Category: Web/Tech

A scientific benefit (and cost) of AI innovation

by  Tyler Cowen in

What changed was that the cost of preliminary exploration collapsed. I could sketch an argument, identify the first serious objections, test whether they were fatal, and reach a provisional verdict in an afternoon rather than a fortnight. This sounds like a simple acceleration, and the more profound effect was on what I was willing to abandon. Dropping a question after an afternoon’s work feels nothing like dropping one after three weeks. When the exploration costs are low, the sunk cost attachment disappears, and you find yourself dropping bad questions earlier and more often, which means the questions you keep are better. I explored far more ideas, and my working portfolio became both larger and better curated. I arrived at this outcome not through any deliberate plan but simply through sustained engagement with a tool that changed what exploration cost.

The skill that improved most, and the one I would never have thought to look for, was something I can only describe as question-identification – the ability to find problems that are both tractable and important. This is the thing an academic career is substantially built on and which nobody, so far as I know, has ever tried to teach directly.

I want to be honest about the costs. My ability to hold together a complex position verbally, under pressure, in a seminar or a conversation, has probably not improved and may have declined somewhat. When preliminary exploration is cheap, you spend less time grinding through arguments from first principles, a grinding that builds fluency that shows up in live exchange. Friends have pressed me on this, and they are right to worry.

That is from Carlo Cordasco, and there is more, via Conor Friedersdorf.

How will AI and the fertility crisis interact?

by  Tyler Cowen in

That is from my latest Free Press column, here is one excerpt:

Each individual will be seen as something special by the other humans. Public spaces will be emptier, so anyone out in public will attract more notice. If you are waiting in line at the movie theater, you will be more likely to start talking to the person next to you. After all, you already have had the option of talking to the AIs all day long.

It has long been the norm in American small towns that you say hi to the people you pass on the sidewalk, or perhaps start chatting with customers in your store who appear to be outsiders. Those kinds of practices will spread to the large cities of today, which will become like smaller towns due to lower population density.

Many of these humans will invest heavily in their appearances, in their charisma, and in their “vibes.” After all, the AIs will, and already do, perform so many useful informational functions. If you, as a human, wish to draw attention to yourself and be seen as noteworthy, you will have to specialize in the remaining human functions. That may include “touching grass,” giving warm and appropriate hugs, looking good or at least looking interesting, and having some kind of unique identity that either is visible upon meeting or which AI smart glasses will communicate during social interactions. (“This guy has sailed around the world three times and punched a shark on the nose.”)

The YouTube celebrity Clavicular has attracted a lot of ridicule for his “looksmaxxing,” which involves a lot of manipulation of his appearance and some plastic surgery. Like it or not, that is a harbinger of how some aspects of this future will operate. Clavicular has achieved nothing of note, except for being immediately recognizable for how he looks. For similar reasons, people are likely to pay more attention to how they dress, what kind of makeup they wear, and other aspects of their appearance, such as how tall they are and how much they weigh. Plastic surgery and the successor drugs to GLP-1s are likely to command even more interest than today.

If a person comes across as extremely nondescript, you might feel there is no reason to speak with that person instead of chatting with your AI. A lot of ordinary social interactions will become more like a gala, where everyone shows up wanting to look a very particular way to draw attention.

To inhabitants of 2026, that might sound stupid, undesirable, and ridiculous. I do not love the thought myself. Yet people today care much more about how they look, and can do much more about it, than could people in medieval times. We are used to those differences, and few of us wish to go back to earlier times. People in this future may well feel the same way.

There are other interesting points at the link.

Rent Control: The Ceiling Trap

by  Alex Tabarrok in

Rent control is in the news again. Check out my new website, Rent Control: The Ceiling Trap. Here is just one bit:

Norway abolished its rent control in 1982, and the economist Are Oust realized the newspapers had been quietly recording the whole experiment. He collected housing classifieds from Oslo’s Aftenposten from 1970 to 2008 and watched the market turn inside out.

Under rent control, Oslo’s listings pages looked nothing like a housing market. It was tenants who advertised, pleading their qualities to landlords — “housing wanted” ads outnumbered “housing for rent.” Ten to fifteen percent of those ads were placed by the tenant’s employer, vouching for them the way a bank vouches for a borrower. Tenants offered babysitting, gardening, snow-shoveling, and janitorial work on the side to sweeten the deal. Landlords, for their part, could demand a tenant of a particular gender, age, occupation, region of origin — some ads specified “strong Christian beliefs.” Deposits commonly ran to 50 or 60 months’ rent, occasionally 100 or more: tenants effectively lent the landlord the equity of the flat, interest free. And only about 20 percent of “for rent” ads dared print the rent, much of which would have been illegal.

Then the ceiling lifted. Within a few years the page flipped: landlords advertised to tenants, roughly 80 percent of listings printed an asking rent, the mega-deposits vanished, and the demands for snow-shoveling Christians of specified gender dwindled to nothing. The price went back to doing the rationing — so nothing else had to.

Check out the whole thing–it’s fabulous.

Differentiation drives the erosion of positivity on social media

by  Tyler Cowen in

We live in a digital age, where billions of people engage in dialogue within topic-bound communities and threads. In an archival analysis of over 2 billion Reddit comments and an experiment, we show that this dialogue becomes more negative over time. Further analyses suggest that negativity rises over time because social media users seek to make unique comments on the same topic, and it is easier to differentiate oneself through negative comments than through positive comments. As threads and communities evolve, and it becomes more difficult to make unique observations, users turn to negativity. Our studies show how basic human motives interact with the structure of social media platforms, posing an acute challenge for sustaining healthy online dialogue.

Here is the article by Hongkai Mao, et.al. Via the excellent Kevin Lewis.  For some of you commenters, how does it feel to be a puppet in the unfolding of this game?

Jackson Dahl podcasts with me and Nabeel on aesthetics

by  Tyler Cowen in

Filmed at home, this ran about two hours, and yes that is Nabeel Qureshi, with a cameo from Spinoza toward the very end.  From Jackson:

Links

From the episode summary:

Tyler and Nabeel are good friends, and given how prolific Tyler is, I decided to use Nabeel as an entry point and interview them together. We discuss sacred commitments, AI acceleration, mentorship, friendship, and more, but I focused the majority of the conversation on art and aesthetics. Tyler and Nabeel are unlikely aesthetes given their day jobs, but in fact take art deeply seriously. They have a shared love for and similar tastes in art, music, and film, in particular. We discuss strange and beautiful art, aesthetic stagnation, and a wide range of favorites: The Beatles, Mozart, Mondrian, Springsteen, Lana Del Rey, Kanye West, Cassavetes, The Sopranos, Apichatpong Weerasethakul, and more.

https://www.youtube.com/watch?si=wC78q_BeD27XDnLN&v=qPHV-BezoIc&feature=youtu.be

Excerpt:

Tyler: (18:31) I think I’m very mundane in many ways. When Marc Andreessen had that famous tweet about not being too introspective, I know he got slammed for that, but I sympathize with that in many ways. I have my work. I focus on it. I want to go see places I haven’t seen before. That really drives me. I feel pretty well motivated. I do think all kinds of deep thoughts, but to me those deep thoughts feel more superficial than my so-called superficial urges to go around doing things. And I’m fine with that.

…Jackson: (23:25) Do you experience art primarily by thinking or by feeling?

Tyler: (23:29) I don’t even know what those words mean. I experience it by looking at it. I don’t think I have very deep emotional responses. I think it’s pleasure and I feel I learn a lot from it. When I go out and look at other works of art or just the world, I see a lot more than people who don’t live with art. I don’t think I feel that much. I’ve never cried in front of a painting. When I read these accounts of someone seeing a Madonna and weeping, it makes no sense to me. It’s like people who do sports gambling. Why do you do that? There are positive-sum gambles for you. Here are a few.

There is much more of interest, self-recommending!

AI cheating on math econ at Brown

by  Tyler Cowen in

The temptation to use artificial intelligence (AI) to cheat is shaking up elite universities in the United States. Professor Roberto Serrano, who is the Harrison S. Kravis University Professor of Economics at Brown University, has detected a massive fraud in one of the classes he teaches, ECON 1170, an advanced undergraduate course in mathematical economics. He has conclusive evidence that at least 50 students cheated on the March midterm exam, making it the biggest known scandal at Brown and in the entire Ivy League, which brings together the East Coast’s eight most elite private universities, including Princeton, Harvard, Yale, Columbia, Cornell, Dartmouth College and University of Pennsylvania.

When he reported the case to high-ranking officials at Brown, he got a cold reaction. The response from the president, he said, was absolute silence. The dean did not comment either until Serrano took the case before the Academic Code Committee.

Here is the full story, via Anecdotal.

Typewriters and fertility

by  Tyler Cowen in

Workplace technological changes were instrumental in creating new tasks for women over the last century. This paper studies the adoption of the typewriter into US workplaces. Exploiting exogenous variation in typist demand across sectors, I document that the typewriter increased women’s labor force participation, leading to lower rates of marriage and fertility. These developments stemmed from a transition of White women from households into office work and an indirect crowding-in effect drawing Black women into household services. Acting as a “meeting technology,” the typewriter reshaped social interactions, enabling White women to marry above their socioeconomic backgrounds and achieve upward mobility.

That is from a recent paper by Myera Rashid.  Via Kris Gulati.

An infovore shares his chats

by  Tyler Cowen in

In conjunction with OpenAI, here is a short piece by me on how to use GPT Pro for travel and a few other matters.  Excerpt:

6. Learn what to look for in an art museum

I’m visiting the Detroit Institute of Arts. Which pictures in particular should I be looking for, and what is some useful context for viewing them?

Why was this chat valuable? “After two visits in three days, I would second the advice I got from this chat.”

Here is the link to the answer.  And know when to ask it to read the local newspapers for you, to learn about a place.

There are some good photos in the piece, including some hitherto unpublished photos of Spinoza.

Aaron Levie on current implicit AI regulation

by  Tyler Cowen in

We now have de facto AI regulation. It’s not obvious why from here on out models that have certain levels of capability or are trained on certain compute sizes won’t have to be reviewed by the government before release.

Realistically, as AI models became more and more powerful this was going to be inevitable (I think it’s too early, but here we are). So now it’s mostly just interesting to think about the implications and scenarios from here. A few would be:

* America gets to control who gets access to frontier intelligence and when. This generally works as long as we remain at the frontier at all times and don’t have a risk of being surpassed. At the moment we have a clear lead in frontier intelligence so this is a good bet, but lots of motivated parties would love to change that.

* This likely creates backlog of AI releases which means that we will see less rapid fire back and forth jumps in model progress. Bull/fine case is that we just get bigger step functions per release at a slower rate and we end up at the same point we would have. Bear case is those incremental smaller jumps were necessary for the continued flywheel of innovation.

* Other countries likely have even more incentive to at least hedge their bets with sovereign AI strategies so aren’t dependent on access to US AI all times. Previously this was relatively moot because the alternative wasn’t good enough, but that could change out of necessity and what we’re seeing in China.

* Open weights obviously a big winner here as it becomes what likely sovereign AI gets built out on, and what (for now) can still be released to the market without the same controls. One interesting question would be how regulation eventually extends to open models, which would have its own set of long term consequences.

Anyway some big updates to everyone’s mental models of AI regulation as a result of the capabilities we’re now seeing in AI. Wild times.

Here is the link.  I would say I have long thought something like this was coming, but am pleased we got in so much early progress “under the wire” up until now.  And here is more from Aaron.

Translated from the Chinese

by  Tyler Cowen in

I think this is the Cursor moment for academia.

The Stanford REAP team has made their move, CoPaper.AI is mass-terminating the manual labor of traditional empirical papers. Link: copaper.ai/landing

If using large models to write papers before was just about polishing and compiling references for you, then this Project from Professor Ross Griebenow’s team at Stanford is like dropping a nuclear bomb in the empirical circles of social sciences and economics.

The greatest truth is the simplest; the heaviest sword has no edge. Its functions are straightforward. Feed in the raw dataset, and within 30 minutes, it can generate a complete DOCX paper complete with full Stata/R code and publication-quality charts.

It chains together EDA, variable definition, econometric model building (from OLS to advanced DID, regression discontinuity, causal forests) all using an Agent workflow.

Every chart it produces comes with 100% reproducible Stata, R, EViews source code underneath. How many low-quality paper mills and data drones’ jobs will this smash?

Data drones and paper ghostwriters are collectively facing unemployment countdown. Because from now on, for social science papers, AI handles all the entropy-increasing drudgery—humans only need to define the problem.

Here is the link.  Mostly that is not true, so perhaps the Chinese are trying to demoralize us.  But will it never ever be true?  In two years be true?  Less?

How research in math will change (from my email)

by  Tyler Cowen in

From GA:

I am a mathematician…and some of your recent comments on MR about the role of AI in Econ research as well as the (disappearing?) role of academic papers inspired this response. (It is partially but not exclusively about academia, so I hope it is ok that I’m sending it to your GMU address. Also, I’m hoping it doesn’t get flagged as spam because of the ai in the title…)

In no particular order:

-In the course of a math career, one accumulates lots of computational guesses, now one can test those with minimal effort.

-One also accumulates lots of incomplete and half formed drafts, proofs of special cases, etc, etc. Running those past claude and chatgpt can (does!) pay off. A lot of math is cleverly applying linear algebra and while I’m very good at linear algebra, I’m not as good at it as the AI’s are.

-The lower hanging fruit here are slightly off the beaten track, but not esoteric subjects. If you have a good overview of such, you can pretty quickly prod ai’s into making progress on them. (Before, you needed to have a school of grad students for that). Basic techniques (graph theory, algebra, calculus..) that ai’s are already good at can push these forward already. Making progress on truly hot topics is harder.

-There are some quite smart people trying to measure just how good autonomous ai’s are at math (e.g. the first batch project). That’s a fun game, but for practical purposes right now, what is relevant is how good an ai is when guided by a motivated human. I suspect we’ll see some remarkable things on that front in the next few years once math people really grok the good routines.

-For instance, getting claude and chatgpt to referee each other’s arguments is fun, and they genuinely have different insights on parts of the same problem.

-The kids will be all right. Right now, they are making pocket change doing ai training developed a better “feel” for the different ai’s that I probably ever will. And they learn things by asking the ai to explain an argument to them instead of trying to decipher a math book or paper.

-Which brings me to your papers point. I notice that a project informed with the right context is much more informative to me than the physical pdf of a math paper, and much easier to extract information out of by just asking the thing.

-Refereeing will look very different very soon. All the referee reports that have been collected by the journals should be valuable, hard to get data. And running all the accepted and published math papers through ai’s as the `control’ will end with quite a few people having egg on their face. It’s like self-driving cars, but there is no refereeing union.

-The last really big math revolution was all the stuff in the wake of Witten and 4-manifold stuff predicted by string theory in the early 90’s. This is going to be so much bigger than that. Buckle up.

AI-Native Firms

by  Tyler Cowen in

Very important work from Hyunjin Kim and Rembrand Koning. Insead and HBS respectively:

We study how firms built around AI capabilities-“AI-native” firms-are organized. Drawing on Y Combinator batches W20-F24 and U.S. venture-backed startups whose first financing closed between 2020 and 2024, we classify each firm’s AI-native status and link it to workforce microdata on team size, function, seniority, and hierarchy. Relative to non-AI startups in the same industry-cohort, AI-native firms are 25% smaller. Their share of engineers is 13% greater, and the shares of entry-level workers and managers are each roughly 15% lower. Their hierarchies are half a seniority level flatter-yet valuations are comparable, implying more value created per employee. We argue these patterns reflect two channels: a process channel, in which AI changes how people work inside the firm, and a product channel, in which AI capabilities are built into what the firm sells. Using text from product descriptions and job postings, we find that embedding AI into the product, beyond layering on AI tools into existing workflows, is a primary way startups are scaling “knowledge work” without large teams of knowledge workers.

The tweet storm on the new paper is especially useful.  Via Luis Garicano.  And note those results predate the very latest and best tools.