Andrey Fradkin
@AndreyFradkin
Followers
2K
Following
13K
Media
67
Statuses
1K
Professor @BUQuestrom and @MIT_IDE, climber, immigrant, nerd dinner convener. Co-host of Justified Posteriors Podcast. On leave, AI + markets @ Amazon.
San Francisco / Boston
Joined December 2010
We also discuss the mysterious lack of agreement among raters regarding which artifacts are better. We posit that the ability to cater the artifact to the reader is an important part of quality that is not captured by this eval. Please listen! https://t.co/sWRNSGiijQ
0
0
2
A key reason is that the artifacts (PowerPoints, schematics, and documents) produced are not that valuable without the social coordination, consensus-building, and decision-making that happens within organizations.
1
0
2
In our latest episode, we discuss OpenAI's GDPval, a carefully done evaluation of AIs ability to do highly compensated work. We were surprised by how well the AIs did at producing artifacts, and yet skeptical that this capability is enough to automate workers. @SBenzell
1
4
6
Microsoft research has created an interesting environment for testing agents interacting in a virtual market. It's a great first step in moving towards useful AI agents for economic activity. https://t.co/bRhfWuOJjt
microsoft.com
Open-source agentic marketplace platform. Study how AI agents search, negotiate & transact. Reveals biases & vulnerabilities.
0
1
5
I think this article gets at the point in our NBER chapter beautifully. An AI that tirelessly searches for the "best deal" on your behalf will surely affect price dispersion https://t.co/3lN56mODE1
As AI goes mainstream, it will remove one of the most enduring distortions in modern capitalism: the information advantages that sellers and service providers enjoy over consumers
1
1
26
Papers in all fields should have an explicit contribution statement, even if it's just 'all authors contributed equally.' This would help the algorithms attribute credit and reduce a lot of confusion about author ordering.
The metricization of science is the death of science. Any scientist "welcomes this initiative" should be permanently disbarred. https://t.co/hvJBxAeaCP
1
0
8
To be clear, we should be rigorously thinking about policy for AI, but in separate papers and (for peer review) in outlets where policy evaluation experts are refereeing.
0
0
1
I see a related pathology in AI papers. Many AI papers, due to earnest concern or just norms, end up having sections discussing how AI may automate labor and then recommend some vague and politically coded policy solution. It's probably best to omit this section.
Agree 1000%. Nature journals can't help but keep doing things that would cause a neutral observer to say "so academics care more about politics than truth seeking, huh?" My recommendation:
2
0
7
I am super excited to share a new AI tool, Refine. Refine thoroughly studies research papers like a referee and finds issues with correctness, clarity, and consistency. In my own papers, it regularly catches problems that my coauthors and I missed. 1/
38
205
1K
Very excited that the clinical trial results for our PRIMA retinal prosthesis are published today in the New England Journal of Medicine! This is the first time that patients who are blind due to photoreceptor loss have been able to intuitively see again.
57
219
2K
(1/11) How do AI agents change markets? We speculate on this topic in our new article for the NBER “Economics of Transformative AI” volume with @gilirusak, @BenSManning, @AndreyFradkin, and @johnjhorton: “The Coasean Singularity? Demand, Supply, and Market Design with AI Agents”
12
49
169
My sister-in-law's new math book for kids, Modultown!, just launched on Kickstarter. It teaches modularity through a fantasy story about magical beings with periodic powers. Please check it out! https://t.co/SfraAd65YT
kickstarter.com
Experience a whimsical puzzle hunt in a new fantastic world full of friendship and surprises, from the authors of Funville Adventures
0
0
3
In the latest Justified Posteriors episode, we discussed AI as Governance by Henry Farrell. My takeaway, AI will change governance structures and I'd like to read more from political scientists on the topic. Any recs? Seems that they've been pretty silent.
1
0
2
Minor life accomplishment, and had to share. Beat A20 in slay the spire. What a great game.
0
0
2
I wrote a long blot post on the economics of AI, prompted by two great workshops (Windfall and NBER), and me leaving OpenAI for METR: TLDR: we driving in the fog. 1. There is no standard model of AI’s economic impact. Economists have been using a wide range of assumptions to
22
94
473
Idea prompted by my attempt to create this t-shirt. https://t.co/G75Z9S8PRB
I noticed a missing market and decided to correct it. Purchase link below.
0
0
1
I noticed a missing market and decided to correct it. Purchase link below.
1
1
5
New AI benchmark proposal. Can the AI use a browser to create a t-shirt design and successfully sell it online without human intervention. I shall call it t-shirt bench.
2
1
8