My position on the Triumphal Arch in DC is they should only do it if the victorious general is driven around the city for the triumph with someone whispering “menento mori” to them

It's not a crazy job to stick around after AI displaces a bunch of stuff

I would think it goes the opposite way - people do so much e-commerce now you care about it more. I still lose packages in a doorman building and it’s annoying

Looks like this building has a package room. That was the biggest problem I had in a non-doorman building in Chelsea (the only building destroyed by Hurricane Sandy; still an empty lot): my packages were just constantly being stolen.

I always thought full doorman services were a legacy of high crime New York and would fade, but we’ll see! It’s also possible that‘s just how people like spending higher income.

It was le best of times; it was le worst of times

Not really clear to me why Vietnam has done so much better than Burma

5. Finding Needles in Haystacks AI in fraud detection and compliance

Week five of AI in Finance: Fraud detection and compliance

• How embeddings, LLMs, and graph methods can improve fraud detection
• Where AI addresses the compliance and regulatory burden
• The arms race of AI advancing cybersecurity risks and defenses

arpitrage.substack.com/p/finding-ne...

There is much more in the piece, including comparisons with previous efforts at Industrial Policy in the US and other countries, and a discussion of the political economy.

We hope it spurs more people to think about the challenge of improving construction productivity

5⃣And finally — government land-banks to help overcome parcel assembly frictions which make it hard for developers to create large plots for development.

We suggest an industrial policy agenda to change this:

1⃣Support for R&D
2⃣Aggregated procurement and advanced purchase commitments to smooth out demand
3⃣Incentives and risk management tools to limit downside of factory investments
4⃣Harmonizing building code requirements

We start with the basic problem: due to a number of overlapping frictions, housing construction is prone to boom-bust cycles and fragmented markets.

This results in a capital-light, contractor-heavy model of on site construction which has seen *negative* productivity growth

Industrial Policy for Housing Construction Since the 1970s, productivity in the housing sector has steadily declined. The way forward requires combining continued land-use reform with a light-touch industrial policy for factory-built housing. ...

I have a new article in National Affairs on Industrial Policy for Housing Construction!

With Steve Teles, we outline an agenda for deregulation and government support to grow factory built housing:

piece: nationalaffairs.com/publications...

substack: arpitrage.substack.com/p/industrial...

Conspiracy theory: the algorithms have figured out showing ads to all members of the household unit increases conversion rates

Always wondered whether visiting UChicago in April played a role in going there

Hey, this is our piece!

Industrial Policy for Housing Construction Since the 1970s, productivity in the housing sector has steadily declined. The way forward requires combining continued land-use reform with a light-touch industrial policy for factory-built housing. ...

i look forward to harassing steven next week about his industrial housing policy piece with @arpitrage.bsky.social

nationalaffairs.com/publications...

GitHub - dmilo75/ai-zoning Contribute to dmilo75/ai-zoning development by creating an account on GitHub.

Our data on housing regulations, parsed across municipalities, is publicly available here.

We'll be adding to this soon in the form of a more easily accessible website. And, in the longer-term, with historic ordinances and zoning maps.

github.com/dmilo75/ai-z...

Am several years late to the party in watching Fred again.. at that London rooftop - pretty good!

The issue is whether we could “expect” that not on the basis of rough human priors, but large amounts of fitted data of very related choices that extrapolate a bit to the policy in question, combined with some degree of reoptimization and price adjustment in model

“Solves”

It’s world models all the way down

It’s taken everything I have to avoid tweeting ”Through transformers, economics will achieve its destiny of becoming the field of psychohistory”

AI one-shot papers Diff-in-diff everywhere

My solution to the @paulgp.com AI diff in diff slop challenge is to develop economic world models based on transformers which simulate households, firms, and governments; so policy evaluation requires spinning up some counterfactuals and simulations

paulgp.substack.com/p/ai-one-sho...

What did Larry see?

Good way of phrasing it!

4. The End of Market Intelligence and the Last Analyst The escalating arms race in text analysis, and whether you can simulate your customers

Week four of AI in Finance:

• The economics of market intelligence
• GPT can read a 10-K, but so can everyone else, so where does alpha go?
• How to simulate your customers

arpitrage.substack.com/p/4-the-end-...

I’m heading troublesome reports that assassinating the leader of the spice rich regions is not working as per plan; since the son has survived and is leading a religious rebellion blocking spice exports.

Let’s see what happens if Imperial Sardaukar are deployed.

"rising housing costs since 1990 are responsible for 13 million (11%) children not being born, 51% of decrease in fertility from the 2000s to the 2010s, and a 7pp decrease in the share of 20-29 year olds that have started families"

via @arpitrage.bsky.social
drive.google.com/file/d/1BK6j...

3. The Cause of and Solution to all of Life's Problems AI in risk assessment and management, and why better models create new risks

Week three of the AI in Finance course:

• The economics of credit scoring from ledgers to ML and digital footprints
• How AI lets banks lend to strangers
• Why better models create new risks

arpitrage.substack.com/p/3-the-caus...