Yes, the idea that LLMs pay uniform attention to everything in a paper (or are "better at paying attention that humans") is a widely believed myth.

Financial crisis? "According to Moody’s, structured finance has become a popular way to pay for new data center projects, with more than $9 billion of issuance in the commercial mortgage-backed security and asset-backed security markets during the first four months of 2025." (Ian Frisch, NYTimes)

"Accenture, (...), in its most recent quarter reported a $100 million increase in new generative AI bookings quarter over quarter. That is down from a $200 million quarter-over-quarter increase the previous two periods." www.wsj.com/articles/how...

PSA if you travel to SMS-SF: I used the SMS-provided promo code on United and compared it with exactly the same flight without the promo code. The code made my airfare $40 MORE expensive. Nice anecdote for price discrimination or business ethics? www.colorado.edu/today/2025/0...

On the Charlie Kirk assassination: "Violence is the last refuge of the incompetent." (Isaac Asimov, Foundation) www.cnn.com/2025/09/12/u...

The LLM turning point is coming ( = we may be close to the "peak of inflated expectations" in the Hype Cycle). www.economist.com/business/202...

Have we learned nothing from conflicts of interest during the subprime housing crisis? "Both authors have a financial interest in www.expectedparrot.com. Horton is an economic advisor to Anthropic."

Excited to have my research (co-authored with the dream team of @mjyang.bsky.social, Zachary Kroff, and @erikbryn.bsky.social) featured in the New Yorker!

Incredibly excited for my brand new class at Rotman this fall: Progress! Econ history + theory + history of thought + philosophy on why rare orgs at rare times in rare places accomplish new things. Trying to put rigor onto an idea that is very much in the air. kevinbryanecon.com/Bryan-Progre... 1/3

Seems Sam Altman thinks "Yes", AI is in a bubble? Yet, he wants to let OpenAI employees convert stock to cash with an implied valuation of $500 Billion? (with a B!). What happend to getting "PhD level" ChatGPT-5 expertise, .. (checks notes) .. two weeks ago?!? www.theverge.com/ai-artificia...

There is an.....interesting op-ed in today's WSJ from a 22 year old entrepreneur telling everyone that work life balance will "keep you mediocre" It is completly wrong. Let me tell you why with citations. [gift link to the very wrong op-ed] 1/n www.wsj.com/opinion/work...

Odd, or signal that ChatGPT-5 is lower quality, but cheaper to run, because it decides when a high-token test-time compute run is triggered..

There are two Bs!

ChatGPT-5 is here.

You should read @noahpinion.blogsky.venki.dev on the data centre investments and whether they will lead to a financial crisis. I think he downplays the risk. It is potentially very high. www.noahpinion.blog/p/will-data-...

Disruption of the traditional MBA is here, whether we want it or not..

Consider: Google has a 90% market share in web and mobile searches. And has Gemini LLM results on top by default on any search. Are 1 trillion processed tokens really a sign of usefulness? (Same question about MSFT forcing LLMs on their developers..) We see "revealed preference", just not for LLMs.

I have been thinking about this a lot, lately, after finding the troubling productivity impacts of AI on older firms ( conference.nber.org/conf_papers/... ), combined with prior work showing the impacts of digitization on older workers (www.nber.org/papers/w28094). #EconSky

Naked self-promotion: My fantastic co-author @kmcelheran.bsky.social will present our empirical paper on "Industrial AI" (self-driving forklifts, self-optimizing production lines, autonomous quality control) and its effect on productivity in US manufacturing. www.nber.org/conferences/...

Its not me, its the LLMs: in domains including coding, math, mapping, logic and now simple physics, they overfit and are unable to generalize. Still waiting for ChatGPT-5 btw..

I agree that teachers (=experts) using LLMs has clear benefits. I think the evidence of benefits of LLMs for students is less consistent and more context dependent. LLMs are not a hoax, but also clearly over-hyped.

Part of the avalanche of recent papers showing incoherence of LLMs: @sendhil.bsky.social and coauthors show that LLM answers contradict their reasoning in the majority of cases, among 7 current LLM/LRM models. Another datapoint on "LLM reasoning is just pretense". arxiv.org/abs/2506.21521

Finished "Empire of AI" by @karenhao.bsky.social: important & compelling read for anybody worried about the broader social consequences of LLMs, even if we never reach AGI. I was initially skeptical about the "empire/colonialism" framing but now appreciate it. www.amazon.com/s?k=empire+o...

More evidence on the educational costs of LLMs: users underperform peers using traditional web-search. papers.ssrn.com/sol3/papers....

Yup, that OpEd reminds me of the recent MIT-brain study on LLM use..

Very impressive study showing that LLM-use reduces brain activity for an essay writing task, over a 4-month time horizon. Letting college students use LLMs can significantly reduce educational benefits, not increase them.

LLM-generated code is highly likely to generate cybersecurity risks. Yet, a majority of IT professionals say they believe that LLM-generated code is more secure than human code. Maybe because MSFT and others claim "up to 30%" of their code is LLM-based? Wait a sec.. youtu.be/NxOehfOfB6c?...

Sad to see, but true: LLM companies are systematically gaming leaderboards, to make LLMs seem smarter than they are. This is why the Apple paper from 2 weeks ago had so much impact: it is part of a string of work that shows the misleading claims of (leading) LLM companies. arxiv.org/abs/2504.20879

Interesting new developments on the Quantum Computing (QC) front: IBM reports that it developed new error-correction to cut the required physical qbits for one logical qbit by 90%. Commercially useful QCs might be with us by the end of the decade! youtu.be/ZNQIF_q3NVA?...

Yup, "LLM reasoning" is just pretense. But OpenAI, Google and Anthropic still charge extra for these "frontier" models. My question: What will LLM companies come up next to try and convince us that AI scaling does not run into diminishing returns?

Does #AI Prediction Scale to #DecisionMaking? The authors argue that there are fundamental limits to AI’s predictive paradigm. Authors: Mari Sako & Teppo Felin Read More: http://spkl.io/63325fAN9T

Interesting RCT from the World Bank on the impact of LLM usage in Nigerian high schools on English language proficiency. Effects are very large (6 weeks of 2x per week, 1.5 hrs ChatGPT use = 1.5 years of English lessons). Hallucinations obviously no problem here. Would this work with math?

Looks like consumer inflation expectations have lost their anchor. If the Fed does not aggressively increase rates, inflation will shift up. But if rates go up, we will hit a recession this year. Stagflation here we come?

Hard not to be optimistic, when you see the recent revolutionary advances in medicine: CRISPR and mRNA/siRNA related technologies could fully AND cheaply (!) heal a large number of rare "Orphan diseases", that are currently too unprofitable for big Pharma R&D. www.nytimes.com/2025/05/15/h...

It IS suprising that LLM hallucination rates have been increasing over the last year, especially given the prominent claims about "AI scaling", "Emergent properties" etc, through more data and more NVidia chips. This looks more like diminishing returns to me..

Hallucination rates in LLM reasoning models are off the charts and have gotten WORSE: OpenAI (not some Luddite) reports hallucinations for o3/o4 of 33%/48%. How is this consistent with all the hype for reasoning systems beating all these benchmarks? www.nytimes.com/2025/05/05/t...

Dear Bluesky: Trump is not running for a 3rd term. Trump: "I’ll be an eight-year president; I’ll be a two-term president. I always thought that was very important, to be honest with you." Can we please acknowledge this and move on? www.nbcnews.com/meet-the-pre...

Meta's bet: the open-source LLMs are like the open system of Microsoft in the 80s, while proprietary LLMs like ChatGPT are like Mac in the 80s. Key question: how valuable are complementor-users like academia, who need published weights like in Llama and will buy Nvidia GPUs? My (biased) sense: very.

This looks promising for Quantum Computing! Once the current uncertainty is over, we might see more progress on the path to Quantum Computers solving practical problems.. www.meritalk.com/articles/nsf...

I agree: LLMs are real and NOT a scam, with real, even tangible benefits especially for people who are already domain experts. However, LLMs are also over-hyped, as exemplified by the "reasoning" they show, which is mostly pretense. Just to say that doesn't make one a Luddite, either.

Great blog entry by @dmckenzie.bsky.social on the question of "How to manage governments better". If you read "Abundance" by Klein-Thompson, but want to know more about what the implications for government design are beyond "de-regulation", you will enjoy this excellent and thoughtful blog!

Successfully used Gemini-2.5-Pro for a coding task for which neither ChatGPT-o3-high-mini nor Claude-3.7 worked well. Google might really pull ahead here, despite the DeepSeek scare a couple of weeks ago. This is consistent with Clayton Christensen: LLMs are a "sustaining innovation" for Google..

#econsky: How likely is it that the upcoming tax cuts will stimulate the economy, if they are deficit-financed and Japan, China and the EU will not buy additional US treasuries?

Agreed, this looks promising!

Looking forward to the next @nber.org Organizational Economics Working Group meeting on April 10/11, coorganized with @andreaprat.bsky.social. We have a great line-up of new research papers in Org Econ! Program and livestream link: www.nber.org/conferences/...

Hopefully stocks will recover today, but keep in mind: Even if tariffs removed tomorrow, the political uncertainty triggered by the last week will continue to hurt the most innovative firms (startups AND corporations) the most! Lasting damage has been done, even if we don’t see it right away.

The logic behind Trump's trade war: IF we will have a trade-weighted US-$ appreciation (vs China, MEX), THEN (1) US consumer prices will not be affected, (2) tariff revenue will be raised, (3) the depreciation of CHN/MEX currencies means foreigners pay for the tariff revenues.

Erza Klein claims in his interview of Paul Krugman that LLMs came up with the 57% tariff on Chinese imports. Not really. When prompted ChatGPT-4.5 proposes a tariff of 20.5%, ChatGPT-o3-mini 13.3% and Claude-3.7 a tariff of 40%... NONE use bilateral trade in their math.

If you keep hearing "high tariffs are a negotiating tool", but keep wondering "to what end??". Answer: to push a change to global financial markets that could trigger the next financial crisis.. youtu.be/iEF1ZqZtS9I?...

What is the most reasonable case for the current trade war? See this ambitious blueprint for a new "Mar-a-Largo" global system, by CEA chair Stephen Miran. Key point: bully all major economies to de-value the US-$, while keeping it a reserve currency. www.hudsonbaycapital.com/documents/FG...