Gen Z Musician First Time Reaction to Chicago - 25 or 6 to 4

This is a reaction video, a minor interest of mine these days. As the title says, we have a Gen Z musician reacting to a Boomer hit, Chicago's "25 or 6 to 4." This particular tune interests to me because I performed it many times back in the Ancient Days with the Saint Matthew Passion, a jazz-rock band in Baltimore based out of Johns Hopkins. Andiroo is himself a musician, keyboard and producer, so he loves making technical comments about the music. He obviously loves this tune, the intricate craftsmanship of it, as though that kind of intricate craftsmanship was a rare commodity in music.

To me, and the guys I played with, that's just how music is, and none of us were professional musicians. We were just college kids – well, young adults – performing music we loved. None of us regarded our own technical skill as particularly evolved, not on the level of Chicago. But we performed this tune, and others, well. To me, and many musicians I've played with since then, that's simply what music is about. No (particularly) big deal.

I've not forgotten [irises]

Claude’s further observations about evolution [MR-AUX]

Here’s some more supplementary material from my investigation of Tyler Cowen’s recent monograph, The Marginal Revolution: Rise and Decline, and the Pending AI Revolution (2026). First I have a short note about the discovery of marginal utility and the discovery of evolution. Then I have some more discussion about the conceptualization of evolution, starting with locating phylogenetic identity in the gene pool. Later on there’s a discussion of second order patterns which is important to epistemology. After all, evolutionary biology was not a mathematicised discipline in Darwin’s time. Economics and statistical mechanics got their Rank 4 “rigidity” from mathematics. What was the source of rigidity for evolution? After than we talk about the Galapagos finches and conclude with some discussion of cognitive ranks conceptualization.

* * * * *

Why the discovery of marginal utility is like the discovery of evolution

[Note: This section is a continuation of a previous dialog. The prompt that originated that dialog is irrelevant. These three paragraphs are Claude’s.]

The imaginative leap you identify is the specific cognitive achievement that makes the Rank 4 move possible. You have to be able to hold in mind simultaneously two things that are in tension: the real difference between species — the fact that a fish and a mammal really are different kinds of things — and the theoretical continuity across those differences that the genealogical model posits. This requires treating the continuity not as something observed but as something inferred from a model, and committing to the model's reality even though it posits entities and processes — ancestral populations, selection pressures acting over millions of years — that are not directly observable.

This is structurally identical to the move marginalism makes. The marginal utility of water to a person who already has plenty is not directly observable. It's a theoretical posit — an entity defined by the model — that makes sense of observable price behavior. Both Darwin and Jevons are positing theoretical entities that are real in the model but not directly accessible to observation, and using those entities to explain patterns in the observable world that the previous descriptive framework couldn't account for. That shared structure is why they're both Rank 4, and why they appear in the same half-century window.

Cowen's inattention to the identity problem is a genuine gap in his account. He treats the emergence of evolutionary theory as primarily a matter of accumulating observations and finding the right conceptual key — the Malthusian insight about population pressure, the geological uniformitarianism that provided deep time. These are real and important. But he doesn't attend to the prior conceptual problem of what kind of object the theory needs to posit — the genealogical continuum across real morphological difference — and why that object required a cognitive move that Rank 3 natural history, for all its sophistication, was not equipped to make.

The problem of phylogenetic identity

Going back to the conceptual problem of identity through the phylogenetic line, just what is it that establishes that identity? I’m not asking about observations, experiments, and inferences, I’m asking about physical stuff? It is the gene pool, a continuous line of genetic transmission?

Yes, and stated that precisely it's a beautiful answer to the identity problem — because it transforms what looked like a philosophical puzzle about categories into a physical fact about causal continuity.

The gene pool is a continuous physical process. Genes are replicated and transmitted from generation to generation through actual molecular mechanisms — copying of DNA, meiosis, fertilization. The chain of physical causation is unbroken. What connects a fish ancestor to its mammalian descendant, across hundreds of millions of years and radical morphological transformation, is not conceptual continuity — not our decision to classify them as related — but a physical causal chain of genetic transmission that actually happened, molecule by molecule, generation by generation, without interruption.

This is what makes Darwin's theory so much more powerful than Lamarck's, in retrospect. Lamarck had the right general idea — species transform over time — but no physical basis for the continuity. The inheritance of acquired characteristics is a conceptual story, not a physical mechanism. Darwin had natural selection but also lacked the physical mechanism — he didn't know about genes, and the blending inheritance he assumed was actually a problem for his theory, as Fleeming Jenkin pointed out. The physical grounding came only with Mendel, rediscovered in 1900, and the Modern Synthesis of the 1930s and 40s that fused Darwinian selection with Mendelian genetics.

What this does to the identity problem

The philosophical puzzle — how can radically different organisms be the same thing? — dissolves when you specify that the identity is physical and causal rather than morphological or essential. The fish and the mammal are not the same kind of thing in the Linnaean sense — they don't share the properties that would put them in the same category under any synchronic classification scheme. But they are connected by an unbroken physical chain of genetic transmission, which is a different and more fundamental kind of identity.

This is precisely the Rank 4 move applied to the identity question. The Rank 3 naturalist asks: what kind of thing is this organism? The answer is given by its observable properties, organized by the classification scheme. The Rank 4 move asks: what process generated this organism, and what connects it to other organisms through time? The answer is given by the physical mechanism of genetic transmission, which is not observable in the organism itself but is inferred from the theory and confirmed by the molecular biology that came later.

The depth of the conceptual achievement

What this means is that Darwin was positing an identity grounded in a physical mechanism he couldn't see and didn't fully understand. He knew that offspring resemble parents — that heritable variation is real — but he didn't know why. The physical substrate of the identity he was positing — the gene pool as continuous causal chain — was invisible to him. He was inferring the existence of a physical continuity from its population-level effects, which is an extraordinarily bold theoretical move.

This is analogous to what the marginalists were doing with utility — positing a theoretical entity that explains observable price behavior without being directly observable itself. In both cases the Rank 4 achievement consists precisely in positing the right kind of unobservable entity: one that is physically real, causally efficacious, and theoretically tractable, even though it can only be approached indirectly through its effects on observable phenomena.

The gene pool as the physical basis of phylogenetic identity also resolves the tension you identified — that the continuity is in a sense an illusion, that the different life forms really are different. They are morphologically different, behaviorally different, ecologically different. But they are physically continuous through the causal chain of genetic transmission. The continuity is not projected onto them by our conceptual scheme — it's a real physical thread running through the apparent discontinuity of form. Darwin intuited this without being able to state it at the physical level. The Modern Synthesis stated it. Molecular biology confirmed it and made the physical chain directly readable in DNA sequences, where the evidence of common descent is written in the genome itself.

The Sphinx riddle now has a deeper answer than the one Oedipus gave. What persists through the morphological transformations of the phylogenetic line isn't a Platonic essence or a conceptual category. It's a physical process — the replication and transmission of genetic information — that is as real and continuous as any other causal chain in nature. The identity is not constructed by the observer. It was there all along, in the chemistry.

Darwin’s leap of faith

No matter how you look at it, Darwin had to take a leap of faith to arrive at this conclusion. Nothing else made any sense, but still.... He was unaware of Mendels’s experiments, but it wasn’t until the mid-20th century that we began to understand the molecular mechanism. I believe Watson and Crick published their discovery in 1953.

Yes, Watson and Crick published the double helix structure in 1953 — the famous one-page paper in Nature, with its almost comically understated final line noting that the structure "suggests a possible copying mechanism for the genetic material." One of the great understatements in scientific history.

And your point about the leap of faith is important — and connects to something deep about how Rank 4 thinking works. Darwin committed to a theoretical structure — genealogical descent with modification through natural selection, grounded in a physical continuity of inheritance he couldn't specify — on the basis that nothing else made the observed patterns cohere. The fossil record, the geographical distribution of species, the nested hierarchies of Linnaean taxonomy, the embryological similarities across radically different adult forms — all of these independently pointed in the same direction. The convergence of independent lines of evidence was so strong that the theory had to be right even though the physical mechanism was unknown.

This is actually a recurring feature of great Rank 4 achievements. Newton committed to universal gravitation as a real physical force without being able to say what it was or how it acted at a distance — he famously refused to speculate, hypotheses non fingo, but the mathematical structure was so powerful and predictively successful that he committed to it anyway. The physical understanding of gravity came with general relativity, more than two centuries later. Thermodynamics was developed and successfully applied throughout the 19th century without anyone understanding what heat actually was at the molecular level — that required the statistical mechanics of Boltzmann, and even then wasn't fully accepted until the early 20th century.

From July 4, 2025

After 1990 AI stopped citing work in psychology [empirical evidence]

I'm bumping this post from 2024 to the top of the queue as it testifies to the process by which AI has converged on the intellectual monoculture it has become in the wake of ChatGPT.

* * * * * 

I’ve frequently noted that, while researchers in artificial intelligence (AI) and machine learning (ML) often have a lot to say about when their machines will approach, overtake, and even surpass human intellectual achievement, they don’t seem to know much about psychology, linguistics, and the cognitive scientists. I made an explicit argument at some length in a recent article I published in 3 Quarks Daily, Aye Aye, Cap’n! Investing in AI is like buying shares in a whaling voyage captained by a man who knows all about ships and little about whales. In making the argument the only evidence I present is anecdotal – Geoffrey Hinton and Ilya Sutskever in that article, though my beliefs on the issue are based on my reading of the current literature, which is opportunistic and by no means ‘complete,’ which, in any case, would be impossible as the literature is so large.

Now I can present a bit of systematic empirical evidence in the matter. M.R. Frank et al. undertook a bibliometric investigation of citation patters in AI and other disciplines and discovered that, while in the early years, AI interacted with other fields quite a bit, that interaction dropped off over the years. The following chart shows how AI cited other fields:

Its citation of psychology peaked in the middle 1960s and then dropped off steadily until 1990. Its citation of mathematics rose steadily through the period. That’s understandable; I have no complaint about that. The drop in citations to psychology is also understandable, but somewhat more problematic. For it implies that, when AI experts offer judgements about human cognitive capabilities, whether directly or indirectly through comparison with AI, that don’t know what they’re talking about. I suppose that last clause is a bit harsh. Perhaps it would be a bit more accurate to say something like: They don’t know any more than a bright college sophomore who’s taken a psych course or two.

Here's the article and abstract:

Frank, M.R., Wang, D., Cebrian, M. et al. The evolution of citation graphs in artificial intelligence research. Nat Mach Intell 1, 79–85 (2019). https://doi.org/10.1038/

As artificial intelligence (AI) applications see wider deployment, it becomes increasingly important to study the social and societal implications of AI adoption. Therefore, we ask: are AI research and the fields that study social and societal trends keeping pace with each other? Here, we use the Microsoft Academic Graph to study the bibliometric evolution of AI research and its related fields from 1950 to today. Although early AI researchers exhibited strong referencing behaviour towards philosophy, geography and art, modern AI research references mathematics and computer science most strongly. Conversely, other fields, including the social sciences, do not reference AI research in proportion to its growing paper production. Our evidence suggests that the growing preference of AI researchers to publish in topic-specific conferences over academic journals and the increasing presence of industry research pose a challenge to external researchers, as such research is particularly absent from references made by social scientists.

A toy to think with: The Erector Set

What Happened to the Erector Set? | The Toy That Taught Every Boy in America to Build

From the YouTube page:

For generations, the Erector Set wasn’t just a toy — it was a hands-on introduction to engineering, creativity, and real-world building.

Long before modern STEM kits and digital learning tools, children across America learned how to think, design, and construct using metal beams, bolts, and gears. Created by A. C. Gilbert in the early 1900s, the Erector Set became a symbol of innovation and imagination, inspiring future engineers and inventors.

In this video, we explore the complete story of the Erector Set — from its invention and rise during the golden age of American toys, to its cultural impact and eventual decline in a rapidly changing world of plastic toys and digital entertainment.

Discover how it competed with brands like LEGO, why its popularity began to fade, and what ultimately happened to one of the most educational toys ever created.

From steel structures to childhood dreams, this is the story of the toy that taught a generation how to build.

If you’ve ever wondered what happened to the legendary Erector Set… this nostalgic deep dive reveals it all.

Three up close

On Steven Spielberg, three vignettes

Wesley Morris, Why We Still Need Steven Spielberg, NYTimes, June 7, 2026.

The cultural importance of movies:

Lately, the idea of a Steven Spielberg has felt endangered. For more than 50 years, his imagery has epitomized American movies, maybe even epitomized America. He has been at the center of an industry that, if it’s not dying, is certainly diminished. The sort of original movies that made Spielberg Spielberg are virtually nonexistent, even though the two major flavors that now define the industry — global box-office smash and best picture nominee — are, with Spielberg, indistinguishable (start with “Jaws,” “Raiders of the Lost Ark,” “E.T. the Extra-Terrestrial”). More than once, he inhabited both modes within one calendar year: “Jurassic Park” in the summer of 1993, for instance, then “Schindler’s List” at the end of Hanukkah, perhaps the most triumphant single-year change-up any Hollywood director has had. (He’s still the most commercially successful director ever, and he’s tied, at 13, with William Wyler for directing the most best picture Oscar nominees.)

Popular art has always bonded us to one another, no matter what might have been cleaving us apart, no matter how different our lives or how our responses to that art diverged. And Spielberg’s films have been a premium adhesive. Not only the ones he directed but the dozens of swooshing, indelibly kooky hits unleashed by Amblin Entertainment, his production company: “Poltergeist,” “Gremlins,” “The Goonies,” the “Back to the Future” trilogy, “Who Framed Roger Rabbit,” “Arachnophobia.”

Spielberg’s stardom arose from the collision of capitalism, audacity and creative vision. His movies emerged alongside the arrival of cable television and proliferating advances in personal computing and home entertainment. I watched “E.T.” at the movies, devoured it on cable, played it on my Atari and let Michael Jackson sing me a lullaby the movie inspired him to write. (Spielberg: so titanic that the other king of pop worshiped his thrillers.)

But a kind of cultural malnourishment has set in. While you once needed a pair of hands to count the major studios, we’re on the verge of barely needing one. And the best, most lucrative ideas entail microwaved nostalgia that we all know by its legal nickname: I.P. The takeovers and reheating, the obscure metrics that ensure we never quite know exactly how popular anything is, it’s dispiriting: Pac-Man eating ghosts, algorithms keeping secrets.

Working through his parent's divorce:

“For years, I was working out my mom and dad’s divorce through my stories,” he told me. The split happened when Spielberg was about 15, but the marriage had begun to fade years before. The family dissolved, too. He went off to live with his father, Arnold, a computer engineer, in Los Angeles, while his three sisters remained in Phoenix with their mother, Leah, a classical pianist who used to operate a kosher deli. But it seems that living under the same roof didn’t significantly change how remote Arnold could feel to his son.

Over dinner one night, Spielberg told me about working on “Indiana Jones and the Last Crusade” with George Lucas. The movie reunites Harrison Ford’s Indiana with his father, a Holy Grail expert played by Sean Connery. “My contribution was: ‘OK, but I want to meet Indy’s dad, and I want them to have had years of estrangement and father neglecting son because the father was a workaholic. And this story will bring them back together again.’” When Spielberg said this, he still sounded wishful and a touch sad. Watched through the lens of his childhood, his movies can seem newly forlorn, someone blowing on a birthday cake gated with stubborn candles.

For centuries, we’ve lived with a myth that genius — male genius — expresses itself as wild eccentricity or madness, that the personality warrants a cult or a harem. Spielberg disorients in that regard. I, at least, needed a moment to absorb how familiar he felt, how familial. The man who made “E.T.” was eerily reminiscent of the woman who took me to see “E.T.” Both of them have in common a special intuition to anticipate needs we don’t know we have. My mother did it for a household. For more than half a century, Spielberg has been doing the same for a planet. [...]

Spielberg hasn’t been to see a therapist since he was in college. Instead, the movies are the arena in which he has worked on some of the mysteries he couldn’t solve on his own. What we experience as sorcery is, for him, a process of exorcism. “I can’t express enough how therapeutic and healthy it is for me to keep doing this job over and over and over again,” he said deliberatively, almost as if he was feeling this out. “I work so much out through this process. So much out. I get to bleed off some of the darkness instead of letting it fester inside me. You get to let it fester inside you.”

A six year old:

As we were getting ready to head over to “Oh, Mary!” Spielberg told me a story about the time his mother’s brother, Bernard, took him and his cousin, Paul, to visit the Lincoln Memorial. It was 1952 or ’53. He would’ve been 6. The three of them climbed the steps. “Suddenly I was standing at the foot of a scary giant,” he recalled. “I remember glancing up and being so terrified I could only look at the hands.” He fixated on how they “were overhanging the armrests” and felt the urge to flee. But something held him back. “When I turned around, I looked up at his face. At this statue. Of Lincoln. A calm washed over me. An instant connection washed over me.” His fear ceased. What arrived in its place was unabated curiosity. He began to read all about Lincoln and started making silhouette cutouts of him, an obsession you can see re-enacted in the opening scene of “Minority Report,” when a child makes a paper mask of Lincoln.

There's much more at the link.

Terence Tao Explains The Math Behind AI

On the YouTube page:

Terence Tao has read more mathematics than almost anyone alive, and he uses AI tools every day. So when one of the most cited mathematicians on Earth says these systems still can't ask a genuinely new question, it's worth understanding exactly where he draws the line — because it isn't where the headlines put it.

Watch the full conversation: Terence Tao: Nobody Understands Why AI Actually Works

If AI has absorbed every textbook ever written, why can't it discover anything new? Tao, a Fields Medal winner and professor at UCLA, separates what these systems do brilliantly from what they can't do at all, and the boundary turns out to be sharper and stranger than most people assume.

We cover why reproducing a famous proof is less impressive than it sounds, what a neural network found hidden inside a million knots that humans had missed, why we still can't predict which tasks AI will actually be good at, the "Keating Test" — the benchmark that would actually demonstrate machine thought — and where exhaustive recall ends and real conceptual origination begins.

AI can pass every exam. It just can't ask a question nobody has asked before — yet.

Chapters:
00:00 The question AI can't ask
00:48 Read every textbook, discover nothing
01:42 Why a reproduced proof proves less
02:39 A million knots, one hidden pattern
03:54 The competence we still can't predict
05:11 The Keating Test for machine thought
06:18 Where recall ends and discovery begins

📬 Get the transcript, fascinating bonus content, and my Monday M.A.G.I.C. Message: https://briankeating.com/yt

Demis Hassabis and Yann LeCun on Computational Compressibility

Dædalus currently has a double issue, AI & Science: What Is the Future of Discovery?, edited by James M. Manyika. Manyika interviews both Hassabis and LeCun and they offer remarks relevant to the issue of computational compressibility as I discussed it in my recent working paper, On Method: Computational Compressibility in Complex Natural and Cultural Phenomena, though they don’t use the term. Here are some passages from those interviews.

Demis Hassabis

In this first passage Hassabis is talking about a well-known problem in computer science, known as P versus NP, which is about how long it takes to solve a problem as a function of the size of the input. Roughly speaking, what’s at stake goes like this: If you are presented with a proposed solution and can verify it quickly, could you also find the solution from scratch relatively quickly (in polynomial time, P) or is finding a solution so difficult as to be all but impossible (NP, Nondeterministic Polynomial time). You don’t need to understand that to understand this passage , pp. 36-38:

It does, and I think those are the interesting limits to test and understand. P equals NP–which attempts to categorize the difficulty of a problem by how much computation it would take to find and check a solution, respectively–is one of the most important questions in science to resolve. I suspect P is not equal to NP, and there are some problems out there that are just not tractable to solve in a practical amount of time without invoking the help of, say, a quantum computer, but we need to understand this a lot better because there may be more nuance here than we previously realized. In our work with AlphaGo and AlphaFold, we’re showing that if you do a lot of precompute, which is not normally considered in these kinds of scenarios, you can seemingly answer some highly complex questions approximately optimally in P (polynomial) time. Neural networks are effectively using massive amounts of precompute to compress knowledge into some efficient artifact. That computed artifact is then available at test time and, for a lot of natural systems, you can use it to narrow down your search space so you don’t have to consider all the possible configurations they could potentially take, but only a much smaller subset that are actually plausible.

Those last two sentences are about computational compressibility. Hassibis then goes on to illustrate:

Let’s take proteins. There are roughly 10300 possible conformations of an average protein. It would take longer than the age of the universe to enumerate that exhaustively to find the one specific shape it takes, so you have to do something much smarter. You have to learn what patterns there are for different amino acid sequences and then only search a tiny fraction of the possibilities to find the approximately correct solution. That seems to be what we managed to do with AlphaFold. Maybe not perfectly, but to an approximation that is at least good enough for practical purposes. [...]

AlphaFold was our solution to the protein folding or protein structure prediction problem. You start with an amino acid sequence–you can think of it very roughly as the genetic sequence for the protein, a one-dimensional string of letters. In the body or in nature, that string folds up into a 3D structure, and that shape goes a long way toward defining the function of that protein, which is really important for drug discovery and disease understanding.[...]

The way we did it is that there were about 150,000 known structures that had been painstakingly put together by structural biologists over the past thirty to forty years with very expensive equipment like electron microscopes. That was just about enough data to give our AI system clues as to the topology of proteins. Of course they don’t just fold up randomly; there are some constraints, and the AI system learned them. Eventually it was able, within a few seconds, to come up with a plausible structure for an unseen protein.

In this next passage, the first two conditions are about compressibility, p. 39:

We look for three aspects of a problem in determining whether it is suitable to tackle with the AI techniques we have today. First, can the problem be described as or converted into a description of a massive combinatorial space? Perhaps it’s intractably large and normal brute force techniques won’t work. Second, if that’s true, do you have enough data to learn some sort of model of the topology of that space? Or maybe a simulator is available or learnable that can generate some additional synthetic data. Ideally, you have both. Third, you need a clear objective that you’re trying to minimize or maximize. In games, that is winning or maximizing the score. In a natural system, that might be minimizing the free energy in that system. If you can quantify that, you can then use a model to search with the guidance of the objective function toward the optimal solution.

Yann LeCun

In the following passage LeCun talks about an abstract representation space. That space contains a compressed representation of the phenomenon, pp. 47-48

I think this is a crucial point and is what I am presently devoting all of my efforts to: devising AI systems that can find an abstract representation of the phenomenon and make predictions in that abstract representation space. This abstract representation eliminates a lot of details about the original observations. And that’s a crucial point because LLMs (large language models) and other generative models are trained to predict every detail of the input. In language, it’s not too much of a problem. You cannot predict exactly which word follows a sequence of words, but you can produce a probability distribution over words. That’s easy because there’s a finite number of possible words. But when you train the model to predict future frames in a video, you can’t represent a useful distribution. You have to make predictions in an abstract representation space, not at the pixel level. So a lot of people in the last few years instinctively said, “let’s just tokenize the world.” Let’s take images from videos and cut them into little squares and turn that into a vector that doesn’t look different from the one that represents a word, and feed this to a gigantic model to predict the next few frames. Frankly, it doesn’t work that well. The reason why is that you simply cannot predict what’s going to happen in a video at the pixel level. There are so many details that are just not in the input. We don’t know how to produce a probability distribution over all possible video frames because it’s mathematically intractable. It’s a problem people have struggled with for decades in statistical physics.

Instead, what we do as scientists is to find a representation of the input that eliminates all the details we cannot predict, and we make predictions in that representation space. That’s not a generative architecture.

Later, p. 55:

Manyika. Given the advances in AI, and particularly if we go beyond human cognitive levels and AI systems come to understand more than we do, what are the implications for philosophy of science, how we do science, and the nature of scientific understanding?

LeCun. I think that question is not a new one. When we solved PDEs (partial differential equations) numerically with computers, did the computational fluid dynamics simulator understand physics better than we did? It can make a prediction and it’s using an algorithm based on equations that humans came up with.

The next step AI enables is training a machine-learning system to make predictions from data without the manual step of reducing the process to equations. AI allows us to skip having to first build a model of reality that can then be computed. This is powerful because many phenomena in science are collective complex phenomena.

That is the compression step. LeCun continues:

A pile of sand behaves in a particular way, and the theory for this is not entirely clear. The property of materials, particularly complex ones, cannot be directly derived from the elementary equations of quantum mechanics. It’s just too complicated. Another example is the magic angle, 1.1 degrees, at which you rotate two stacked monolayers of carbon, called graphene, to form a superconductor. That’s a collective phenomenon that is extremely difficult to explain. There are various properties of materials of this type that cannot be usefully reduced to a small number of equations from which you can derive this collective behavior. How does intelligence emerge from neurons in interaction? That’s a philosophical question of how a super complex property like intelligence can emerge from a large number of relatively simple elements in interaction, but that’s a pretty high-level thing. At a lower level are questions of how life emerges from the interaction between proteins. This transition is what has baffled scientists for a long time: the transition from the microscopic to the mesoscopic. This is where interesting things happen, like life, for example.

So now there’s a new way of doing science, which is neither completely qualitative and observational nor reductionist, but is a data-driven, AI-powered phenomenological model that may allow us to bridge the gap between microscopic and macroscopic.

That last paragraph is about compression.

Two tablescapes (without food)

Dealing with the Devil, John Malone and American business

Matt Stoller, A Billionaire Explains Why American Business Now Feels like the Mafia, BIG, June 5, 2026.

The article opens:

In 1981, a consultant named Elmer Smalling, an expert in pay-TV systems, was negotiating on behalf of Jefferson City, Missouri, to see about better prices for residents. Across the table was an executive for a giant corporation named TCI, which had 25% of the U.S. cable market and was known for the hard-charging tactics of its CEO, John Malone. The negotiations were not going well, because the city was thinking of taking away the franchise and going with someone else. Paul Alden, one of Malone’s subordinates, lived up to that reputation.

“We know where you live, where your office is and who you owe money to,” he told Smalling. ”We are having your house watched and we are going to use this information to destroy you. You made a big mistake messing with T.C.I. We are the largest cable company around [.] We are going to see that you are ruined professionally.”

The bulk of the article is a review of Malone's autobiography, Born to Be Wired: Lessons from a Lifetime Transforming Television, Wiring America for the Internet, and Growing Formula One, Discovery, Sirius XM, and the Atlanta Braves, 2025.

Stoller observes:

And this book, while not honest, is as close as I’ve ever seen to getting to the core of how the billionaires who took over American society really think. Malone’s book helped me understand the generation of media billionaires, before the tech oligarchs, who had to contend with the dying embers of New Deal regulations. And they knew a world where it wasn’t ok to do what they were trying to do, and yet they did it anyway, with energy, creativity, and a malevolent zeal to make the world safe for capital.

I'll give you three paragraphs from the review:

At age 27, he presented to the AT&T board of directors something that was then novel in corporate governance. He argued for them to buy back their own stock, which would reduce the total number of shares and thus increase earnings per share. The Chair of AT&T at the time, Fred Kappel, rejected the suggestion, while congratulating Malone on a fabulous presentation. Kappel told him, “If in your whole career, you can do a single thing that changes the Ma Bell system in even the smallest way, you would be very successful.”

It was a jarring moment for Malone. Though a rising star at AT&T, he realized he’d have to leave Bell Labs because it wouldn’t satisfy his need for control. But it’s also, unwittingly, the moment we see a young aggressive corporate leader’s reaction to that of an executive who was genuinely engaged in caretaking of a vast enterprise. Bell Labs had violated what would become a cardinal rule for Malone, which is that its leaders valued the corporation itself over the efficient use of capital in maximizing returns. Malone told his colleagues that Bell Labs, the legendary scientific lab, was a bureaucratic mess, a dinosaur, headed for extinction.

When he left Bell Labs for McKinsey, Malone writes as if it is an act of rebellion. “I wondered whether I had just jumped from the biggest, safest vessel I would ever board,” he wrote, “I felt ill.” This kind of odd faux rebellious streak courses through the book.

That middle paragraph is about a man who's sold his soul to Homo economicus.

There's much more in the article.

In the park looking out

The American Way of War

I've got a short series of posts from awhile back on the American way of war. I decided to add to it with this short conversation with Claude.

* * * * *

American wars since the collapse of the USSR

Give me a chronological list of the wars that America has been involved in since the collapse of the USSR.

The project knowledge touches on this theme thematically but doesn't provide a factual list. Here's a straightforward chronological account from general knowledge, using 1991 (the USSR's formal dissolution) as the starting point:

Gulf War (1991) — Already underway when the USSR collapsed in December; the US-led coalition expelled Iraq from Kuwait. A short, decisive conventional war.

Somalia (1993) — Humanitarian intervention that escalated into urban combat, culminating in the Battle of Mogadishu ("Black Hawk Down").

Haiti (1994) — Military intervention to restore the elected government of Jean-Bertrand Aristide.

Bosnia (1995) — NATO air campaign (Operation Deliberate Force) in which the US played the lead role, ending the Bosnian War.

Kosovo (1999) — NATO air war against Yugoslavia to halt ethnic cleansing; again US-led.

Afghanistan (2001–2021) — Invasion following 9/11, ousting the Taliban; the longest war in American history.

Iraq (2003–2011, then resumed 2014–) — Invasion that toppled Saddam Hussein, followed by a prolonged insurgency. US forces returned in 2014 to fight ISIS and remain in smaller numbers today.

Libya (2011) — US participated in NATO air operations that led to Gaddafi's fall.

Syria / ISIS (2014–present) — Air campaign and special operations forces deployed against the Islamic State across Syria and Iraq.

Iran (2020) — The drone strike killing General Qasem Soleimani; a targeted killing rather than a sustained war, but a direct military act against a state actor.

Yemen (ongoing, indirect)** — The US has provided intelligence, refueling, and arms to the Saudi-led coalition, with direct drone strikes against Al-Qaeda in the Arabian Peninsula.

Iran (2025) — You noted in your War Series writing (with a 2026 addendum) that a Trump-era attack on Iran occurred, which fits this list as the most recent major escalation.

A few observations relevant to your own analytical framework: this list illustrates the pattern you describe in your War Series — the post-Cold War redirection of American aggression outward, particularly toward the Arab/Muslim world as the new "Other" following the loss of the Soviet scapegoat. Nearly every conflict after 1993 takes place in the Islamic world, which tracks your 2014 addendum about the "well-crafted Orientalist mythology of the exotic and crazy Arab Other."

Talcott Parsons 1947

At the heart of that War Series is a 1947 essay by Talcott Parsons, “Certain Primary Sources of Aggression in the Social Structure of the Western World.” I read that early in my undergraduate years at Johns Hopkins. I regard it as speculative, plausible but speculative. How would one gather empirical evidence on it? It’s not like any nation ever announces, “We’re going to war to let off some steam and ease tensions at home?”

That's exactly the right epistemological challenge — the mechanism Parsons proposes operates at the level of collective unconscious motivation, which is precisely what makes it both compelling and resistant to direct verification. But there are several indirect empirical strategies that could build a cumulative case.

Quantitative correlational approaches

The most tractable starting point is looking for statistical relationships between domestic stress indicators and the initiation of military action. You'd want to examine whether wars or significant military escalations cluster after periods of elevated internal tension — economic downturns, spikes in domestic unrest, rising inequality, political polarization indices. The null hypothesis would be that war timing is random with respect to domestic conditions; a systematic pattern would be suggestive. Turchin's work on political stress cycles (his "cliodynamics" project) is actually well-suited to this — he's already built quantitative indices of internal instability across centuries, and one could ask whether war-initiation correlates with his stress peaks.

The "rally around the flag" literature

There's actually a substantial body of political science research on this, though it approaches the question from a slightly different angle. The "rally" effect — the well-documented spike in presidential approval following military action — shows that war does relieve domestic political pressure in measurable ways. John Mueller's work on war and presidential popularity is foundational here. The harder inferential step is from "war relieves pressure" to "anticipated relief motivates war-making," but the effect at least confirms the direction of the mechanism.

Elite communication analysis

Parsons' theory doesn't require that anyone consciously decides to go to war for psychodynamic reasons — the motivation can be largely unconscious and rationalized afterward. This actually makes elite rhetoric more useful as evidence, not less. You'd look for patterns in how war is framed domestically — whether war rhetoric systematically spikes domestic us-versus-them language, whether it correlates with reduced internal scapegoating (your Klinkner & Smith thesis about civil rights gains during wartime), and whether the timing of hawkish escalation by elites tracks with rising domestic discontent. Content analysis of congressional debates, newspaper editorials, and presidential speeches across the full arc of conflicts could reveal structural patterns invisible to any individual case.

The comparative/historical method

Parsons' thesis should predict that nations with other mechanisms for handling internal aggression — stronger corporatist institutions, more robust social contracts, functioning scapegoats already in place — should be less prone to foreign military adventurism. Scandinavian nations, for instance, would be a useful comparison class. You could also look at cases where domestic conditions that should predict war didn't produce it, and ask what was different. The comparative method won't confirm the mechanism, but systematic cross-national patterns would be hard to explain otherwise.

The Klinkner-Smith natural experiment

The civil rights thesis is actually the most testable piece, and as you note in your writing, they've already done substantial work on it. Their finding — that Black Americans' civil rights gains cluster tightly around periods of external military threat — is exactly the kind of pattern that's hard to explain without something like Parsons' mechanism. It's not proof of the psychodynamics, but it's consistent evidence that war functions to redirect internal aggression in measurable ways.

The fundamental limitation

None of this gets you to the causal mechanism Parsons actually proposes — that repressed aggression generated by Western child-rearing practices flows through nationalist scapegoating and finds periodic release in war. That specific chain is probably permanently beyond empirical reach, because the mediating variable (unconscious aggressive impulse at civilizational scale) isn't directly observable. What you can do is show that the functional relationship holds — that war operates as if it were serving this purpose — and that alternative explanations (pure rational calculation of national interest, for instance) leave systematic residuals.

Which is, frankly, roughly where Freudian-influenced social theory in general sits: plausible, structurally coherent, supported by functional evidence, but not falsifiable in any clean Popperian sense. The honest assessment is that Parsons is probably tracking something real but operating at a level of abstraction that empirical social science can approach but not fully close.

Commencement speakers: Out to lunch & out of touch

Molly Jong-Fast, Why Those Commencement Speakers Deserved Those Boos, NYTimes, June 5, 2026.

Commencement address season hasn’t been going well — for the commencement speakers. [...]

When Eric Schmidt, a former chief executive of Google, told graduates at the University of Arizona about their A.I.-shaped future, the shouting got so intense that he paused and said that graduates feared “that the future has already been written, that the machines are coming, that the jobs are evaporating, that the climate is breaking, that politics are fractured, and that you are inheriting a mess that you did not create.” Mr. Schmidt told them to make the best of it. “The question is not whether A.I. will shape the world. It will. The question is whether you will help shape artificial intelligence.”

Mr. Schmidt’s solution to world-upending technological change is … what? To pull yourself up by your bootstraps? His approach is peak billionaire brain, directed at the young people who have, for the better part of a decade, been treated as woke, lazy, avocado-toast-eating snowflakes. All these speakers just don’t get it. The problem isn’t woke; the problem is work. It’s a lack of social mobility. It’s that college may no longer elevate a graduate to the middle class. It’s that nobody even bothers to pretend that a house, a good job and the ability to start a family are at all guaranteed.

Think of this from the graduates’ perspective: Wealthy old people telling you your future is being pulped by acres and acres of electricity-sucking, water-guzzling data centers feels dystopian because it is. Companies are trying to automate your future away. No wonder you’re furious.

The truth about AI:

Right now, A.I. is in its dark hype period — great for Anthropic’s I.P.O. — but who knows how useful any of this actually will be in the end in creating efficiencies (a.k.a.: replacing the youngs with bots). It’s within young people’s power to stop. Demand regulation of tech companies. Elect people who will legislate that regulation. Organize against data centers in your hometowns.

Don’t just boo — do something.

There's more at the link.

11th St., a red flower, and Washington St. [Hoboken]

Microsoft AI builds a “hill-climbing” machine

* * * * *

The Microsoft AI Team, MAI-Thinking-1: Building a Hill-Climbing Machine

Abstract: Progress in AI is driven not by a single model, but by the ability to continually improve upon the current state of models. Achieving this requires treating model development as a system-level optimization problem, for which the solution is building a hill-climbing machine for rapid improvement. Our process includes a scaling-focused framework for pre- training modeling decisions, as well as a robust reinforcement learning recipe and infrastructure that sustains long, log-linear performance improvement. The first model developed using our process is MAI-Thinking-1, a 35B active / 1T total parameter MoE that stands among the strongest models of similar size on STEM reasoning and coding tasks (e.g., 52.8% on SWE-Bench Pro, 97.0% on AIME 2025, and 87.7% on LiveCodeBench v6). MAI-Thinking-1 is trained from-scratch, exclusively on clean, enterprise-grade data, without distillation from third-party models. In this technical report, we offer a deep dive into the development of MAI-Thinking-1. By sharing our technical details and learnings we hope to cultivate a transparent and science-driven approach to further development in AI.

Final paragraph of the introduction:

MAI-Thinking-1 is the first model developed using our hill-climbing machine: the integrated process of building data pipelines, training infrastructure, reinforcement learning environments and rewards, evalua- tion suites, and safety tests that turn model development into an empirical optimization loop on a specified domain. The hill-climbing machine allows us to advance AI while grounding progress around human needs from the ground up.

Was kitty wishing upon a star?

When dreams become true... pic.twitter.com/0VIPyCpQ17

— The Figen (@TheFigen_) June 4, 2026

Seeing China Clearly | Robert Wright & Tianyu Fang

YouTube page:

Trans-Pacific tech watcher Tianyu Fang joins Robert Wright to take a close look at China—including how the government works (and doesn’t), online censorship, how China sees AI and the chip war, the legacy of the one-child policy, and more. Plus: Silicon Valley’s hawkish turn. [...]

0:00 Teaser
0:45 How Tianyu became a “China explainer”
3:15 American vs Chinese exceptionalism
8:09 What you can and can’t say in China
18:32 Is the government more accountable in the US or China?
27:03 How China sees America
36:32 The one-child policy and its consequences
42:53 Tianyu: The biggest misconception about China is…
46:20 Heading to Overtime

Fountain in Shipyard Park

Does waning interest in the World Cup signal a thinning of genuine nationalist sentiment?

David Wallace-Wells, Why Does No One Care About the World Cup This Year? NYTimes, June 3, 2026.

They used to call the World Cup, unequivocally, the planet’s biggest sporting event. But it is about to start, right here in North America, and no one much seems to care. Thousands of tickets remain unsold, and just weeks ago, others were being resold well below their official price. [...] And I actually do think this might be telling us something, beyond the world of sports, about the global landscape of politics and culture.

In the States, the indifference might not be surprising, even though the event is being played mostly on U.S. soil. The U.S. team is more talented than in the past but hasn’t looked impressive for years. Soccer is still a growth sport rather than a dominant one in this country, and many Americans aren’t exactly feeling the flush of simplistic patriotism these days. On top of which, the tickets have been priced punishingly high.

What is more striking to me is the muted interest of the rest of the world, which every four years for decades seemed almost to pause for a month to engage in a truly global but appealingly low-stakes performance of tribal nationalism. [...]

What makes this shift so striking is that it has happened alongside a rising tide of political nationalism around the world, which you might think would produce a great surge in soccer nationalism, too. Instead, the age of global populism has coincided with intense interest in the biggest club teams — for-hire rosters assembled largely from international talent by megacorporations boasting jersey sponsorships from foreign conglomerates. [...] But no one could even pretend to illustrate the age of global populism by talking about the intensity of popular feeling about national teams.

That’s from the beginning of the article. After this, that, and the other, Wallace-Wells concludes:

Namely, that what we identify as nationalism in global affairs might be better described as a form of parochialism, with populists making particular claims not about the nation per se so much as the ways it should be reformed — presumably toward some reactionary ideal, its contours often more local than genuinely national. In this reading, globalization hasn’t just generated a backlash among those who resent deindustrialization, capital flight and the stateless lives of the world’s billionaires. It has also made the nation itself seem like a somewhat untrustworthy unit of political and social organization to many people on the right. For them, what might once have served as a source of patriotism and pride now produces feelings of resentment and regret. Not that liberals aren’t queasy about nationalism these days, either. For all of us, rooting for Arsenal or P.S.G. might now be more appealing precisely because it’s essentially meaningless.

Ed Zitron says AI is a losing bet – "doing exit liquidity for venture capital"

To go out publicly and say what he is saying, @edzitron has balls the size of the Las Vegas sphere pic.twitter.com/oHjv3LCvXh

— JustDario (@DarioCpx) June 4, 2026

From Saturday's walk in NW Hoboken

Correcting Cowen’s misleading presentation of large language models [MR #10]

Surprise! There’s been a change of plans. The last time I’d posted about Cowen’s monograph on marginalism I figured I had one more (longish) blog post, one about the fourth and final chapter, “Why Marginalism Will Dwindle, and What Will Replace It?” But the more I thought about it, the longer and more convoluted it got. So I’ve decided to simplify things by writing three posts, each substantial, but focused, instead of a long rambling affair like the one I did on biology. So, I ‘m writing one post about large language models (this post), which Cowen brings up at the end of the chapter. Then I’m writing one about high dimensional models in economics, which Cowen introduces early in the chapter. My final post will be a general response to Cowen’s ideas about where this is all headed.

In this post I want to do three things: 1) First I’ll talk about the surprise nature of the success achieved by GPT-3 and then ChatGPT. 2) Then I will present three passages from Cowen’s text and comment on them. 3) Finally, I want to give a brief rundown of tradition of statistical work that stands behind LLMs.

Surprise!

OpenAI released GPT-3 in 2020 to a limited audience of insiders, who recognized that it represented a breakthrough. This level of performance came as a surprise. No one predicted it. GPT-3 was scaled up from GPT-2, which was in turn scaled up from GPT-1, but no one was making explicit predictions about the level of performance to be achieved at each step. These were experiments: “Let’s try it and see what happens.” That’s fine. That’s a good way to make progress, to try things out and see what happens. But don’t mistake a lucky trial for genuine knowledge.

Cowen mentioned GPT-3 on Marginal Revolution on July 19, and then published a Bloomberg column on it on July 21, which he excerpted in Marginal Revolution the next day: “...think of GPT-3 as giving computers a facility with words that they have had with numbers for a long time, and with images since about 2012.” I published a working paper in August, GPT-3: Waterloo or Rubicon? Here be Dragons, in which I both acknowledged about the breakthrough and cautioned about becoming too satisfied with the technology that occasioned the breakthrough.

Two and a half years later, in November of 2022, OpenAI released ChatGPT to the general public. It spread like wildfire. Now the proverbial everyone witnessed what only a small group had witnessed in the summer of 2020. The machine speaks. Sorta’. But more convincingly than any machine had spoken before and in a way that had unimaginable implications for the future.

A threshold HAS been crossed, but it is not, so far as I can see, a threshold in our understanding, either of AI or anything else. It is a threshold in performance along a continuous line of scientific understanding and engineering design and construction, something I have documented in some detail in a recent working paper, The Origins of LLMs. As far as I can tell, there has been no paradigm shift, in Thomas Kuhn’s sense, no rank shift, in terms of cognitive rank theory. There were no fundamentally new ideas in the world by, say, late July of 2020 as a consequence consolidating GPT-3 and making it available in limited release.

“What about the scaling hypothesis,” you might ask. “Isn’t that new?” Ilya Sutskever first explored the idea in 2014. Rich Sutton’s famous 2019 essay, The Bitter Lesson, generated broad discussion about the issue. Then OpenAI published a paper in 2020 that cemented matters, “Scaling Laws for Neural Language Models.”

Given the nature of computing, scaling up is not trivial. Hundreds if not thousands of technical details need to be worked out as the size of the training corpus increases by factors of 10 or more, time after time, and as more and more GPUs are ganged together to assemble the computing power needed. The scaling hypothesis gave researchers a reason to expect improved performance with scaling, but without having to make fundamental breakthroughs in understanding, not of machine learning, artificial neural nets, and certainly not about language and cognition. Consequently our sense of possibility has expanded enormously. Our knowledge and deep understanding have remained the same and the scaling hypothesis made it easy to believe that that was just fine.

Passages from Cowen’s Text

Unfortunately Cowen seems to have bought this story. Not only that, but he doesn’t even acknowledge that there is considerable current debate about whether or not LLMs will be sufficient to achieve AGI (artificial general intelligence) when they are scaled up enough. The most visible opponent of this idea is Gary Marcus, a student of Steven Pinker, who argues that we need to incorporate insights and technology from “old school” symbolic computing (sometimes known as GOFAI, good old-fashioned AI). Marcus is certainly not alone, there are many others. But I don’t want to reprise that debate. I just want to mention that it exists and that Cowen completely ignores it.

What I would like to do in this section is quote some passages from his text and comment on them.

The Marginal Revolution: Rise and Decline, and the Pending AI Revolution, pp. 106-107:

Suffice to say, LLM construction has for the most part ignored linguists and philosophers, and that also means ignoring their intuitions. LLM construction also ignored a lot of people in the AI field who insisted neural nets were a dead end. Instead, in a relatively short number of years humans invented new ways of modeling language and reasoning through language. That research program has proven wildly successful, as we have much better models of language and reasoning than almost anyone had been expecting.

That first sentence is true, sorta’. It is also misleading. As I have documented in that working paper, The Origins of LLMs, this technology is based on a continuous line of statistical thinking that extends back to the 1950s (I take a brief look at this in the next section) . It is the syntacticians, semanticists, and cognitive scientists who been ignored. The second sentence is a bit of an exaggeration. AlexNet put neural nets firmly back on the agenda in 2012.

The big problem is Cowen’s use of “model” in the last two sentences. Large language models are not causal models like those economists use. They don’t tell us anything about how language and thought work. They are algorithmic models. They are about turning input into output; just how that is done is a mystery. Until we understand the internal operations of LLMs they tell us almost nothing about language and reasoning. They give a boost to the idea that some kind of statistical process is involved, but that’s it.

This situation is deeply paradoxical. These algorithmic models perform much better than the computer models created during the “classical” era of cognitive science, the 1960s and 1970s, models that were based on linguistic theory. We knew how those models worked. We don’t know how these models work. We have purchased performance at the cost of ignorance – a formulation I have from the late Martin Kay.

AI won't unfold in society as fast as the Silicon Valley pundits think it will [Tyler Cowen]

From YouTube: 

Economist and author Tyler Cowen delivers a provocative keynote on how AI will reshape growth, work, status, and geopolitics. Mixing clear‑eyed realism with long‑run optimism, he argues that AI is both our “plan A” for avoiding fiscal crisis and a technology that will leave many people disoriented—and some high‑status winners of the old world worse off.

What’s in this video:
—Why AI will radically change jobs and status without causing mass unemployment
—Two big new job categories: running experiments and gathering data for AI
—The “human bottlenecks” that limit AI’s impact to ~2% → ~2.5% growth
—How AI could be “plan A” for stabilizing public debt and avoiding fiscal crisis
—Who gains and who loses: global poor and initiative‑takers vs. elite professionals

Cowen’s message: AI’s benefits are enormous—higher growth, longer lives, more opportunity for the poor—but they come with psychological, political, and institutional friction. If you work in or care about AI, you’re not just building products; you’re helping write the only credible plan for a sustainable and prosperous future.

Recorded live at Sana AI Summit 2026, New York, May 21st, 2026.