVA; Past lives: PhD @MITEcon, @Uber, @AQRCapital, @UChicago

Katılım Ocak 2012
1.5K Takip Edilen7.5K Takipçiler
Sabitlenmiş Tweet
Basil Halperin
Basil Halperin@BasilHalperin·
Introducing the Stripe Economics of AI Fellowship: The economics of AI remains surprisingly understudied. The fellowship aims to help fill that gap, by supporting grad students and early-career researchers with $, data, a conference, and community –
Basil Halperin tweet media
English
27
183
1.1K
423.7K
Alexander Berger
Alexander Berger@albrgr·
@michael_nielsen (For the record I just thought this was funny, as a number of the replies point out I think there are a lot better securities to look at for AI timelines)
English
2
0
13
826
Basil Halperin
Basil Halperin@BasilHalperin·
"Trying to escape the permanent underclass" is like an Incan trying to save enough money to escape Pizarro, sorry -- Either the political system works (and there is nothing to escape) or you're just screwed ¯\_(ツ)_/¯
Jan Kulveit@jankulveit

The "permanent underclass" meme is primarily bad futurism, where people admit AGI massively changes one domain, but somehow everything else stays roughly 2025. Not impossible, but small slice of futures

English
25
48
776
75.3K
Tim Hua 🇺🇦
Tim Hua 🇺🇦@Tim_Hua_·
The Cope Theory of everything suggests that the “permanent underclass” meme comes from people who (1) feel the AGI and (2) don’t feel like they could affect its trajectory. But people still need meaning and purpose. Thus, a new concept is invented: “if you work hard to…
Basil Halperin@BasilHalperin

"Trying to escape the permanent underclass" is like an Incan trying to save enough money to escape Pizarro, sorry -- Either the political system works (and there is nothing to escape) or you're just screwed ¯\_(ツ)_/¯

English
2
3
46
3.9K
Basil Halperin
Basil Halperin@BasilHalperin·
@captgouda24 tfw your historically-informed analogy is historically uninformed
English
2
0
32
2.5K
Nicholas Decker
Nicholas Decker@captgouda24·
@BasilHalperin But with better tactics, they could have done it. Incan battle tactics were appallingly bad, they had a pathway to beating the Spaniards but nooo they had to fight pitched battles with terrible losses.
English
6
0
25
2.9K
Basil Halperin
Basil Halperin@BasilHalperin·
Scott is very generous; but also what he's too modest to say is -- I'm just trying to be in the tradition of Scott Sumner :) He's also right that David really brought out a lot of great discussion in our podcast -- check it out!! ⬇️
David Beckworth@DavidBeckworth

Scott Sumner on @BasilHalperin: "I see Halperin as a pragmatist in the tradition of Bennett McCallum, which is one reason why he’s my favorite young macroeconomist... Like McCallum, Basil Halperin seems to have absorbed both the best of New Keynesian economics and the best of Milton Friedman thought. He also favors NGDP targeting. He also seems to have excellent intuition about which sort of macro models are plausible and which are not—a skill that’s hard to teach. Even their personalities seem a bit similar, as both come across as being very polite." (1/2) scottsumner.substack.com/p/basil-halper…

English
1
0
41
5.2K
Basil Halperin
Basil Halperin@BasilHalperin·
New paper: AI is good at lots, but labs think automating one thing might be especially important – AI research itself What happens if you embed this into a standard economic growth model? When do you get an ‘economic singularity’?
Basil Halperin tweet media
Anton Korinek@akorinek

1/🆕 New NBER paper: 𝗪𝗵𝗲𝗻 𝗗𝗼𝗲𝘀 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗻𝗴 𝗔𝗜 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗣𝗿𝗼𝗱𝘂𝗰𝗲 𝗘𝘅𝗽𝗹𝗼𝘀𝗶𝘃𝗲 𝗚𝗿𝗼𝘄𝘁𝗵? Under empirically grounded calibrations, a singularity could arrive within just a few years of automating AI research. 🧵 📄 nber.org/papers/w35155

English
5
24
90
16.6K
Basil Halperin
Basil Halperin@BasilHalperin·
1. In the Cobb-Douglas world, the "fraction of tasks automated" = the post-automation *cost-weighted* fraction of tasks that are performed by AI - i.e. you want to measure with expenditure shares 2. In the model, we don't explicitly have new task creation, but new tasks doesn't change that^ empirical notion 3. Away from Cobb-Douglas, the object we'd want to measure is that "elasticity of innovation to AI inputs" described in section 7 4. @tomwhoulden usefully points out that perhaps the version of our proposition under new task creation is "explosive growth if f > 0.13 … and f stays there" - i.e. is not diluted by new task creation (Useful q!!! We have been setting aside new task extension)
English
0
0
2
165
sam manning
sam manning@sj_manning·
Glad this is out! Really great stuff! One thing I'm unsure how to interpret: What do you mean by the "fraction of tasks in a sector automated by AI"? Won't the denominator change there as new tasks emerge in response to automation of other tasks? Like its hard to put a number on the fraction of tasks in finance that are automated by excel formulas, since there is still a "full" set of tasks that humans perform in finance jobs. I would guess that a large share of tasks in software development as of a 2022 task list are now automated by AI, but there is still a full set of non-automated tasks that comprise software development today that you could use as your denominator. Essentially, am curious how we would know if that 5% automation in other sectors condition was reached. Maybe one way would be if labor productivity rose 5%?
English
1
0
4
239
Basil Halperin
Basil Halperin@BasilHalperin·
Automating 1% of all semiconductor research ≈ automating 10% of all general research (in this sense) For me, staring at this has raised my perception of how important it is to monitor automation in the chip industry. Even a little bit of automation there goes a long way...
English
1
2
13
3.4K
Basil Halperin
Basil Halperin@BasilHalperin·
While ideas are getting harder to find for the economy as a whole (the level of diminishing returns is β=3.1)... ...in software they are less so (β≈1)... ...and in hardware much less so (β=0.2)!
English
1
0
2
446