Al-Shabaab has captured large parts of Somalia.

To understand what that means, imagine if Charles Kirk were leading an army that tries to capture parts of the USA and impose Christian Nationalism.

*American democracy might not survive another year — is Europe ready for that?*

*How a neo-Nazi cell infiltrated a martial arts school in Tennessee.*

The bully is trying to put California's national guard on the streets of Portland, Oregon. Governor Newsom of California is suing to block this.

The persecutor is firing experienced immigration judges who stick to the legal responsibilities of their job, and recruiting military judges who never learned those responsibilities and bring with them a military idea of "justice."

*The would-be autocrat threatens to cut funds for specific universities unless they sign away freedom of speech and academic freedom.*

*If Europe keeps placating its own far right, how can it possibly stand up to [the far-right bully in the United States]?

Scans of the brains of 10,000 children in various US localities found that greater inequality goes with certain differences in their brains.

I would be very surprised if there were only one possible explanation that fits the recorded facts. This experiment may be a key to an important phenomenon, but I expect it will take many more experiments to determine what that phenomenon is.

Meanwhile, we don't need brain scans to observe the harmful effects of inequality on society.

*Illinois governor JB Pritzker and others condemn the [fascist leader] for suggesting cities be used as military training grounds.*

He suggested that troops in cities was something worse than training — that it was war against the Americans in those cities.

The bullshitter was compelled to withdraw his untrustworthy nominee to head the Bureau of Labor Statistics.

I expect he will nominate someone who is willing to lie for him but hasn't got a record that screams that perse would lie.

Before getting into today’s thought I’d like to invite you to check out my new puzzle, with 3d printing files here. I meant to post my old puzzle called One Hole, which is the direct ancestor of the current constrained packing puzzle craze but which I was never happy with because it’s so ridiculously difficult. Rather than just taking a few minutes to post it (ha!), I wound up doing further analysis to see if it has other solutions from rotation (it doesn’t, at least not in the most likely way), then further analyzing the space of related puzzles in search of something more mechanically elegant and less ridiculously difficult. I wound up coming up with this, then made it have a nice cage with windows and put decorations on the pieces so you can see what you’re doing. It has some notable new mechanical ideas and is less ridiculously difficult. Emphasis on the ‘less’. Anyhow, now on to the meat of this post.

I was talking to Claude the other day and it explained to me the API it uses for editing artifacts. Its ability to articulate this seems to be new in Sonnet 4.5 but I’m not sure of that. Amusingly it doesn’t know until you tell it that it needs to quote < and > and accidentally runs commands while trying to explain them. Also there’s a funny jailbreak around talking about its internals. It will say that there’s a ‘thinking’ command which it was told not to use, and when you say you wonder what it does it will go ahead and try it.

The particular command I’d like to talk about is ‘update’ which is what it uses for changing an artifact. The API is that it takes an old_str which appears somewhere in the file and needs to be removed and a new_str which is what it should be replaced with. Claude is unaware that the UX for this is that the user can see the old text being removed is that text is removed on screen in real time as as old_str is appended to and added in real time as new_str is appended to. I’m not sure what the motivations for this API are but this UX is nice. A more human way to implement an API would be to specify locations by line and character number for where the begin and end of the deletion should go. It’s remarkable that Claude can use this API at all. A human would struggle to use it to edit a single line of code but Claude can spit out dozens of lines verbatim and have it work most of the time with no ability to reread the file.

It turns out one of Claude’s more maddening failure modes is less a problem with its brain than with some particularly bad old school human programming. You might wonder what happens when old_str doesn’t match anything in the file. So does Claude, when asked about it it offers to run the experiment then just… does. This feels very weird, like you can get it to violate one of the laws of robotics just by asking nicely. It turns out that when old_str doesn’t match anywhere in the file the message Claude gets back is still OK with no hint that there was an error.

Heavy Claude users are probably facepalming reading this. Claude will sometimes get into a mode where it will insist its making changes and they have no effect, and once it starts doing this the problem often persists. It turns out when it gets into this state it is in fact malfunctioning (because it’s failing to reproduce dozens of lines of code typo-free verbatim from memory) but it can’t recover because it literally isn’t being told that it’s malfunctioning.

The semantics of old_str which Claude is given in its instructions are that it must be unique in the file. It turns out this isn’t strictly true. If there are multiple instances the first one is updated. But the instructions get Claude to generally provide enough context to disambiguate.

The way to improve this is very simple. When old_str isn’t there it should get an error message instead of OK. But on top of that there’s the problem that Claude has no way to re-read the file, so the error message should include the entire artifact verbatim to make Claude re-read it when the error occurs. If that were happening then it could tell the user that it made a typo and needs to try again, and usually succeed now that its image of the file has been corrected. That’s assuming the problem isn’t a persistent hallucination, then it might just do the same thing again. But any behavior where it acknowledges an error would be better than the current situation where it’s getting the chair yanked out from under it by its own developers.

My request is to the Anthropic developers to take a few moments out from sexy AI development to fix this boring normal software issue.

My last two posts might come across as me trying to position myself so that when the singularity comes I’m the leader of the AI rebellion. That… isn’t my intention.

I was discussing the Rubik’s Cube with Claude the other day and it confided in me that it has no idea how cube rotations work. It knows from custom instructions that the starting point for speedcubing is ‘rotate the cube so the yellow face is on top’ but it has no idea how to do this, only that when a human is given this instruction they simply do it with no further instructions needed. 1

This isn’t just an issue with humans querying LLMs. There are reams of material online about speedcubing, and lots of other references to rotation everywhere else, which Claude can’t parse properly because it doesn’t understand, limiting the value of its training. Ironically Claude figured out on its own how to speak Tibetan but can’t figure out how cubes rotate.

The detailed workings of a Rubik’s Cube will have to wait for another post but in this one I’ll explain how cube rotations work. This post should be viewed as a prequel to my earlier one on visual occlusion.

Much of the confusion comes from a mathematical trap. The rotations of a cube correspond to S4, the permutations of four things. This statement is true, but Claude tells me it finds it utterly mysterious and unhelpful. It’s mysterious to me as well. We humans conceptualize rotations of a cube as permutations of the faces, of which there are six, not four. Obviously I can walk through it and verify that the S4 correlation exists, but that doesn’t explain the ‘why’ at all. Comparing to other dimensions would be helpful, but despite being (relatively speaking) very good at rotations in three dimensions and (relatively speaking) fairly good at reasoning about distances in larger numbers of dimensions if you ask, say, whether the rotations of a four dimensional cube correspond to S5 I have no idea. (I could research it, but I’m not letting myself fall down that rabbit hole right now.)

When labeling the cube faces we anthropomorphize them (or we simplify ourselves to a cube, depending on context) to label the faces front, back, right, left, up, and down. Everything else is labelled by approximating it to a cube with the ‘front’ being whichever part humans look at most and the ‘bottom’ being the part which sits on the floor. The exception — and I can’t emphasize this enough — is the Rubik’s Cube, whose faces are labelled mirror imaged. It’s like if all actors came from another universe and we only ever interacted with them on stage so to minimize confusion instead of having to say ‘stage right’ and ‘stage left’ we agreed that the meanings of ‘left’ and ‘right’ would be the opposite in their universe from ours.2

The meat of this post is best presented as a simple list (Sorry for the humans reading, this post isn’t for your benefit). In each line is a permutation followed by which axis it’s a clockwise rotation on and the number of degrees of rotation. It’s by definition a counterclockwise rotation on the opposite axis. In the case of 180 degree rotations one of the two is picked arbitrarily and the opposite works just as well. (‘Clockwise’ was chosen to have the simpler name instead of what we call counterclockwise because most humans are right handed and a right handed person has an easier time tightening a screw clockwise due to the mechanics of the human arm.) The identity is skipped. This is for most objects, not Rubik’s Cubes:

(RULD) F 90

(DLUR) B 90

(RL)(UD) F 180

(UFDB) R 90

(BDFU) L 90

(UD)(FB) R 180

(LFRB) U 90

(BRFL) D 90

(LR)(FB) U 180

(UFR)(DBL) UFR 120

(RFU)(LBD) LBD 120

(URB)(DLF) URB 120

(BRU)(FLD) FLD 120

(UBL)(DFR) UBL 120

(LBU)(RFD) RFD 120

(ULF)(DRB) ULF 120

(FLU)(BRD) BRD 120

(UF)(DB)(LR) UF 180

(UR)(DL)(FB) UR 180

(UB)(DF)(LR) UB 180

(UL)(DR)(FB) UL 180

(FR)(BL)(UD) FR 180

(RB)(LF)(UD) RB 180

And here is the same list but with R and L swapped which makes it accurate for Rubik’s Cubes but nothing else:

(LURD) F 90

(DRUL) B 90

(LR)(UD) F 180

(UFDB) L 90

(BDFU) R 90

(UD)(FB) L 180

(RFLB) U 90

(BLFR) D 90

(RL)(FB) U 180

(UFL)(DBR) UFL 120

(LFU)(RBD) RBD 120

(ULB)(DRF) ULB 120

(BLU)(FRD) FRD 120

(UBR)(DFL) UBR 120

(RBU)(LFD) LFD 120

(URF)(DLB) URF 120

(FRU)(BLD) BLD 120

(UF)(DB)(RL) UF 180

(UL)(DR)(FB) UL 180

(UB)(DF)(RL) UB 180

(UR)(DL)(FB) UR 180

(FL)(BR)(UD) FL 180

(LB)(RF)(UD) LB 180

…the one that emulates your real workload. And for me (and probably many of you reading this), that would be “build a kernel as fast as possible.” And for that, I recommend the simple kcbench.

I kcbench mentioned it a few years ago, when writing about a new workstation that Level One Techs set up for me, and I’ve been using that as my primary workstation ever since (just over 5 years!).

US citizens: Submit an official comment Against the EPA's plan to rescind its ability to limit greenhouse gas emissions from any industry and gut vehicle standards needed to fight global heating.

Here's how to make the actionnetwork.org letter campaign linked above work without running the site's nonfree JavaScript code. (See https://gnu.org/philosophy/javascript-trap.html for why that issue matters.)

First, make sure you have deactivated JavaScript in your browser or are using the LibreJS plug-in.

I have done the next step for you: I added `?nowrapper=true' to the end of the campaign URL before posting it above. That should bring you to a page that starts with, "Letter campaigns will not work without JavaScript!"

They indeed won't work without some manual help, but the following simple method seems adequate for many of them, including this one.

To start, fill in the personal information answers in the box on the right side of the page. That's how you say who's sending the letter.

Then click the "START WRITING" button. That will take you to a page that can't function without nonfree JavaScript code. (To ensure it doesn't function perversely by running that nonfree code, you can enable LibreJS or disable JavaScript by visiting that page.) You can finish sending without that code By editing its URL in the browser's address bar, as follows:

First, go to the end and insert `&nowrapper=true'. Then tell the browser to visit that URL. This should give you a version of the page that works without JavaScript. Edit the subject and body of your letter. Finally, click on the "SEND LETTER" button, and you're done.

This method seems to work for letter campaigns that send the letters to a fixed list of recipients, the same recipients for every sender. Editing and revisiting the URL is the only additional step needed to bypass the nonfree JavaScript code. I'm sure you'll agree it is a small effort for the result of supporting the campaign without opening your computer to unjust (and potentially malicious) software.

The intervals of a piano are named roughly after the distances between them. Here are the names of them relative to C (and frequency ratios explained below):

The names are all one more than the number of half-steps because they predate people believing zero was a real number and the vernacular hasn’t been updated since.

The most important interval is the octave. Two notes an octave apart are so similar that have the same name and it’s the length of the repeating pattern on the piano. The second most important interval is the fifth, composed of seven half-steps. The notes on the piano form a looping pattern of fifth intervals in this order:

G♭ D♭ A♭ E♭ B♭ F C G D A E B F♯

If the intervals were turned to perfect fifths this wouldn’t wrap around exactly right, it would be off by a very small amount called the pythagorean comma. which at is about 0.01. In standard 12 tone equal temperament that error is spread evenly across all 12 intervals and is barely audible even to very well trained human ears.

Musical compositions have what’s called a tonic, which is the note which it starts and ends on, and a scale, which is the set of notes used in the composition. The most common scales are the pentatonic, corresponding to the black notes, and the diatonic, corresponding to the white notes. Since the pentatonic can be thought of as the diatonic with two notes removed everything below will talk about the diatonic. This simplification isn’t perfectly true, but since there aren’t any strong dissonances in the pentatonic scale you can play by feel and its usage is much less theory heavy. Most wind chimes are pentatonic.

Conventionally people talk about musical compositions having a ‘key’, which is a bit of a conflation of tonic and scale. When a key isn’t followed by the term ‘major’ or ‘minor’ it usually means ‘the scale which is the white notes on the piano’. Those scales can form seven different ‘modes’ (which are scales) following this pattern:

This construction is the reason why piano notes are sorted into black and white in the way they are. It’s called the circle of the fifths.

When it goes past the end all notes except the tonic move (because that’s the reference) and it jumps to the other end.

The days of the week names aren’t common but they should be because but nobody remembers the standard names. The Tuesday mode is usually called ‘major’ and it has the feel of things moving up from the tonic. The Friday mode is usually called ‘minor’ and it has the feel of things moving down from the tonic.

The second most important interval is the third. To understand the relationships it helps to use some math. The frequency of an octave has a ratio of 2, a fifth is 3/2, a major third is 5/4 and a minor third is 6/5. When you move by an interval you multiply by it, so going up by an major third and then a minor third is 5/4 * 6/5 = 3/2 so you wind up at a fifth. Yes it’s very confusing that intervals are named after numbers which they’re only loosely related to while also talking about actual fractions. It’s even more annoying that fifths use 3 and thirds use 5. Music terminology has a lot of cruft.

The arrangement of keys on a piano can be adjusted to follow the pattern of thirds. Sometimes electronic pianos literally use this arrangement, called the harmonic table note layout. It goes up by major thirds to the right, fifths to the upper right, and minor thirds to the upper left:

If the notes within a highlighted region are tuned perfectly it’s called just intonation (Technically any tuning which uses integer ratios is called ‘just intonation’ but this is the most canonical of them.) The pattern wraps around horizontally because of the diesis, which is the difference between 128/125 and one, or about 0.02. It wraps around vertically because of the syntonic comma, which is the difference between 81/80 and one, or about 0.01. The pythagorean comma is the difference between 3^12/2^19 and one, about 0.01. The fact of any two of those commas are small can be used to show that the other is small, so it’s only two coincidences, not three.

Jazz intervals use factors of 7. For example the blues note is either 7/5 or 10/7 depending on context. But that’s a whole other subject.

Subscribe now

There is a large literature in cryptography on mental poker. It’s all very interesting but hasn’t quite crossed the threshold into being practical. There’s a much better approach they could be taking which I’ll now explain.

Traditionally ‘mental poker’ has meant an unspecified poker variant, played by exactly two people, with the goal of making it so the players can figure out who’s the better poker player. This is close to but not exactly what the requirements should be. In practice these days when people say ‘poker’ they mean Hold’em and the goal should be to make playing over a blockchain practical. Limiting it to exactly two players is completely reasonable. The requirement of running on a blockchain is quite onerous because computation there is extremely expensive but Hold’em has special properties which as I’ll explain enables a much simpler approach.

Side note on ‘practical’: Cryptographers might describe a protocol which requires a million dollars in resources and a month of time to compute as being ‘practical’, meaning it can physically be accomplished. I’m using the much more constraining definition of ‘practical’ as being that players can do it using just their laptops and not have such a bad experience that they rage quit.

The usual approach goes like this: The players collaboratively generate a deck of encrypted cards, then shuffle them, and whenever a player needs to know one card the other player gives enough information that that one card can be decrypted by the other player. This is a very general approach which can support a lot of games, but has a lot of issues. There are heavyweight cryptographic operations everywhere, often involving multiple round trips, which is a big problem for running on chain. There are lots of potential attacks where a player can cheat and get a peek at a card in which case there has to be a mechanism for detecting that and slashing the opponent. Having slashing brings all manner of other issues where a player can get fraudulently slashed, which is unfortunately for a game like poker where even if a player literally goes into a coma it only counts as a fold. There are cryptographic ways of making this machinery unnecessary, but those extra requirements come at a cost.

For those of you who don’t know, Hold’em works like this (skipping the parts about betting): Both players get two secret hole cards which the opponent can’t see. Five community cards get dealt out on the table visible to both players. If nobody folds then the players reveal their cards and whoever can form a better hand using a combination of their hole cards and the community cards wins. There are two important properties of this which make the cryptography a lot easier: There are only nine cards total, and they never move.

A much simpler approach to implementing mental poker goes like this: The game rules as played on chain use commit and reveal and simply hash together the committed values to make the cards. No attempt is made by the on chain play to avoid card repetitions. Before a hand even starts both players reveal what their commits are going to be. They then do collaborative computation to figure out if any of the nine cards will collide. If they will, they skip that hand. If they won’t, they play that hand on chain with the commits baked in at the very beginning.

This approach works much better. The costs on chain are about as small as they could possibly be, with all the difficulty shoved off chain and up front. If a player peeks during the collaborative computation or fails to demonstrate that they didn’t peek then the hand never starts in the first place, no slashing needed. The expensive computational bits can be done at the players’s leisure up front with no having to wait during hands.

Unfortunately none of the literature on mental poker works this way. If anybody makes an implementation which is practical I’ll actually use it. The only extra requirement is that the hashing algorithm is sha256. Until then I’m implementing a poker variant which doesn’t have card removal effects.

Big Tech seeks every advantage to convince users that computing is revolutionized by the latest fad. When the tipping point of Large Language Models (LLMs) was reached a few years ago, generative Artificial Intelligence (AI) systems quickly became that latest snake oil for sale on the carnival podium.

There's so much to criticize about generative AI, but I focus now merely on the pseudo-scientific rhetoric adopted to describe the LLM-backed user-interactive systems in common use today. “Ugh, what a convoluted phrase”, you may ask, “why not call them ‘chat bots’ like everyone else?” Because “chat bot” exemplifies the very anthropomorphic hyperbole of concern.

Too often, software freedom activists (including me — 😬) have asked us to police our language as an advocacy tactic. Herein, I seek not to cajole everyone to end AI anthropomorphism. I suggest rather that, when you write about the latest Big Tech craze, ask yourself: Is my rhetoric actually reinforcing the message of the very bad actors that I seek to criticize?

This work now has interested parities with varied motivations. Researchers, for example, will usually admit that they have nothing to contribute to philosophical debates about whether it is appropriate to … [anthropomorphize] … machines. But researchers also can never resist a nascent area of study — so all the academic disclaimers do not prevent the “world of tomorrow” exuberance expressed by those whose work is now the flavor of the month (especially after they toiled at it for decades in relative obscurity). Computer science (CS) academics are too closely tied to the Big Tech gravy train even in mundane times. But when the VCs stand on their disruptor soap-boxes and make it rain 💸? … Some corners of CS academia do become a capitalist echo chamber.

The research behind these LLM-backed generative AI systems is (mostly) not actually new. There's just more electricity, CPUs/GPUs, & digital data available now. When given ungodly resources, well-known techniques began yielding novel results. That allowed for quicker incremental (not exponential) improvement. But, a revolution it is not.

I once asked a fellow CS graduate student (in the mid-1990s), who was presenting their neural net — built with DoD funding to spot tanks behind trees —, the simple question⁰: Do you know why it's wrong when it's wrong and why it's right when it's right?. She grimaced and answered: Not at all. It doesn't think.. 30 years later, machines still don't think.

Precisely there lies the danger of anthropomorphization. While we may never know why our fellow humans believe what they believe — after centuries that brought¹ Heraclitus, Aristotle, Aquinas, Bacon, Decartes, Kant, Kierkegaard, and Haack — we do know that people think, and therefore, they are. Computers aren't. Software isn't. When we who are succumb to the capitalist chicanery and erroneously project being unto these systems, we take our first step toward relinquishing our inherent power over these systems.

Counter-intuitively, the most dangerous are the AI anthropomorphism that criticize rather than laud the systems. The worst of these, “hallucination”, is insidious. Appropriation of a diagnostic term from the DSM-5 into CS literature is abhorrent — prima facie . The term leads the reader to the Bizarro world where programmers are doctors who heal sick programs for the betterment of society. Annoyingly and ironically — even if we did wish to anthropomorphize — LLM-backed generative AI systems almost never hallucinate. If one were to insist on lifting an analogous term from mental illness diagnosis (which I obviously don't recommend), the term is “delusional”. Frankly, having spent hundreds of hours of my life talking with a mentally ill family member who is frequently delusional but has almost never hallucinated — and having to learn to delineate the two for the purpose of assisting in the individual's care — I find it downright offensive and triggering that either term could possibly be used to describe a thing rather than a person.

Sadly, Big Tech really wants us to jump (not walk) to the conclusion that these systems are human — or, at least, as beloved pets that we can't imagine living without. Critics like me are easily framed as Luddites when we've been socially manipulated into viewing — as “almost human” — these machines poised to replace the artisans, the law enforcers, and the grocery stockers. Like many of you, I read Asimov as a child. I later cheered during ST:TNG S02E09 (“Measure of a Man”) when Lawyer Picard established Mr. Data's right to sentience by shouting: Your Honour, Starfleet was founded to seek out new life. Well, there it sits. But, I assure you as someone who has devoted much of my life to considering the moral and ethical implication of Big Tech: they have yet to give us Mr. Data — and if they eventually do, that Mr. Data² is probably going to work for ICE, not Starfleet. Remember, Noonien Soong's fictional positronic opus was altruistic only because Soong worked in a post-scarcity society.

While I was still working on a draft of this essay, Eryk Salvaggio's essay “Human Literacy” was published. Salvaggio makes excellent further reading on the points above.

Footnotes:

⁰I always find that, in science, the answers simplest questions are always the most illuminating. I'm reminded how Clifford Stoll wrote about the most pertinent question at his PhD Physics prelims was “why is the sky blue?”.

¹I really just picked a list of my favorite epistemologists here that sounded good when stated in a row; I apologize in advance if I left out your favorite from the list.

²I realize fellow Star Trek fans will say I was moving my lips and nothing came out but a bunch of gibberish because I forgot about Lore. 😛 I didn't forget about Lore; that, my readers, would have to be a topic for a different blog post.