Andrey Fradkin
@AndreyFradkin
Professor studying digital technology and AI @BUQuestrom and @MIT_IDE, climber, immigrant, nerd dinner convener. Co-host of Justified Posteriors Podcast.
Want to hear and read two crazy smart, funny, and frighteningly well-read economists talking about AI and innovation? Then @SBenzell and @AndreyFradkin are your guys. Links to their stuff are in the next post...
Could AI Save Us From Making Hard Choices About the Budget? An analysis of some recent arguments from Peter Thiel and Tyler Cowen
whatever you do next, aspire to do something that doesn’t make sense to most at the moment.
Just hit our first milestone for the Justified Posteriors podcast! In the next phase, we're looking to mix it up with interview episodes, any suggestions?
If social simulations become good enough and simple enough to make, they will be very useful. A decade ago, I tried to build simulations of Airbnb, and they yielded interesting insights. But at the same time, they were too bespoke and small scale to be really useful for planning…
x.com/i/article/1939…
Louder, for the people in the back. It's a fast moving target, and academia is so slow. At the same time, most companies are interested in PR rather than the truth. Not a great situation for learning about the effects of AI.
Almost all randomized controlled trials on the impacts of AI on innovation, productivity, & job performance pre-dates reasoner models. The ones we do have (a couple of tests of o1-preview in law & medicine) suggest they may lead to a large jump in many fields, but we don’t know
Good to see this one finally in print. Market design for jobs platforms is still grossly understudied given its importance. Glad we could make a modest contribution.
Up for pre-order: The Microeconomics of AI (MIT Press) Yes, I know it's pricey. amazon.com/Microeconomics…
Folks interested in marketplaces who are in the Bay Area, I'll be giving a talk called "Marketplace design: what have we learned?" on Thursday at Lyft HQ. Do come by and say hi! I'll also be at Manifest this weekend.
In the latest Justified Posteriors episode, we discuss empirical studies of AI's ability to persuade. One interesting aspect of this topic is what counts as persuasion. Does offering an incentive count as persuasion, or must it be pure rhetoric? Thoughts?
Highly recommend convening a regular nerd dinner, it's fun and may result in much good for the world.
15 years ago I hosted a weekly Nerd Dinner with a bunch of friends in their early 20s at Dar Bar Indian Cuisine in Palo Alto. Lots of chicken tikka. BCIs were a hot topic at these dinners. We could never know it at the time, but years later many of those friends would go on to…
Hey Justified Posteriors Fans! What do you want to see us discuss next on the pod? We want to hear from you! Put your thoughts in the comments.
I want to second Neal's comment here. Teaching Mullainathan's work for my PhD class reminded me of just how creative and impactful his work is (if you're interested, check out our podcast episode about AI for ER docs). Susan Athey's work is also foundational.
I think some of the best work on issues broadly relevant to AI policy (taken very loosely, I don’t mean policy only like government regulation at all) come from economists. I teach a bunch of it. I would highlight especially Susan Athey and Sendhil Mullainathan.
Thanks to @tylercowen for the shout-out! Stoked to be podcasting with @SBenzell about econ and AI.
I think some of the best work on issues broadly relevant to AI policy (taken very loosely, I don’t mean policy only like government regulation at all) come from economists. I teach a bunch of it. I would highlight especially Susan Athey and Sendhil Mullainathan.
Paper on the economics of AI competition shows new AI models are adopted very rapidly but are not always substitutes for each other (better models in the same family often substitute, but others expand the pie). Also, people use a mix of models. andreyfradkin.com/assets/demandf…