AI Goes to College…for the Free Money

Last year, the state [CA] chancellor’s office estimated 25 percent of community college applicants were bots.

Everyone understands that students are using AI; sometimes to help them learn, sometimes to avoid learning. What I didn’t appreciate is that community colleges offering online courses are being flooded with AI bots who are taking the courses:

The bots’ goal is to bilk state and federal financial aid money by enrolling in classes, and remaining enrolled in them, long enough for aid disbursements to go out. They often accomplish this by submitting AI-generated work. And because community colleges accept all applicants, they’ve been almost exclusively impacted by the fraud.

The state has launched a Bladerunner-eque “Inauthentic Enrollment Mitigation Taskforce” to try to combat the problem. This strikes me, however, as more of a problem on the back-end of government sending out money without much verification. It’s odd to make the community colleges responsible for determining who is human. Stop sending the money to the bots and the bots will stop going to college.

Sam Altman, as usual, is ahead of the game.

*Hope I Get Old Before I Die*

That is the new and fun book by David Hepworth.  It focuses on the careers of rock stars who simply keep on going and do not retire.

Can we admit that Paul McCartney and also the Rolling Stones have made the best of this?

Here is one bit:

Of the ten most-visited graves in the USA, just one is the resting place of a president.  The rest are all the graves of entertainers.

I liked this line:

‘Sometimes I feel like I work for Liz Phair,’ she [Liz Phair] says.  ‘And I have years off but then, like, I work for her.’

You can order the book here.

Xenophon’s consultation of the Pythia

1. Statement of prayer-question – Xenophon begins by verbally addressing Apollo, asking “to which of the gods should I sacrifice and pray in order best and most successfully to perform the journey I have in mind and return home in safety?” Only once this plea is uttered does Apollo’s priesthood record the god’s reply.

2. Ritual hymn & payment – Like all individual consultants, he had to buy a pelanos (sacrificial cake) and burn it on the altar while reciting the short Delphic paean in Apollo’s honour; the spoken hymn and the offering together signalled respect and opened the way for prophecy.

3. Sacrificial plea – A goat was sprinkled with water; if it shuddered, Apollo was deemed willing to speak. The consultants (or an attendant priest) then voiced a brief prayer “Hear me, Lord Apollo…” over the animal before it was sacrificed. Only after this spoken plea did the Pythia mount the tripod and deliver the oracle.

That is an o3 answer in response to one of my queries, namely whether you had to make incantations to oracles before they would respond.  You did!  If you scroll down, you will see that original answer is amended somewhat and improved in accuracy.  For instance “…drop the idea that each visitor had to intone a fixed hymn. At most, priests might intone a brief paean while the cake was burned…”

In any case, you could not do “one shot” with the oracle — you had to put a bit of effort into it.  If you simply approached them and asked for a prophecy of the future (and did nothing else) you would get no meaningful response.  In contemporary terminology, you needed a bit of prompting.

To return more explicitly to the current day, many people complain about the hallucinations of top LLMs, and indeed those hallucinations are still present.  (o3 is much quicker than o1 pro, but probably has a higher hallucination rate.)  If you ask them only once, you are more likely to get hallucinations.  If you ask a follow-up, and request a correction of errors, the answer usually is better.

Almost everyone evaluates the LLMs and their hallucinations on a one-shot basis.  But historically we evaluated oracles on a multi-shot basis.  It would be easy for us to do that again with LLMS, and of course many users do.  For the faster models the follow-up query really does not take so long.

Or just start off on the right foot.  Marius recommends this prompt:

Ultra-deep thinking mode. Greater rigor, attention to detail, and multi-angle verification. Start by outlining the task and breaking down the problem into subtasks. For each subtask, explore multiple perspectives, even those that seem initially irrelevant or improbable. Purposefully attempt to disprove or challenge your own assumptions at every step. Triple-verify everything. Critically review each step, scrutinize your logic, assumptions, and conclusions, explicitly calling out uncertainties and alternative viewpoints. Independently verify your reasoning using alternative methodologies or tools, cross-checking every fact, inference, and conclusion against external data, calculation, or authoritative sources. Deliberately seek out and employ at least twice as many verification tools or methods as you typically would. Use mathematical validations, web searches, logic evaluation frameworks, and additional resources explicitly and liberally to cross-verify your claims. Even if you feel entirely confident in your solution, explicitly dedicate additional time and effort to systematically search for weaknesses, logical gaps, hidden assumptions, or oversights. Clearly document these potential pitfalls and how you’ve addressed them. Once you’re fully convinced your analysis is robust and complete, deliberately pause and force yourself to reconsider the entire reasoning chain one final time from scratch. Explicitly detail this last reflective step.

I haven’t tried it yet, but it doesn’t cost more than a simple “Control C.”  Perhaps some of you can do better yet, depending of course on what your purpose is.

There is no reason why you cannot ask for better, and get it.  Beware those who dump on hallucinations without trying to do better — they are the Negative Nellies of LLM land.

And oh — o3 pro is coming soon.

Spain facts of the day

  • In 1990, less than 1% of the Spanish population were foreign residents. The foreign-born population was even smaller, with immigrants accounting for about 0.5% of residents.
  • In 2023, Spain alone accounted for 23% of all naturalizations in the European Union

As of 2025…

  • 14% of residents in Spain are foreign nationals.
  • Nearly 20% of Spain’s population was born outside the country.
  • 1 in 7 residents of Madrid were born in Latin America.

That is from the Show Notes to Rasheed Griffith’s podcast,

The Library Burned Slowly

A powerful but grim essay by John McGinnis, Professor of Constitutional Law at Northwestern. For decades, the federal government—driven by the left—expanded its control over universities. The right, most notably Ronald Reagan, tried to resist, shielding civil society from state overreach. They failed. Now, a new right has turned to the left’s playbook and is imposing its own vision of the good society. Chris Rufo mocks classical liberals like myself and their naive ideas of neutrality, fairness and open institutions. Principles are for losers. Seize power! Crush your enemies. Rufo does know how to crush his enemies. But what happens when the devil turns? Bludgeoning your enemies is fun while it lasts but you can’t bludgeon your way to a civilization. Hayek’s civil society dies in the rubble.

It seems remarkable that seemingly antisemitic protests by undergraduates, such as those at my own university of Northwestern, could threaten the biomedical research funding of its medical school. But the structure of civil rights laws as applied to universities has long allowed the federal government to cut off funding to the entire university based on the wrongful actions of particular units or departments.

Ironically, the left, now alarmed by the federal government’s intrusive reach, bears direct responsibility for crafting the very legal weapons wielded against the universities it dominates. Almost four decades ago, progressive legislators demanded sweeping amendments to civil rights law, expanding federal oversight over higher education. The sequence of events reveals a cautionary tale of political hubris: progressive confidence that state power would reliably serve their ends overlooked the reality that governmental authority, once unleashed, recognizes no ideological master. Today’s circumstances starkly illustrate how expansive federal control over civil society, originally celebrated by progressives, returns to haunt its architects. The left’s outrage ought to focus not on this particular administration but on its own reckless empowerment of the state.

…Clumsy governmental dictates on contentious matters such as transgender rights do not merely settle disputes; they inflame societal divisions by transforming moral disagreements into winner-takes-all political battles. Civil society, by contrast, thrives precisely because it embraces diversity and facilitates compromise, allowing pluralistic communities to coexist peacefully without being conscripted into ideological warfare. The left, fixated upon uniform outcomes, consistently undervalues the power of voluntary cooperation and cultural persuasion. Their shortsightedness has delivered into the hands of their opponents the very instruments of coercion they forged, vividly confirming an enduring truth: the power you grant government today will inevitably be wielded tomorrow by your adversaries.

Read the whole thing.

Large Language Models, Small Labor Market Effects

That is a new paper from Denmark, by Anders Humlum and Emilie Vestergaard, here is the abstract:

We examine the labor market effects of AI chatbots using two large-scale adoption surveys (late 2023 and 2024) covering 11 exposed occupations (25,000 workers, 7,000 workplaces), linked to matched employer-employee data in Denmark. AI chatbots are now widespread—most employers encourage their use, many deploy in-house models, and training initiatives are common. These firm-led investments boost adoption, narrow demographic gaps in take-up, enhance workplace utility, and create new job tasks. Yet, despite substantial investments, economic impacts remain minimal. Using difference-in-differences and employer policies as quasi-experimental variation, we estimate precise zeros: AI chatbots have had no significant impact on earnings or recorded hours in any occupation, with confidence intervals ruling out effects larger than 1%. Modest productivity gains (average time savings of 2.8%), combined with weak wage pass-through, help explain these limited labor market effects. Our findings challenge narratives of imminent labor market transformation due to Generative AI.

Not a surprise to me of course.  Arjun Ramani offers some interpretations.  And elsewhere (FT): “Google’s core search and advertising business grew almost 10 per cent to $50.7bn in the quarter, surpassing estimates for between 8 per cent and 9 per cent.”

Slow takeoff, people, slow takeoff.  I hope you are convinced by now.

The macroeconomics of tariff shocks

There is a new paper by Adrien Auclert, Matthew Rognlie, and Ludwig Straub.  It seems timely?:

We study the short-run effects of import tariffs on GDP and the trade balance in an open-economy New Keynesian model with intermediate input trade. We find that temporary tariffs cause a recession whenever the import elasticity is below an openness-weighted average of the export elasticity and the intertemporal substitution elasticity. We argue this condition is likely satisfied in practice because durable goods generate great scope for intertemporal substitution, and because it is easier to lose competitiveness on the global market than to substitute between home and foreign goods. Unilateral tariffs tend to improve the trade balance, but when other countries retaliate the trade balance worsens and the recession deepens. Taking into account the recessionary effect of tariffs dramatically lowers the optimal unilateral tariff level derived in standard trade theory.

I wonder what the policy implications might be.  Here is a good thread on the paper.

Thursday assorted links

1. Are these the most beautiful colleges in America?  Where is the brutalism?  UC Irvine has some.  I feel like I have been to way, way too many of those campuses.

2. Facts about egg supply chains.

3. Good robot photos in this NYT story about China.

4. “Apply to our 1517 garage science cohort – we’re looking for renegade scientists to kickstart a revolution

5. Greek primary surplus is 4.8% of gdp, remarkable.  And Argentina forecasts are now for 5.5% growth for 2025, also wonderful.

6. “Australian Radio Network (ARN), the media company behind KIIS, as well as Gold and iHeart, used an AI-generated female Asian host to broadcast 4 hours of midweek radio, without disclosing it.

Whose disinformation?

Meta, which owns Facebook and Instagram, blocked news from its apps in Canada in 2023 after a new law required the social media giant to pay Canadian news publishers a tax for publishing their content. The ban applies to all news outlets irrespective of origin, including The New York Times.

Amid the news void, Canada Proud and dozens of other partisan pages are rising in popularity on Facebook and Instagram before the election. At the same time, cryptocurrency scams and ads that mimic legitimate news sources have proliferated on the platforms. Yet few voters are aware of this shift, with research showing that only one in five Canadians knows that news has been blocked on Facebook and Instagram feeds.

The result is a “continued spiral” for Canada’s online ecosystem toward disinformation and division, said Aengus Bridgman, director of the Media Ecosystem Observatory, a Canadian project that has studied social media during the election.

Meta’s decision has left Canadians “more vulnerable to generative A.I., fake news websites and less likely to encounter ideas and facts that challenge their worldviews,” Dr. Bridgman added.

You can argue this one all sorts of ways, but perhaps there is a lesson in here…?  Here is the full NYT piece, via the excellent Kevin Lewis.

The Prophet’s Paradox

The political problem of disaster preparedness is especially acute for the most useful form, disaster avoidance. The problem with avoiding a disaster is that success often renders itself invisible. The captain of the Titanic is blamed for hitting the iceberg, but how much credit would he have received for avoiding it?

Consider a pandemic. When early actions—such as testing and quarantine, ring vaccination, and local lockdowns—prevent a pandemic, those inconvenienced may question whether the threat was ever real. Indeed, one critic of this paper pointed to warnings about ozone depletion and skin cancer in the 1980s as an example of exaggeration and a predicted disaster that did not happen. Of course, one of the reasons the disaster didn’t happen was the creation of the Montreal Protocol to reduce ozone-depleting substances (Jovanović et al. 2019; Tabarrok and Canal 2023). The Montreal Protocol is often called the world’s most successful international agreement, but it is not surprising that we don’t credit it for skin cancers that didn’t happen. I call this the prophet’s paradox: the more the prophet is believed beforehand, the less they are credited afterward.

The prophet’s paradox can undermine public support for proactive measures. The very effectiveness of these interventions creates a perception that they were unnecessary, as the dire outcomes they prevented are never realized. Consequently, policymakers face a challenging dilemma: the better they manage a potential crisis, the more likely it is that the public will perceive their actions as overreactions. Success can paradoxically erode trust and make it more difficult to implement necessary measures in future emergencies. Hence, politicians are paid to deal with emergencies not to avoid them (Healy and Malhotra 2009).

Since politicians are incentivized to deal with rather than avoid emergencies it is perhaps not surprising to find that this attitude was built into the planning process. Thus, the UK COVID Inquiry (2024, 3.17) found that:

Planning was focused on dealing with the impact of the disease rather than preventing its spread.

Even more pointedly Matt Hancock testified (UK COVID Inquiry 2024, 4.18):

Instead of a strategy for preventing a pandemic having a disastrous effect, it [was] a strategy for dealing with the disastrous effect of a pandemic.

From my paper, Pandemic preparation without romance.

My excellent Conversation with Chris Dixon

Here is the audio, video, and transcript.  Here is the episode summary:

Chris Dixon believes we’re at a pivotal inflection point in the internet’s evolution. As a general partner at Andreessen Horowitz and author of Read Write Own, Chris believes the current internet, dominated by large platforms like YouTube and Spotify, has strayed far from its decentralized roots. He argues that the next era—powered by blockchain technology—can restore autonomy to creators, lower barriers for innovation, and shift economic power back to the network’s edges.

Tyler and Chris discuss the economics of platform dominance, how blockchains merge protocol-based social benefits with corporate-style competitive advantages, the rise of stablecoins as a viable blockchain-based application, whether Bitcoin or AI-created currencies will dominate machine-to-machine payments, why Stack Overflow could be the first of many casualties in an AI-driven web, venture capital’s vulnerability to AI disruption, whether open-source AI could preserve national sovereignty, NFTs as digital property rights system for AIs, how Kant’s synthetic a priori, Kripke’s modal logic, and Heidegger’s Dasein sneak into Dixon’s term‑sheet thinking, and much more.

Most of the talk was about tech of course, but let’s cut right to the philosophy section:

COWEN: What’s your favorite book in philosophy?

DIXON: I’ve actually been getting back into philosophy lately. I did philosophy years ago in grad school. Favorite book, man. Are you into philosophy?

COWEN: Of course, yes. Plato’s Dialogues; Quine, Word and Object; Parfit, Reasons and Persons; Nozick. Those are what come to my mind right away.

DIXON: Yes. I did analytic philosophy. I actually was in a graduate school program and dropped out. I did analytic philosophy. Actually, Quine was one of my favorites — Word and Object and Two Dogmas of Empiricism, all those kinds of things. I like Donald Davidson. Nozick — I loved Anarchy, State, and Utopia. Reading that with Rawls is a great pairing. I used to love Wittgenstein, both early and later. I was into logic, so Frege and Russell. This was a grad school.

Now I’m trying to finally understand continental philosophy. I never understood it. I’ve actually spent the last three months in a philosophy phase. I’ve been watching a lot of videos. Highly recommend this. Do you know Bryan Magee?

COWEN: Sure, yes.

DIXON: Amazing. I watched all of his videos. This guy, Michael Sugrue, was a Princeton professor — great videos on continental philosophy. I’ve been reading — it sounds pretentious; I’m not saying I understand this or I’m an expert on it, but I’m struggling in reading it. I’m trying to read Being and Time right now — Heidegger. I really like Kripke. I follow Kripke. I liked his books a lot. Nelson Goodman was one of my favorites. Funny enough, I just bought it again — Fact, Fiction, and Forecast. Kripke — Naming and Necessity is his legendary book on reference and language.

COWEN: I’ve never been persuaded by that one. It always felt like sleight of hand to me. He’s very, very smart. He might be the sharpest philosopher, but I like the book on Wittgenstein better.

DIXON: He basically invented modal logic. I don’t know if you know that story. He was in high school, something.

COWEN: He was 15 years old, I heard. Yes.

DIXON: [laughs] He’s like a true prodigy. Like a lot of philosophy, you have to take it in the context, like Naming and Necessity I think of as a response — gosh, I’m forgetting the whole history of it, but as I recall, it was a response to the descriptive theory of reference, like Russell. Anyways, I think you have to take these things in a pairing.

Actually, last night I was with a group of people. I got a lecture on philosophy, and it was great because he went through Hume, KantHegel, Nietzsche. I don’t want to go too much into that, but I’ve always struggled with Kant. Then he went into Hegel and explained that Hegel struggled with Kant in the same way that I did, and then improved on it. I’m not trying to go into details of this; it’s too much. The point is, for me, a lot of it has to be taken in as a dialogue between thinkers over multiple periods.

COWEN: Are you getting anything out of Heidegger? Because I sometimes say I’ve looked at every page of that book, but I’m not sure I’ve read it.

DIXON: It’s a good question. I have a friend who’s really into it, and we’ve been spending time together, and he’s trying to teach me. If you want, I’ll send you some videos that I think are really good.

COWEN: That’d be great.

DIXON: They’ve helped me a lot. I’ve always got it from an intellectual history point of view. If you want to follow the history of postmodernism, there’s Heidegger and then Derrida, and just what’s going on in the academy today with relativism and discourse and hermeneutics. I think it’s modern political implications that were really probably kicked off by Nietzsche and then Heidegger. I’ve always understood in that sense.

What I struggle with, and I understand him as a theory of psychology, I think of describing the experience of the Dasein and being-in-the-world. To me, it’s an interesting theory of psychology. You’re thrown into the world. This whole idea is very appealing to me. Just that whole story he tells — you’re thrown into the world, ready at hand versus present at hand. I think this idea of knowing how versus knowing that, different kinds of knowledge is a very interesting idea. Do you watch John Vervaeke?

COWEN: No.

You will find the (very interesting) tech segments all over the rest of the dialogue.  And I am happy to refer you all to the new paperback edition of Chris’s new book Read Write Own: Building the Next Era of the Internet.

Wednesday assorted links

1. Jon Hartley podcast with Robert Barro.

2. Dwarkesh questions about AI.

3. Why it is hard for women to date well.

4. “Social media constantly incentivizes disagreement, and therefore also demasculinization.

5. WEF update.

6. It is hard to type mathematical symbols.

7. Ezra talks with Steven Hahn about Trump and American continuity (NYT).  A little slow to get going, but has many important points.

Smith Reviews Stiglitz

Vernon Smith reviews Joe Stiglitz’s book The Road to Freedom:

Stiglitz did work in the abstract intellectual theoretical tradition of neoclassical economics showing how the standard results were changed by asymmetric or imperfect information. He is oblivious, however, to the experimental lab and field empirical research showing that agent knowledge of all such information is neither necessary nor sufficient for a market to converge to competitive supply-and-demand equilibrium outcomes.

Consequently, both the standard and the modified theories are irrelevant because buyers and sellers in possession only of dispersed, private, decentralized, value information easily converge to competitive price-quantity allocations in experimental markets over time via learning in repeat transactions.

…The first experiments, showing that complete WTP/WTA information was not necessary, were reported in Smith (1962), and none of us could any longer accept the standard and Stiglitz-modified theories. Further experiments, showing that such information was not sufficient, and that equilibrium prices need not require that markets clear, were reported in Smith (1965). (For propositions summarizing and evaluating observed empirical regularities in these experimental markets, see Vernon L. Smith, Arlington Williams, W. K. Bratt, and M. G. Vannoni, 1982, “Competitive Market Institutions: Double Auctions vs. Sealed Bid-Offer Auctions,” American Economic Review 72, no. 1, 58–77; and Vernon L. Smith, 1991, Papers in Experimental Economics, Cambridge: Cambridge University Press.) It was natural, in the first market experiments, to investigate those questions, such as the information state of traders, that were central to the abstract economic theory of the time.

So, the Akerlof-Stiglitz modifications of theory were founded on a false conditional and thus were not germane to practical market performance. They were born falsified.

…The needed policy implications are quite clear, and they have nothing to do with Stiglitz’s market failure and everything to do with how markets function. Indeed, the appropriate policy recommendation is to fully support the market-system maximization of prosperity, as did Friedman and Hayek, then use incentive mechanisms to improve the relative positions of those who are disadvantaged in that system. Never kill the goose that lays eggs of gold.