Important academic paper
Get me off Your Fucking Mailing List
Figure 2: Get me off your fucking mailing list.
Figure 2: Get me off your fucking mailing list.
having written something out I think that the "c isn't a language anymore" article, in a zoomed out abstract view, applies to JavaScript even though js is only 28 years old.
The long-term popularity of any given tool for software development is proportional to how much labour arbitrage it enables. […] The components sourced from an intern fixing ChatGPT’s output just enough for it to run and the exhaustively tested ones from a senior developer are equivalent in the eyes of management.
spent my morning coffee today reading Ed Zitron's piece on the "AI" bubble, whose bubble-ness the relevant ruling class seems remarkably in denial of across the board. it got me thinking about the whole absurd cycle that brought us here, VC disruption startups to bitcoin to NFTs and now "AI", each hyped as a civilization-scale revolution based on little more than the promises of rich tech CEOs and their sycophants. it must be understood first and foremost that capitalists are crooks and skittish herd animals, that always beneath the story they sell the world their only real aim is to make money. like yes, perhaps the Sam Altmans of the world do on some level believe their own hype, believe in the supposed promise of their tech... but if the tech fails, if the promise is broken, that's a minor inconvenience because at least they made money.
but there is something dreadfully fascinating about the story they're telling, right? there's this insulated ecosystem where these tech guys fail spectacularly, pissing millions or billions of dollars into the toilet and laying off countless employees, yet are only rewarded for doing so. each failure is met not by self-reflection or material analysis, but by a cult-like doubling down, an exponential increase in the scale of the promise (and the capital required for its realization), no, for real, this time we're going to fix everything.
the promises they make about how bitcoin, NFTs, "AI" will change the world are beyond absurdity, totally unmoored from the material conditions of the world they exist in, yet it has to be the case that a lot of these people do believe it. however disconnected from reality they are, the fact of reality cannot help but assert itself at the margins-- climate change, rising fascism, political division, you name it. whether they identify the problem as "too much woke" or "not enough tech" (probably both), either way their actions are a response to tensions at the very least orthogonal to the very real tensions being felt by the global working class.
profit motive aside, i think the core of what's going on here is a class of isolated, secular rich people desperate to invent a god that will liberate them from having to care. no, we don't need to radically transform society, redistribute wealth, dismantle the organs of capitalist society... we just need "AI"! they're like children being told santa claus isn't real, except they're fucking adults at the center of a cult of personality around the very concept of world-changing tech. their delusions are echoed by court stenographers larping as journalists, an echo chamber they themselves helped to create by radically undermining the traditional news business model, by encouraging every entity under the sun to turn into a Wall Street financial vehicle.
these people more or less bought the public facing apparatus that reproduces their hype, they've pushed their libertarian ideology absolutely everywhere it can fit, and while it has been immensely profitable in the short term (by a layman's definition of profit, anyway), there remains the unfortunate inconvenience of a material reality that trundles on without care for their machinations. there is no use-case for "AI" on the scale they're demanding, if any at all. OpenAI could own every single outlet tasked with covering their tech and push a unilateral & uncompromising "AI" vision with no opposition, and it still wouldn't change that reality. there is no making healthcare, transit, schools, or any other public infrastructure "profitable," no matter how many overtures are made to that effect. there is no fixing climate change that doesn't involve radically reducing car dependency (which means directly attacking the mob-like car dealer fiefdoms that exert heavy influence over local politics), investing heavily in nuclear energy (which means directly attacking the fossil fuel companies that invest enormous sums of money into maintaining the public's irrational fear of nuclear power plants), and strictly limiting the ability of profit-seeking companies to exploit public infrastructure to its breaking point (which means directly attacking tech companies voraciously consuming electrical grid capacity for their pet projects, as well as companies that fuel droughts by stealing public water and reselling it at a massive profit). you just can't affect the base by gussying up the superstructure, man! and this cold, hard reality is one that no capitalist is prepared for or interested in facing, let alone the libertarian tech elite whose vision of the future is "sci-fi stories are cool, let's do that, no i don't know what "dystopia" means."
"AI" is, in many ways, the ultimate endpoint of the capitalist's dream, a self-generating economy with no need to cater to specific customer needs, no need to hire pesky laborers (especially unionized ones), no need to bother with regulations. it's an infantile vision of the world doomed to failure, yet it's an infectious vision. wouldn't it be nice if we could invent our way out of this mess? wouldn't it be nice if things could just keep on going as they are, stewarded into a perfect technofuturist utopia by a caring God of our own invention? if you're a marginalized person of any stripe, your answer is probably an instinctively repulsed FUCK NO, but their audience is the disconnected suburbanite, the white (likely Christian) middle class home-owner aspiring to perfect independence. in this way, the secular tech world has become an extension of fundamentalist evangelical ideology. there is no cost too great if it means ushering in our Savior. all the death and misery and inequality and exploitation, disease and genocide and climate change and gun violence and bigotry, these are Signs and Portents my friend, signalling the end times.
"the end times" are, really, just a poetic verbalization of the civilization-level need for a people's revolution in everything, an objectively extant pressure correctly identified but wildly misdiagnosed. it's a refusal to countenance any possibility that we can become better than ourselves in a way that actually matters, not through technology or devotion or purity, but through action. it is a desperate plea to the universe for an all-knowing adult to just take charge already so we can stop pretending like any of this is actually our responsibility. please, someone, save us from ourselves, bring on the Rapture, bring on the Grey Goo, bring on the Zombie Apocalypse, bring on the Alien Invasion, something, anything else, just please, make it quick!
their pleas will not be answered. there will be no spectacular end, just as Twitter didn't explode the way we all wanted, just as Covid didn't wipe out civilization overnight the way we all feared, just as the Roman Empire never truly fell but simply crumbled under its own weight over centuries. no amount of delusional thinking & frantic technological interventionism on the part of the ruling classes can change this fact. they tell us, and themselves, stories of the world they wished we live in, but material reality trundles along unchanged as it always has and always will do.
no one can save us but us.
The unified shader architecture was introduced with the Nvidia GeForce 8 series, ATI Radeon HD 2000 series, S3 Chrome 400, Intel GMA X3000 series, Xbox 360's GPU, Qualcomm Adreno 200 series, Mali Midgard, PowerVR SGX GPUs and is used in all subsequent series.
s3 made a modernish dx10 gpu with unified shaders?!
The scenarios sketched out in this hotel room in Potsdam all essentially boil down to one thing: people in Germany should be forcibly extradited if they have the wrong skin colour, the wrong parents, or aren’t sufficiently “assimilated” into German culture according to the standards of people like Sellner. Even if they have German citizenship.
What is true is that a majority of people—across virtually all demographics and ages—agree life in America is getting worse, and we must steel ourselves for the grim future ahead. […] This is reflected much closer to home, as well; polling shows that a quarter of Americans fear being attacked in their own neighborhood.
And our vehicles reflect these anxieties.
https://jon-e.net/surveillance-graphs/
(not mine ofc)
about the history of search engines and how those evolved into the knowledge graphs and voice assistants and chatbots we have today, and the consequences of so much of our information being enclosed by huge tech/search/ad companies
We’re being made an offer we can’t refuse: it’s a shame that you can’t find anything on the internet anymore, but the search companies are here to help.
So I stumbled a few weeks ago on this video about the world's smallest conlang and I've been thinking about it ever since.
(You will probably have the idea more than fully by the ten minute mark.)
This fascinated me. I think languages are really interesting but I'm also really bad at them. I cannot hack the memorization. I've mostly avoided conlangs (that's Constructed Languages, like Esperanto or Klingon) because like, oh no, now there's even more languages for me to fail at memorizing. So like, a language so minimal you can almost learn it by accident, suddenly my ears perk up.
Toki pona is really interesting, on a lot of different axes. It's a thought experiment in Taoist philosophy that turns out to be actually, practically useful. It's Esperanto But It Might Actually Work This Time¹. It has NLP implications— it's a human language which feels natural to speak yet has such a deeply logical structure it seems nearly custom-designed for simple computer programs to read and write it². Beyond all this it's simply beautiful as an intellectual object— nearly every decision it makes feels deeply right. It solves a seemingly insoluble problem³ and does it with a sense of effortlessness and an implied "LOL" at every step.
So what toki pona is. Toki pona is a language designed around the idea of being as simple as it is possible for a language to be. It has 120 words in its original form (now, at the twenty year mark, up to 123), but you can form a lot of interesting sentences with only around twenty or thirty (I know this because this is roughly the size of my current tok vocabulary). The whole tok->en dictionary fits on seventeen printed pages⁴. There are no conjugations or tenses to memorize. There are no parts of speech, generally: almost every single word can be used as almost any part of speech, drop a pronoun or a preposition⁵ in the right place and it becomes a verb, drop a verb in the right place and it becomes an adjective. There are almost no ambiguities in sentence structure; I've only found one significant one so far (using "of" [pi] twice in the same clause) and the community deals with this by just agreeing not to do that. There's in one sense quite a lot of ambiguity in the vocabulary but in another sense there's none; for example nena is listed in the dictionary as meaning hill, mountain, button, bump, nose, but the official ideology of Toki Pona is that this is not a word with five definitions, it is a word with exactly one definition, that being the concept encompassing all five of hills, mountains, buttons, bumps, and noses. Toki pona words are not ambiguous, they're just broad. I will now teach you enough toki pona to understand most sentences (assuming you consult the dictionary as you go) in a paragraph shorter than this one:
Every sentence has an "verb" or action clause. Most sentences also have a subject clause, which goes before the verb separated by "li". Some sentences also have an "object" or target clause, which goes after the verb clause separated by "e". So "[subject] li [verb] e [object]". All three clauses are any number of words, the first word is the noun (or verb) and the rest are adjectives (or adverbs). If the first word in a sentence is mi or sina you can omit the li in that case only. There is no "is", a sentence like "a flower is red" is just "[a plant] li [red]" because there are no parts of speech and adjectives are verbs. Pronounce everything the way you would in English except "j" sounds like "y". If you see "tawa", "lon", "tan" or "kepeken" in adjective/adverb position that's a preposition, so treat whatever comes after it as a fourth clause. "a" and "n" are interjections and can go anywhere in a sentence. If you see "o", "pi", "la", "taso, "anu" or "en", then you've found one of the few special words (particles) and you should probably either read the book or watch the jan Misali instructional videos on YouTube.
That's like… almost it!⁶ Watch one youtube video, or read a half page of grammar summary and keep the dictionary open on your phone, and you're basically ready to jump into conversations. (You'll need those last six particles to make nuanced sentences, but sentences do not have to be nuanced.) In my own case I watched the Langfocus video linked at the top and then lazily over the next week watched the jan Misali overview and two of their instructional videos then read a few chapters of the book, and was already able to construct a full paragraph and be understood by another person:
o! mi sona e toki pona. mi lukin sona e toki mute... taso mi toki ike e toki ale. toki mute li jo e nimi mute. nimi mute li pona ala tawa mi. toki pona li pona e mi tan ni. toki pona li jo nimi lili. nimi lili li pona e me :)
(Translation: Hello! I am learning toki pona. I've tried to learn many languages… but I speak all of the[m] badly. Most languages have a lot of words. Lots of words is not good for me. Toki pona is good for me because of this. Toki pona has few words. Few words is good for me :)⁷)
So that's week one. By the end of week three (that'll be this past Sunday) I had hit three milestones:
(That last one actually kind of excited me when I realized, after the fact, that it had been happening for several days before I noticed it.)
All of the above happened on the toki pona Discord, which seems to be the largest and healthiest outpost of toki pona community online⁸ and, I assume therefore, the world. Here's my toki pona joke; you'll notice that I begin by preemptively apologizing for it, but by the end it becomes clear I was, at that moment anyway, apparently completely on the Discord's level:⁹
¹ Esperanto was meant to be an auxillary language simple enough the entire world could learn it as a second language. But it does not seem to me to be actually all that simple to learn (the verb tenses may be free of exceptions, but there are still like six of them) and it seems hard to convince myself to study it when I could put that effort toward a more widespread language. Toki Pona on the other hand is a language simple enough the entire world actually could learn it, had it reason to, and the buy-in cost in time is a lot lower. Meanwhile Toki Pona just seems to get certain international-accessibility things right, probably due to the advantage of having over a century of conlang practice largely inspired by Esperanto to draw from. The difference that's most interesting to me, but which I'm also least qualified to evaluate, is that the Toki Pona author had some basic familiarity with, and gave Toki Pona input from, east Asian languages. Both Toki Pona and Esperanto draw roots from a variety of languages but in Esperanto's case this winds up being mostly European languages, mostly Italian, whereas Toki Pona has Mandarin and Cantonese in its mix. Both languages were intended to have a minimal, universal phonology but Esperanto's is not very universal at all and has some consonant sounds that are going to be be totally unfamiliar unless you speak Polish or a language related to Polish. Toki Pona's phonology is actually minimal-- it has the letters in the sentence "mi jan pi ko suwi lete" and nothing else and the legal syllabary is basically hiragana without the combination syllables. Now, what I don't know is whether any of this actually helps speakers of Asian languages in practice or if it just makes surface choices that make it seem that way to a westerner like me. In practice if you find a toki pona speaker their native language is almost certainly English, so there doesn't seem to have been a great deal of testing as to how learnable it is to people from varied regions.
² I kept imagining toki pona as being like LLVM IR for human languages or something.
³ Designing an actually learnable, universal(?) international auxillary language.
⁴ By which I mean if you print the unofficial web version it's 17 pages long. But you can fit it in sixteen if you remove the web-specific introductory paragraph, the "joke word" kijetesantakalu, and the web footer with the "trans rights now!" banner in it.
⁵ There's an awkward nuance in the specific case of prepositions: if you drop a preposition into verb position it still winds up "behaving like" a preposition, it's just that the preposition describes an action. It's probably more accurate to say that the preposition phrase becomes a verb rather than the preposition itself. So for example "mi tawa kalama" means "I went toward the sound" [kamala is a noun and the object of the preposition] not "I went loudly" [as it would be if kamala were an adverb modifying tawa as a verb]. Going into detail on this because I don't want to mislead anyone but "you can form a sentence using an unmodified preposition as a verb" is such a fun fact I wanted to include it.
⁶ I guess I could say "what I'm calling 'clauses' are called 'word groups' in official sources; word groups are always separated by a particle or preposition" but I'm not sure that's interesting unless you're a linguist.
⁷ There's a lot of repetition in this paragraph of both kinds that show what's cool about toki pona and what's kinda impractical about it. Cool, it has some sentences where a single word shows up twice conveying two different clear things based only on where it is in the sentence, like "mi toki ike e toki ale" where toki is a verb ("I speak") the first time and a noun ("language") the second. Less cool, because so much of the meaning of the hyper-versatile toki pona words is based on context, you sometimes wind up having to include multiple sentences saying nearly the same thing just to make sure the reader selects the correct context. I may have overdone it here.
⁸ There's also a low-traffic Mastodon instance, multiple forums including one run by the Merveilles crowd, and something called "Wikipesija"
⁹ Translation/explanation:
