HuShifang

I'd be cautious with Sapolsky claims. https://statmodeling.stat.columbia.edu/2025/06/30/its-sapols...

I think this is a great explanation.

The "Ladder of Causation" proposed by Judea Pearl covers similar ground - "Rung 1” reasoning is the purely predictive work of ML models, "Rung 2" is the interactive optimization of reinforcement learning, and "Rung 3" is the counterfactual and casual reasoning / DGP construction and work of science. LLMs can parrot Rung 3 understanding from ingested texts but it can't generate it.

That you seem to consider yourself sufficiently expert and authoritative to declare the idea "overplayed" -- and that you consider this declaration something that will further readers' understanding (which I at least view as the goal here) -- is... let's say, interesting.

In any case, no, I certainly would not attribute to pigs human-level cognition. But certainly, the notion that language is a prerequisite for thought (at least up through the level of "solving a Sudoku puzzle", per some research) has been largely discarded [0]. And certainly, as omnivores (who will kill and/or consume members of their own species, occasionally including even their own offspring, as well as members of other species) I think it's reasonable to suppose that pigs would have an inkling that one being's death is another's meal. And fundamentally, there is ample reason to suppose that pigs are closer to human-like cognition than the vast majority of species on earth. (Anecdata: they make for difficult pets owing to their need for stimulation, they can play video games [1], and -- as any reputable animal scientist will tell you -- once you get past certain bare-minimum things like ending the use of gestation crates, the most important things for pig welfare in industry are not group housing or (beyond a certain point) extra space, but rather giving them toys and a sense of cleanliness via clean bedding.)

But feel free to provide evidence that attests that "[t]heir thoughts on death and loss will be similarly weak." I would certainly take it into consideration.

[0]: For a handy summary, see https://www.livescience.com/can-we-think-without-language

[1]: https://www.bbc.com/news/technology-56023720

Of course, the hypothesis that they are sufficiently aware of everything -- existing and being nurtured solely to be slaughtered, having already seen many friends die, if female having had many children stolen from them -- to be traumatized to the point of utter emotional breakdown and unfeeling is also consistent with said observation. Humans who have survived war crimes and horrific battles tend to go pretty dead-eyed too. Death might even be welcomed, as a deliverance.

I'd just add, these three can be obtained from veg-friendly supplements too.

I'm not an expert in this, but...

> BNNs bring the following advantages over GPs: First, training large GPs is computationally expensive, and traditional training algorithms scale as the cube of the number of data points in the time series. In contrast, for a fixed width, training a BNN will often be approximately linear in the number of data points. Second, BNNs lend themselves better to GPU and TPU hardware acceleration than GP training operations.

If I'm not mistaken Hilbert Space Gaussian Processes (HSGPs) are O(mn+m) (where m is the number of basis functions, often something like m=30, m=60, or m=100), which is also a huge improvement over conventional GPs' O(n^3). I know that there are some constraints on HSGPs (e.g. they work best with stationary time series, and they're not quite as accurate, flexible, or readily interpretable or tunable as conventional GPs), but what would be the argument for an AutoBNN over an HSGP? Is it mainly about the lack of a need for domain expert input?

As an outsider to the "tech" scene, who did a PhD generals field on modern European intellectual history (which included a dollop of recent continental philosophy), it's endlessly fascinating to me how people in tech have fixated upon Girard. While a few of his ilk do come up in standard reading lists, he (generally) doesn't -- he is far more prominent vis-a-vis his peers in this discourse than in his "native" one. I suspect that this owes to path dependency and his metaphysics' compatibility with the industry's participants' socio-intellectual priors (so to speak).

I think one further bit of horror is, it has the potential to erode your memory of the lost person, piece by piece regressing your recollection of them to the mean (so to speak) of the model's training set. At some point, will the bereaved -- to use the example given here -- just passively accept that the lost loved one actually was the sort of person to write about quilting, even when they never did in life?

John Eliot Gardner's "Bach: Music in the Castle of Heaven" is a sort of musical biography featuring a (very) deep dive on the cantatas.

As it happens, there's a PyMC implementation of the 1st and 2nd editions of Statistical Rethinking here:

https://github.com/pymc-devs/pymc-resources

(I think the author of the book discussed above, Osvaldo Martin, is the primary or sole contributor for the Rethinking implementations, in fact -- he had a full implementation in his own repo (https://github.com/aloctavodia/Statistical-Rethinking-with-P...) before deprecating it in favor of the above-linked one.)

FWIW here's Aki Vehtari's approving share on his (and Gelman et al's) blog. https://statmodeling.stat.columbia.edu/2024/02/08/bayesian-a...

The author was also a lead author on a CRC red-series book, Bayesian Modeling and Computation in Python Learning, published a few years ago (short review here):

https://academic.oup.com/jrsssa/article/185/Supplement_2/S76...

Not an endorsement, but here's an argument that AGI is computationally intractable (edit: at least via ML). https://irisvanrooijcogsci.com/2023/09/17/debunking-agi-inev...

There's also this in the Berkeley Hills. Smaller though.

https://redwood-valley-railway.business.site/

Yes, but as I said, I was talking about early Confucians.

For context, he is reasonably well known -- probably the third most-prominent of the early Confucians (albeit a distant third, after Confucius and Mencius), with partial and complete English translations. But, he was often regarded by Confucians throughout Chinese history warily, because of certain "Legalist" tendencies (which, to put it crudely, were more authoritarian and Machiavellian as well as more legalistic), which were frequently disparaged (but quietly drawn on by pragmatic "statecraft" thinkers). Especially because two of his students, Han Fei and Li Si, were arguably the most (in)famous legalists.

That said, I had a teacher years ago from Taiwan who argued that 1) Xunzi was the first full-blown "philosopher" (in the Western sense of offering an epistemology etc and arguing in a logically reasoned way) in Chinese history, and that 2) Xunzi and Zhuangzi were the smartest Chinese thinkers of all time. Years later, I'm not sure she was wrong.

Here's a recent webinar from PyMC Labs' Max Kochurov (with some slides and code too):

https://www.youtube.com/watch?v=QllfKQH-yQ4

That's what books were for (among other things): they were knowledge prosthetics.

EDIT: And to be clear, yes, books were conveying information from other adults. But 1) the capital requirements of engaging or operating a printing press meant that there was some gatekeeping (for better as well as worse) concerning which adults' views made it into print, and 2) there were things like peer review and other social technologies to increase confidence in the accuracy of the information contained in books, i.e. to make that gatekeeping more than merely economic. LLMs are ingesting the unfiltered thoughts of anyone with internet access, and then noisily producing outputs based on them (with some limited inexpert human fine-tuning at the end).

Personally I would vote "no" regardless of whether I had confidence the kid wouldn't be exposed to the dregs of the Internet, for two reasons:

1. LLMs are I think pretty inarguably bad at answering certain types of questions in a factually accurate way, at least without nontrivial prompt engineering. I don't think many 8 year-olds will recognize which types of questions the LLM will give inaccurate answers to, or be ready for the sort of prompt engineering that'd be required to mitigate this.

2. Learning is of course more than simply obtaining and memorizing facts. You don't want to overfit your kid - you want them to have to struggle a bit for their understanding so that it generalizes. Sure, formulating questions is part of that process, but there's something to be said for having to sift through pages of search results (or even better, from a learning standpoint, searching through the pages of a book they had to find on a library or bookstore shelf). It's like the old idiom, give a man a fish and he eats for a day, teach a man to fish and he eats for a lifetime. Better to teach your kid to fish (and for knowledge/understanding, not mere facts).

As for image generation, I suspect you'd find more inappropriate content come up there than with a chatbot - and teaching the kid to draw, paint, etc would probably be better for their development.

Why on earth would this be remotely noteworthy to Tyler Cowen, an economist? Set aside the cultural stuff -- it's obviously just market dynamics at its core. Humanities degrees are -- rightly or wrongly -- not considered to benefit one's future livelihood, so a priori, they are not seen as being as valuable as others in a financial sense, which drives down enrollments in humanities classes. When enrollments go down, it becomes more difficult to make the case to resource-constrained universities that the department deserves the funding used to recruit and retain faculty and graduate students, etc. (It's already a hard case given that humanities departments are less likely to bring in lucrative outside grants than STEM departments.) So, the faculty sing a song, they do a dance -- that is, they make the class as fun and entertaining as possible -- and they give everyone A's afterward. Word gets out (sometimes in a very structured way, via student social network groups), and students enroll (sometimes whole sports teams will swarm a single humanities class). The classes don't get cancelled, and more sections are created, which creates jobs for the grad students as TAs. Some professors are extra eager to inflate grades, because they like the attention and/or they want a following, which boosts their prestige (merit being more subjective in the humanities than STEM, too), and who knows, maybe one day they'll get a book deal or an NPR interview because they taught a famously big class. Some professors don't really want to do it, but the department chair really needs to keep that enrollment up so the provost doesn't axe the whole department in next year's budget, and the grad students have to eat, and what they study is important, so they bear up and do it anyway, and just feel guilty afterward.

It's a real phenomenon, and not good, but c'mon -- let's not pretend to be so shocked.

(EDIT: And the students will be extra disappointed if they don't get that A that was all but advertised to them, so it's all the harder to ever right the ship. And at prestigious institutions, there's always that facile argument that "If you're smart enough to get into this school, of course you're going to be smart enough to get A's at this school!" And all universities want to keep their future alumni donors happy...)

Second the Beast Academy recommendation. I'd add, it's the younger version of Art of Problem Solving. The interactive games are extremely well designed from a pedagogical standpoint, the videos are fun and clear, the visuals are appealing, etc. (And I will say, there's a TTS button on the screen so that a kid who can't read can still hear the instructions, which are usually pretty intuitive anyway, especially after the video walkthrough. And the accompanying comics have proven more supplemental than essential for us.)

I understand why the IP holders want to keep the revenue flowing by adapting the works to meet current market demands/expectations (while offering lofty talk of the "timelessness" of Dahl's art). Setting aside the question of what this sort of editing does to his art (though for the record, I think that the transgressive meanness of Dahl is an integral part of his artistry, whether one likes it or not), the fact remains that most art is timeful, not timeless -- even works that "work" very well in their own time are usually consigned to obscurity once the world moves on. I tried to read "King Solomon's Mines" a couple months ago, and it was truly unreadable to me -- it relied upon cultural assumptions that are just utterly alien to me, and I couldn't enjoy it. And that's okay. I don't want to read a version that excises all the cringe-inducing language, because forgetting one story makes space for new ones (I can enjoy Indiana Jones movies if I want adventure, for instance, and they might not have been made if H. Rider Haggard's work had been picked up for an updating and franchising.) And endlessly bowdlerizing the "classics" means leaving little room for new ones to be written, read, appreciated, and canonized.

No, you'd need to be chugging multiple jugs of soy milk a day for that to (maybe) matter. Better would be to lift, eat better, sleep, etc.

There's a big literature on this in Chinese history -- many arguments concentrate on economics and labor specifically (to crudely sum it up, there was always lots of cheap labor available, and little incentive to invest the capital to develop and build expensive machinery so as to replace it). There are cultural and intellectual arguments too.

But this piece is totally wrongheaded in that it supposes that South China is China. For most of Chinese history, North China has been politically, socially, and culturally dominant (and not infrequently economically dominant too). It's too cold and arid to grow rice, and thus grows wheat and other grains. (This is why the cuisine is dominated by bready dumplings, e.g. mantou) Some of the author's arguments about rivers could apply to the Huang He (the Yellow River), too, but it has historically been much more central to Chinese civilization than the Yangzi ("Yangtze", or Chang Jiang - lit. "long river").

I'd wait a few years on Chicago -- my impression is that they've been undergoing some pretty profound cultural shifts regarding admissions and grading, i.e. their rigor is getting more performative and less real (and it was always at least a little performative, or at least, rooted in being just a little gratuitously mean).

EDIT: Not coincidentally, it was an admin brought in from Yale who got this ball rolling.

I went to Harvard (for grad school) and taught there as a TF; I also have lots of experience of other private and public schools. I don't think Harvard's lower tier students are as smart as the top tier students at an average school. Let's set aside the egregious legacy or athletic admits. Is the second-lowest quintile of Harvard students smarter than the top quintile of "average school" students? More polished, yeah, sure; more cunning (not necessarily in a negative sense, but in the sense of understanding how to work the system, how to give evaluators what they want), sure. And maybe by the end, there's more difference than there was coming in, thanks to competitive motivation and other network effects. But admissions are something like a lottery, and I would say that the mean "smartness" of Harvard students is not much different from that of a generic selective private school (let's say, Harvard's second-lowest quintile would be the middle quintile at another reasonably selective private school), and that the top percentiles at a generic public school would perform quite comparably to Harvard's top quintile (even if Maryland has a lower mean, as Gladwell notes).

Are there very smart people at Harvard, and other prestigious liberal arts universities? Absolutely. But there are lots of people who are merely above average and lucky, by dint of birth or chance, and some people who aren't even above average. (FWIW MIT and Caltech have much better claims to being filled with the smartest people in the world -- rigorous standards of admission, no legacies, NCAA Div3 athletics, etc.)

You can always find examples of irrational behavior, but the difference is scale. Today a major state isn't spending inordinate resources to ensure that the cookies left for Santa are prepared using the finest ingredients (say, a few million dollars' worth of gold) by the most accomplished bakers, and then placed on a dish so large that it can be seen from space (and hundreds or thousands of people died making it).

If you can show me the tomb of a prominent world leader from the last, let's say, 500 years that's decorated with images of his/her enemies' severed genitalia, I'll concede the point.

Yeah, that's why I hesitated to include it on the list, but I don't think such representations are very common in religious contexts today (and these statues do seem to have been religiously significant, e.g. all the Venus statues from prehistoric Europe and the Middle East)

Well, I am a working professional historian of the premodern world, and this is certainly my impression from years of reading Chinese-language primary sources about China during the Shang, Zhou, Qin, and Han dynasties. Spend some time with the literature, and you'll read how Mohists coupled careful logic and quasi-scientistic reasoning with "ghosts who will punish you if you're bad," about how Han dynasty tomb exterior-door inscriptions talk extensively about how the decedent's family loves them but earnestly hopes to never, ever see them again, because if they did it would mean the decedent left their tomb to punish them for their unfilial conduct. (And you'll read how gingerly the subject of human sacrifice in the distant past, or emperors indulging itinerant "Daoist" rainmakers, to the considerable chagrin of more secular-minded officials, is handled.) I can't point to anything synoptic on the Chinese case -- early China scholars tend to make a lot (frankly too much) out of a little (we don't have that many texts -- but Herbert Fingarette's "Confucius: The Secular as Sacred" was formative for all the early China people I studied with on a more intellectual register. One controversy that gets lots of play: whether Confucius urged performing the rituals "as if" the dead were present, or if they were actually present. Linguistically, the phrasing is entirely open to the latter, even if people like to make the thinking presented in the "Analects" out to be more modern-seeming. At the risk of giving too much of a peek behind the curtain, so to speak, broadly speaking -- this being several decades since Foucault made his mark -- careers get made by either emphasizing the past's surprising lack of alterity or by finding some spectacular, flamboyantly surprising new form of alterity. And the former is rather easier to pull off than the latter.

But, I read about the piles of genitalia in Toby Wilkinson's "The Rise and Fall of Ancient Egypt" and about the god-statues in Trevor Bryce's "Babylonia: A Very Short Introduction."

On the flip side, let's also not forget that the earliest civilizations (so to speak) were also rather different from later ones in some striking ways. Put simply, they were a lot more literal-minded, and didn't engage in as much abstraction as did their descendents. Ancient fertility statues (even as late as classical Greece and Rome) are grotesquely over-endowed; ritual sacrifices of food and symbolic objects have in many places taken the place of sacrificed slaves or wives. Mesopotamian city states would fight wars because, like a frat prank, one would steal a statue of a god like Marduk from the other's temple -- only, there was no notion that it was a statue, rather it was the god. Early Egyptian murals speak to the power of kings by showing piles of dicks their soldiers had cut off of defeated enemies (unlike Egyptians they were uncircumcised). And there's the whole mummification thing, which betrays a certain literal-mindedness about immortality. So yeah, it's a spectrum - but both before and after the "rise of civilization"

Skiena's "Algorithm Design Manual" (which recently got a third edition) introduces algos and their application -- when to use them, why you might use one over another, etc. With lengthy "stories from the trenches" about problems he's solved using them, in some cases. And it's a handy reference on the ideas behind the algos, too