I happened to learn recently that that’s probably not from Keynes:

https://quoteinvestigator.com/2011/08/09/remain-solvent/

Aaronson goes on:

Look, obviously the physicists had their reasons for teaching quantum mechanics that way, and it works great for a certain kind of student. But the “historical” approach also has disadvantages, which in the quantum information age are becoming increasingly apparent. For example, I’ve had experts in quantum field theory – people who’ve spent years calculating path integrals of mind-boggling complexity – ask me to explain the Bell inequality to them, or other simple conceptual things like Grover’s algorithm. I felt as if Andrew Wiles had asked me to explain the Pythagorean Theorem.

And then, did anyone clap?

This is a false analogy. I don’t think it’s a surprise, I am not convinced that it’s an actual problem, and if it is, I don’t think Aaronson makes any progress to a solution.

The Pythagorean theorem is part of the common heritage of all mathematics education. Moreover, it’s the direct ancestor to the problem that Wiles famously solved. It’s going to be within his wheelhouse. But a quantum field theorist who’s been deep into that corner of physics might well not have had to think about Bell inequalities since they were in school. It’s like asking an expert on the voyages of Zheng He about how Charlemagne became Holy Roman Emperor. There are multiple aspects of Bell inequalities that someone from a different specialization could want “explained”, even if they remember the gist. First, there are plenty of questions about how to get a clean Bell test in the laboratory. How does one handle noise, how do we avoid subtly flawed statistics, what are these “loopholes” that experimentalists keep trying to close by doing better and better tests, etc. Aaronson has nothing to say about this, because he’s not an experiment guy. And again, that’s entirely fair; some of us are best as theorists. Second, there are more conceptual (dare I say “philosophical”?) questions about what exactly are the assumptions that go into deriving Bell-type inequalities, how to divide those assumptions up, and what the violation of those inequalities in nature says about the physical world. Relatedly, there are questions about who proved what and when, what specifically Bell said in each of his papers, who built on his work and why, etc. Aaronson says very little about all of this. Nothing leaps out at me as wrong, but it’s rather “101”. The third broad category of questions are about mathematical specifics. What particular combination of variables appears in which inequality, what are the bounds that combination is supposed to satisfy, etc. The expressions that appear in these formulae tend to look like rabbits pulled out of a hat. Sometimes there are minus signs and factors of root-2 and such floating around, and it’s hard to remember where exactly they go. Even people who know the import of Bell’s theorem could well ask to have it “explained”, i.e., to have some account given of where all those arbitrary-looking bits came from. I don’t think Aaronson does particularly well on this front. He pulls a rabbit out of his hat (a two-player game with Alice and Bob trying to take the XOR of two bits), he quotes a number with a root-2 in it, and he refers to some other lecture notes for the details, which include lots of fractional multiples of pi and which themselves leave some of the details to the interested reader.

Aaronson leads into this rather unsatisfying discussion thusly:

So what is Bell’s Inequality? Well, if you look for an answer in almost any popular book or website, you’ll find page after page about entangled photon sources, Stern–Gerlach apparatuses, etc., all of it helpfully illustrated with detailed experimental diagrams. This is necessary, of course, since if you took all the complications away, people might actually grasp the conceptual point!

However, since I’m not a member of the Physics Popularizers’ Guild, I’m now going to break that profession’s time-honored bylaws, and just tell you the conceptual point directly.

The tone strikes me, personally, as smarmy. But there’s also an organizational issue. After saying he’ll “just tell you the conceptual point directly”, he then goes through the XOR rigmarole, which takes more than a page, before he gets to “the conceptual point” (that quantum mechanics is inconsistent with local hidden variables). It’s less direct than advertised, for sure. I have not systematically surveyed pop-science explanations of Bell’s theorem prior to 2013, but the “page after page of entangled photon sources…” rings false to me.

Regarding occasional sneer target Lawrence Krauss and his co-conspirators:

Months of waiting but my review copy of The War on Science has arrived.

I read Krauss’ introduction. What the fuck happened to this man? He comes off as incapable of basic research, argument, basic scholarship. […] Um… I think I found the bibliography: it’s a pdf on Krauss’ website? And all the essays use different citation formats?

Most of the essays don’t include any citations in the text but some have accompanying bibliographies?

I think I’m going insane here.

What the fuck?

https://bsky.app/profile/nateo.bsky.social/post/3lyuzaaj76s2o

Aaronson tries to back up his perspective in chapter 9, where he makes the following contention:

Quantum mechanics is what you would inevitably come up with if you started from probability theory, and then said, let’s try to generalize it so that the numbers we used to call “probabilities” can be negative numbers.

This is a bait-and-switch, or more charitably, poor organization. Later he will admit that he needs to introduce not just negative numbers, but complex numbers too. What arguments does he give to justify bringing complex numbers into the picture? Why prefer ordinary quantum theory over what we might call “real-amplitude” quantum theory? He provides three suggestions. The first is based on a continuity argument (“if it makes sense to apply an operation for one second, then it ought to make sense to apply that same operation for only half a second”). He argues that this can only be made to work if the amplitudes are complex rather than only real. But this does not hold. We can simply say that in real-amplitude quantum theory, the time evolution operators belong to the subgroup of the orthogonal group that is continuously connected to the identity. This is actually what would be analogous to regular quantum theory, where we make unitary operators by taking the exponential of -iHt, where H is a Hamiltonian and t is an amount of time. In the real-amplitude theory, we just use an antisymmetric matrix as a generator instead of an anti-Hermitian one.

The second argument is that the number of parameters needed to specify a mixed state scales better for complex amplitudes than for real. This is a style of argument that has a considerable cachet among aspiring reconstructors of the quantum formalism, but it too has shortcomings. Aaronson invokes the principle that states for independent quantum systems combine via the tensor product. He asserts that this is true, and then argues that this makes the parameter counting work out nicely for complex but not real amplitudes. Plainly, then, this case for complex amplitudes can’t be better than the case for the tensor product. It replaces one mathematical “brute fact” with another. People who go into more depth about this invoke a premise they call “tomographic locality”. The conceptual challenge is then, if tomographic locality failed to hold true, would that actually be so bad? Would we find it stranger than, for example, quantum entanglement? See Hardy and Wootters (2010) and Centeno et al. (2024).

The third argument is given almost in passing. It’s a “well, I guess that’s nice” property which holds for the complex-amplitude theory and fails for the real-amplitude version. Bill Wootters noticed it. Of course, he also found something that works out nice only when the amplitudes are real instead. See Wootters (2013) for a more recent explanation of the latter, which he first published in 1980.

What Aaronson calls starting “directly from the conceptual core” strikes me instead as merely discarding some old prefatory material, like the Bohr model of hydrogen, and replacing it with new, like some chatter about classical computation. His “conceptual core” is the same old postulate. He just applies it in somewhat different settings, so he ends up doing matrix algebra instead of differential equations. I once thought that would be easier on students, but then I actually had to teach a QM class, and then I ended up “reviewing” a lot of matrix algebra.

A physicist who learned quantum mechanics the old-fashioned way, and who now sees “quantum” being hyped as the next Bitcoin, might well have some questions at this point. “So, you’re telling me that these highly idealized models of hypothetical, engineered systems bring us closer to the secrets of the Old One than studying natural phenomena will? I’m sure you have your own good reasons for wanting to know if QURP is contained in PFUNK, but I want to understand why ice floats on water, why both iron and charcoal glow the same kind of red when they get hot, why a magnet will pick up a steel paperclip but not a shiny soda can.” And: “I get the desire for a ‘conceptual core’ to quantum physics. But have you actually isolated such a thing? From where I stand, it looks like you’ve just picked one of the important equations and called it the important equation. Shouldn’t your ‘conceptual core’ be a statement with some punch to it, like the big drama premise of special relativity? What’s your counterpart to each observer who feels herself motionless will measure the same speed of light?”

Here’s how Aaronson begins chapter 9:

There are two ways to teach quantum mechanics. The first way – which for most physicists today is still the only way – follows the historical order in which the ideas were discovered. So, you start with classical mechanics and electrodynamics, solving lots of grueling differential equations at every step. Then, you learn about the “blackbody paradox” and various strange experimental results, and the great crisis these things posed for physics. Next, you learn a complicated patchwork of ideas that physicists invented between 1900 and 1926 to try to make the crisis go away. Then, if you’re lucky, after years of study, you finally get around to the central conceptual point: that nature is described not by probabilities (which are always nonnegative), but by numbers called amplitudes that can be positive, negative, or even complex.

This is wrong in a few ways. First, that “years of study”? Yeah, I saw complex probability amplitudes in my first term of college. Before they showed us all the blobby/cloudy pictures of electron orbitals, they took two minutes to explain what was being plotted. Our first full-blown quantum mechanics course was at the advanced age of … sophomore year. And we’re not talking about something squeezed in on the last day before summer vacation. See above regarding how it’s the third equation in the first chapter of the ubiquitous standard undergrad QM textbook. This is not an idea sequestered in the inner sanctum of knowledge; it’s babby’s first wavefunction.

Second, the orthodox method is not really “historical”. It can’t be. The physicists who did all that work from 1900 through 1925–27 knew much more physics than college kids do today. They were professionals! Pick up the Dover reprint of the Sources of Quantum Mechanics collection, and see how many of the papers in it make sense using only first-year physics. Dirac was thinking about Poisson brackets, not a block on an inclined plane. The capsule “histories” in QM textbooks are caricatures, and sometimes quite poor ones at that.

Glob help me, but I’ve actually been reading Quantum Computing Since Democritus, and I’ve been sorely tempted to write an effortful post about it. In particular, it is appealing to ask whether the book delivers on its professed theme. Here’s Aaronson in the preface, laying out what he considers the book’s “central message”:

But if quantum mechanics isn’t physics in the usual sense – if it’s not about matter, or energy, or waves, or particles – then what is it about? From my perspective, it’s about information and probabilities and observables, and how they relate to each other.

This is a defensible claim. All the way back in the 1930s, Birkhoff and von Neumann were saying that we should understand quantum physics by modifying the rules of logic, which is about as close to “quantum information” thinking before the subjects of computer science and information theory had really been invented. Later, E. T. Jaynes was fond of saying that quantum mechanics is an omelette that mixes up nature and our information about nature, and in order to make further progress in physics, we need to separate them. When undergrads came to John Wheeler asking for summer research projects, he liked to suggest, “Derive quantum mechanics from an information-theoretic principle!” But the question at hand is whether Aaronson’s book succeeds at making a case. You can talk a lot about quantum information theory or quantum computing without convincing anyone that it illuminates the fundamental subject matter of quantum mechanics. Knuth’s Art of Computer Programming is not an argument that classical electromagnetism is “about information”.

Here’s Aaronson a bit later:

Here, the physicists assure us, no one knows how we should adjust our intuition so that the behavior of subatomic particles would no longer seem so crazy. Indeed, maybe there is no way; maybe subatomic behavior will always remain an arbitrary brute fact, with nothing to say about it beyond “such-and-such formulas give you the right answer.”

Then he argues,

as the result of decades of work in quantum computation and quantum foundations, we can do a lot better today than simply calling quantum mechanics a mysterious brute fact.

What is this new improved perspective? Here’s how his italicized paragraph about it begins:

Quantum mechanics is a beautiful generalization of the laws of probability: a generalization based on the 2-norm rather than the 1-norm, and on complex numbers rather than nonnegative real numbers.

That isn’t just a “brute fact”. It’s the same “brute fact” that an ordinary textbook will tell you! It’s the “fourth postulate” in Cohen-Tannoudji et al., equation (1.3) in Griffiths and Schroeter, page 9 of Zwiebach. All that Aaronson has done is change the jargon a tiny bit.

Aaronson declares himself indifferent to the needs of “the people designing lasers and transistors”. And fair enough; we all have our tastes for topics. But he has set himself the challenge of demonstrating that studying how to program computers that have not been built, and comparing them to computers that physics says can never be built, is the way to the heart of quantum mechanics.

Aaronson quotes a passage from Carl Sagan, thusly:

Imagine you seriously want to understand what quantum mechanics is about. There is a mathematical underpinning that you must first acquire, mastery of each mathematical subdiscipline leading you to the threshold of the next. In turn you must learn arithmetic, Euclidean geometry, high school algebra, differential and integral calculus, ordinary and partial differential equations, vector calculus, certain special functions of mathematical physics, matrix algebra, and group theory . . . The job of the popularizer of science, trying to get across some idea of quantum mechanics to a general audience that has not gone through these initiation rites, is daunting. Indeed, there are no successful popularizations of quantum mechanics in my opinion – partly for this reason. These mathematical complexities are compounded by the fact that quantum theory is so resolutely counterintuitive. Common sense is almost useless in approaching it. It’s no good, Richard Feynman once said, asking why it is that way. No one knows why it is that way. That’s just the way it is.

Aaronson follows this by saying that he doesn’t need convincing: “Personally, I simply believe the experimentalists” when they say that quantum physics works. Again, fair enough on its own. But I think this is poor media literacy here. Sagan’s Demon-Haunted World is all about the public understanding of science, the difference between authorities and experts, the challenge of becoming scientifically literate, and that kind of thing. What Sagan means by “what quantum mechanics is about” in this context is what physicists use the theory to do, day by day, and why we have confidence in it. Even if you come along with a better explanation of where the mathematics comes from, all that won’t go away!

I don’t know why they decided to put an AI slop image right up top in their banner and then repeat it later.

Falling down the “Rationality Trap” rabbit hole (illustration by ChatGPT 5 Thinking)

Yeah, the washed-out faux-Ghibli aEStheTiC was a bit of a clue. Gonna go scrub my brains with the Tenniel illustrations for Alice now, with a Ralph Steadman chaser.

Afterthought: This kind of brainrot, the petty middle-management style of ends justifying the means, is symbiotic with pundit brainrot, the mentality that Jamelle Bouie characterizes thusly.

It is sometimes considered gauche, in the world of American political commentary, to give words the weight of their meaning. As this thinking goes, there might be real belief, somewhere, in the provocations of our pundits, but much of it is just performance, and it doesn’t seem fair to condemn someone for the skill of putting on a good show.

Both reject the idea that words mean things, dammit, a principle that some of us feel at the spinal level.

The way these people treat the written word confounds me. Whenever I cite a source, it’s because I’ve read it and know what it says. The fact that “AI” facilitates the process of deciding on your conclusion and then filling in bullshit to prop it up makes “AI” corrosive to a person’s moral fiber.

But I just met 'er!

Startup carcass in alley this morning. Tire tread on burst bubble. This Valley is afraid of me. I have seen its true face. The prediction markets are extended gutters and the gutters are full of blood and when the drains finally scab over, all the techbros will drown. The accumulated filth of all their microdosing and Soylent will foam up about their waists and all the accelerationists and effective altruists will look up and shout “Save us!”

And I’ll whisper, maybe later.

Don’t worry, Kelsey Piper managed to use it as an opportunity to be a bluecheck dipshit.

(via)

Education report calling for ethical AI use contains over 15 fake sources

womp, and wait for it, womp

The gateway into the online rationalist world is often effective altruism, which is grounded in the genuinely reasonable idea that those who donate to charity should get the most bang for their buck. […] The focus of the movement shifted to existential concerns around humanity’s survival, such as multi-planet living (so humanity could survive the end of Earth), and artificial intelligence – both to ensure it doesn’t wipe out humanity once it emerges, but also to make sure it does emerge, because of a belief in its massive potential to fix our societal issues.

As so often happens, this downplays the extent to which the batshit was within them all along.

That included a piece of rationalist Harry Potter fan fiction, Harry Potter and the Methods of Rationality, written by the controversial AI theorist Eliezer Yudkowsky, which deconstructs the contradictions of the wizarding world.

“AI theorist” really does just mean “guy who makes shit up”, doesn’t it? And that seems a generous description of HPMoR, which was really more about inventing problems that the HP books didn’t have while ignoring those that they did, since it was really based on fan wikis instead.

(massive bong rip) the aliens already came here and put us all in the Matrix, dude

That’s just yer bog-standard “the best lie has a seed of truth”, ainnit?

(Peer review in its modern form was adopted gradually, with a recognizable example in 1831 from the same William Whewell who coined the word scientist. It displaced the tradition of having the editor of a journal decide everything himself, so whatever its flaws, it has broadened the diversity of voices that influence what gets officially published.)

Behold the power of this fully selective quotation.

The Wall Street Journal came out with a story on “conspiracy physics”, noting Eric Weinstein and Sabine Hossenfelder as examples. Sadly, one of their quoted voices of sanity is Scott Aaronson, baking-soda volcano of genocide apologism.

At least with Ulysses, there’s a brothel chapter.

The Grauniad has a new piece today about the underpaid human labor on which the “AI” industry depends:

https://www.theguardian.com/technology/2025/sep/11/google-gemini-ai-training-humans

Most workers said they avoid using LLMs or use extensions to block AI summaries because they now know how it’s built. Many also discourage their family and friends from using it, for the same reason.