Avatar

The unlit path

@theunlitpath

recent developments in mammal communication have rendered a grammatical sketch of rabbit language possible, as long awaited by those following the field. we analyze rabbit language as consisting of a single morpheme ("(tense silence)") itself composed of a single phoneme (a tense silence). the typological classification of this language is not yet determined, but resemblances to any human languages are scant, leaving the isolate theory most likely. attentive readers may inquire about the linguistic status of other communicative gestures observed in bunnies, such as hopping around or jumping around or bouncing around or stomping their feet cutely. however, we suggest these gestures are non-linguistic, as they lack a systematic structure. the structure of rabbit language is such that all utterances consist (by definition) of a single morpheme, arranged in singleton style, giving a total of one grammatical utterance ("(tense silence)"). chomskian universals appear to remain satisfied: recursion occurs in bunny language, as (given time in the universe is infinitely divisible), any utterance nontrivially contains a morphemically identical utterance (namely, "(tense silence)"). no other utterances are grammatical. it is because of this syntactic structure, as yet unevidenced in bouncing around and jumping around and hopping around and stomping their feet cutely, that we can confidently assign bunny language to the category of "language".

the satisfaction of the narrow faculty of language (FLn) [1] in a non-human animal is of particular theoretical significance and presents a major upset to researchers working in the field of universal grammar. at this time, understanding of meaning in bunny language has not progressed past basic semantics. with present knowledge any translation more than the most primitive semi-structured utterances remains elusive. due to the differences between human and bunny modes of behaviour (we don't even know what the fuck they do all day and what the fuck they are up to), substantive translation may be impossible. though, working in a more promising direction, human researchers have gained native-level faculties in jumping around and hopping around and bouncing around and stomping their feet cutely.

for more present developments in the field of lagomorph syntax, please see my recent publications with dr. amelia hickenlooper in animal communication.

— sofia chkq, reprinted from speech and sign magazine, chicago, usa

[1] hauser, marc; chomsky, noam; fitch, william tecumseh; "the faculty of language: what is it, who has it, and how did it evolve?", in science iss. 298 pp. 1569-1579, doi:10.1126/science.298.5598.1569

Avatar
Reblogged

Needles are pointless; if you can thread a needle, you can push thread through a hole smaller than the needle (and thus any the needle can pass through), so what's the needle for? Just push the thread through the fabric. Duh

i keep a list!

  • analáugns (Gothic): hidden
  • anchorhold (English): the cell of an anchorite, in which they are entombed as a kind of living saint
  • armōsts (Gothic): poorest
  • bearonæss (Old English): wooded headland
  • beinahrúgu (Old Norse): pile of bones
  • carcern (Old English): prison
  • crepuscular (English): pertaining to twilight
  • darkling (English): in darkness
  • daroð (Old English): javelin, arrow, projectile, dart
  • deliquesce (Old English): to become liquid, especially through organic decomposition
  • deosil (English): from Scottish Gaelic deiseil, literally "in the direction of the Sun, southwards, clockwise" but in English meaning "clockwise."
  • effloresce (English): to burst forth into bloom, to flower
  • etiäinen (Finnish): a form of folkloric apparition
  • exarch (English): a Byzantine provincial governor, particularly of an exarchate like Ravenna or Africa. From Greek ἔξαρχος.
  • fralusanō (Gothic): lost, gone away (f. sg.)
  • geniza (English): borrowed from Hebrew "גניזה", a storeroom containing books which cannot be used because of their condition, but which nevertheless cannot be destroyed because they contain the name of God
  • gnist (Danish): spark; see OE gnāst, ON gniesta, SWE gnista, OHG gneisto, MHG gneiste
  • hellwara (Old English): 'of the inhabitants of Hell' (gen. pl. of f. hellwaru or m. hellwaran)
  • hnasqus (Gothic): soft (see OE hnesce, "soft," modern English dialect nesh, "wimpy, soft")
  • idaltu (Saho-Afar?): elder, firstborn (H. s. idaltu is an obsolete classification of the "Herto man")
  • idreigonds (Gothic): repentant
  • iktsuarpok (English): borrowed from Inuktitut ᐃᒃᑦᓱᐊᕐᐳᒃ itsoarpok; "to go outside repeatedly to check if a visitor has arrived"
  • incunabulum (English): an early printed book; something in its infant stages. From Latin incunabula, "swaddling clothes"
  • incus (Latin): anvil
  • inwitwrāsen (Old English): "chain of deceit" (poetic)
  • Iolanthe (Greek): proper name meaning "flower of the violet"
  • irgendwo (German): somewhere, anywhere
  • Listopad (Polish): November
  • mæw (Old English): seagull
  • mamihlapinatapai (Yaghan): glossed as "a look shared by two people wishing the other would initiate something they both desire but which neither wishes to begin." The word is a regular derivation from ihlvpi, "to feel awkward, to be at a loss," with various affixes of voice, aspect, and so forth; it might more accurately be translated "to make each other both feel awkward."
  • narthex (English): the antechamber of a church; from the Greek word for "giant fennel" or "box of ointments"
  • neorxnawang (Old English): "field of heaven" (poetic)
  • opalescent (English): iridescent in a manner resembling opal
  • orcnaw (Old English): evident
  • razda (Gothic): voice
  • reordberend (Old English): 'voice-bearer', i.e., a human being (poetic)
  • ríastrad (Irish): battle frenzy, berserker rage, 'warp spasm'
  • Saoshyant (English): eschatological figure of Zoroastrianism; from Avestan saoš́iiaṇt̰.
  • Sargasso (English): region of the western Atlantic; from Portuguese sargaço, but of unknown ultimate origin.
  • searonet (Old English): 'web of guile, web of deception' (poetic)
  • talast (Old English): 2nd sg active indicative, "thou reckonest, thou dost consider"
  • tīrfæst (Old English): glorious
  • tramountayne (Middle English): the north; the north wind; the north star (rare) (from Latin transmontanus, via Italian)
  • Tuscarora (English): a Native American people, from Skarure (their autonym) skarū’ren’, “hemp gatherers.”
  • Tyree (English): surname; found as a geographic appelation in Scotland (westernmost island of the Hebrides), Antarctica (Mount Tyree, named for a U.S. Navy admiral), and Star Trek (the planet Tyree, where the Orb of the Emissary was discovered)
  • Ushakaron (English): the proper name of a star; I think it's Xi Tauri, but my notes are bad in this respect and I cannot confirm that; my notes also say it is from the Akkadian word "avenger," which I cannot confirm
  • velico (Italian): sailing
  • westengryre (Old English): "terrors of the wasteland" (poetic)
  • whyssyne (Middle English): cushion
  • widdershins (English): literally "anti-sunwise;" in later usage, "counterclockwise," since that is opposite the motion of the Sun in the northern hemisphere.
  • þystro (Old English): darkness

I also have a much shorter list in the same file of my lexicographically favorite terms/phrases that consist of more than one word:

  • hapax legomenon (Greek): a word which occurs only once in a manuscript or particular textual corpus
  • táiknái andsakanái (Gothic): from Luke 2:34, "a disputed sign, a sign which will be spoken against"
  • uncleftish beholding (English): from "Uncleftish Beholding" by Poul Anderson, "atomic theory" as calqued into a notional form of English with only Germanic roots
  • varg í véum (Old Norse): from the Vǫlsunga Saga, "a wolf in holy places," i.e., an outlaw
  • wære fræton (Old English): from the Old English Exodus, "they devoured the treaty," i.e., they went back on their word, they broke their agreement
Avatar
Avatar
Reblogged

"You think Christmas will happen on the 25th again merely because it happened on the 25th every other year. Days aren't auspicious or cursed. Reality cares not for our human calendars. So it is mere superstition to think Christmas will happen on a particular day of our calendar. "

off-by-one error at the toilet paper factory results in a roll that's just 100 concentric cylindrical sheets. useless

Avatar
Reblogged

this seems pretty satisfying:

Space Studies Institute → Foresight Institute → MIRI/CFAR

but this seems to have a missing step:

Esalen Institute → MIRI/CFAR

what were the immediate predecessors?

I feel like I don't get where Anna Salamon and the early CFAR instructors were coming from. Where'd they learn to run classes and workshops?

Some rationalist blogs mention all these other classes that people in the CFAR orbit were going to, on "circling" or "authentic relating"... none of it really means anything to me... i guess i wouldn't understand the answer if I had it, it's probably just something like "a variety of organizations in the bay area self-help scene" just like Foresight Institute above is really standing in for a variety of organizations in the bay area tech nonprofit scene

can't forget burning man!!

this goes back before MIRI/CFAR, Michael Vassar was a burner!

im vague on what actually happens at burning man, but apparently,

All manner of things. Everything from sacred meditation, to single mingles, to Alcoholics Anonymous, to coffee enemas. (Not kidding about that last one)
Burning Man used to have the excellent Palanque Norte lecture series, which brought together techy people like Cory Doctorow and psychonauts like Teafaerie. Burning Man is important to understanding the SF Bay Area intellectual scene I think.

*Palenque Norte, which is still going on. It was founded by Terence McKenna, apparently, who popularized DMT and predicted the Singularity. Psychonauts and rationalists rubbed shoulders here and elsewhere in the Bay Area. The drug culture and, I guess, a "future shock" level were things they shared.

Avatar
Reblogged

Gonna make the numbering system in my fantasy book base 12 and never explain or elaborate

Avatar

I make most societies in my books base 12 or base 8

Avatar

Meanwhile the discord is speculating if the people just have 12 fingers now

The discord has forgotten about knuckle counting

Like for the months?

You have two left over, could go up to 14 with knuckle counting.

But 12 has very many factors, very good number.

Dozenal knuckle counting is usually done like this.

You can get to 12 on one hand using your thumb as the pointer to keep track. You can use the other hand as the second digit to count to 144, if you want.

what’s the deal with base 8?

The Yuki supposedly counted the gaps between fingers.

Did you know that office chairs are rated for a safe load? I was just at the store, and saw that they had one that claimed to be safe for a mere two hundred pounds. For a child, perhaps? I had been in engineering school long enough to know it was time to ask some serious questions.

"I don't know what you're talking about. Dynamic load? What's that?" asked the confused and humbled IKEA Brand Representative as she fumbled futilely for the belt-mounted Customer Service® pepper spray that I had already removed and pocketed while asking for directions.

Senseless. These folks are not in control of their own lives. For answers, I would have to go to the top, rattle some cages. Soon, I was jetting off to Munich for a meeting with the European Commission For The Rights of the Human Butthole. This would be a rare visit for a commoner, they told me. Very few have ever seen a true office-chair crash test.

Friends, have you ever seen a robot ass? I don't mean one attached to one of those horny-on-main Hajime Sorayama chrome ladies. I mean an industrial-strength seat integrity tester. They make these things in Ehime, coincidentally quite close to where Sorayama-san works. An entire factory dedicated to simulating the dynamic and static loads of the human ass, for the purpose of improving survival in common, everyday, office-chair explosions.

Explosions, you ask? Yes indeed. All of these suckers are loaded with a compressed gas cylinder, like the shock absorbers inside your car's suspension. One of these babies goes off, it's going to be a bad scene for you, the attending paramedic, any witnesses, and of course your butthole. That's why we were here, watching a test, figuring out how they get that mysterious "weight rating" that started me on this journey at the IKEA so many years ago.

There is a loud thump, which I feel more than hear through the reinforced concrete wall of the observation bunker. Nobody needed to expose their eyes to the threat: advanced high speed cameras captured several hundred frames of the total atomic level destruction of the crash ass. They have all they need to know.

A young buck, one of the junior scientists assigned to do all the work as dour white-haired experts frown behind him, keys up the PA system.

"Test chair #XC2832 dash 9, test 2: one hundred fifty-five kilos. Total loss of derriere."

Fire extinguishers have already deployed, clearing the burning debris left behind from the failure of the Gǒushǐ Heavy Industrial Concern compressed-gas cylinder deep within that test chair. The alarms are reset, and a new seat is wheeled into the test chamber. It is three and a half hours until lunch.

As I leave the lab, I notice that there are no office chairs anywhere to be seen inside the work areas. Everyone stands. They know the risks.

When my domestic robot makes noodles for me, call that biang-biang control.

Avatar
Reblogged

The problem with "leftism" as a label is that it both encompasses people who see socialism as continuous with liberalism (even if they "hate liberals") and various strands of communitarian critics of liberalism. And the liberal socialist side, it seems like their opponents aren't really leftists, but really kind of family-values conservatives who want contingent modes of social relationality to dominate us. But for the communitarian side, the opponents are cast as generically atomized liberals who don't value the kinds of community that they take to be necessary for creating socialism. And they can end up slinging a bunch of insults at each other over how the other isn't real leftism. But really it's like socialism has roots in both agrarian conservatism and liberal/republican traditions.

disentangling political traditions based on how they handle various scenarios like "an elderly widow is to be evicted from her farm due to debt / property taxes" or "a mine is to be closed as it is no longer profitable to operate it" or

The mine is closed — good! No more resources go to waste on an unprofitable mine. The miners find jobs elsewhere because the central bank keeps the economy running hot and there are few regulatory barriers. If there are no jobs nearby, companies compete to hire everyone and pay for their moving costs. The miners form a union to negotiate with the new employers. In the meantime, they live off private optional employment insurance which everyone has. It's good insurance despite being unregulated because the audit results are widely communicated. And if that fails, there's UBI, which the civ is hopefully prosperous enough to fund to the level that no one starves.

The widow can't afford the land value tax anymore, so she sells the farm. It's not worth very much because it's undeveloped and taxes are so high. For the same reason, the farm is not most of her net worth. She moves into a small house or apartment that she rents with her savings or life insurance payout. It's a personal tragedy for her because she's very attached to the farm. But the bigger tragedy is that she lost her husband and her vigor, and no political system can bring them back — yet.

Avatar
Reblogged

The problem with "leftism" as a label is that it both encompasses people who see socialism as continuous with liberalism (even if they "hate liberals") and various strands of communitarian critics of liberalism. And the liberal socialist side, it seems like their opponents aren't really leftists, but really kind of family-values conservatives who want contingent modes of social relationality to dominate us. But for the communitarian side, the opponents are cast as generically atomized liberals who don't value the kinds of community that they take to be necessary for creating socialism. And they can end up slinging a bunch of insults at each other over how the other isn't real leftism. But really it's like socialism has roots in both agrarian conservatism and liberal/republican traditions.

disentangling political traditions based on how they handle various scenarios like "an elderly widow is to be evicted from her farm due to debt / property taxes" or "a mine is to be closed as it is no longer profitable to operate it" or

The mine is closed — good! No more resources go to waste on an unprofitable mine. The miners find jobs elsewhere because the central bank keeps the economy running hot and there are few regulatory barriers. If there are no jobs nearby, companies compete to hire everyone and pay for their moving costs. The miners form a union to negotiate with the new employers. In the meantime, they live off private optional employment insurance which everyone has. It's good insurance despite being unregulated because the audit results are widely communicated. And if that fails, there's UBI, which the civ is hopefully prosperous enough to fund to the level that no one starves.

Avatar
Reblogged

I think generalist robotics is gonna be a big deal in the next 10 years. I think they've more or less got all the important parts and just need to do the work of putting them together. I think that's gonna have a big impact on the world...

i feel like self driving cars are gonna be a big deal in the next ten years. I think generalist robotics are gonna be a big thing like... eventually... but... i mean, self driving cars seems so much easier and we're still not doing it

idk, maybe i'm being mislead by the completely different safety standards, but this seems like thinking too many tech generations ahead to me i guess

hmm I forgot about self-driving cars. I just have no idea about those. here's elon musk in 2015 saying:

“We’re going to end up with complete autonomy, and I think we will have complete autonomy in approximately two years.” That doesn’t mean city streets will be overflowing with driverless Tesla vehicles by 2018 (coincidentally, the company’s Model 3 should be on roads by then). Musk expects regulators will lag behind the technology. He predicts it will take an additional year for regulators to determine that it’s safe and to go through an approval process. In some jurisdictions, it may take five years or more, he says.

so 10 years ago if you were saying self-driving cars were gonna be a big deal in the next 10 years, you'd be wrong. maybe you'd be right this time... certainly things have come a long way.

comparing them to robotics... well "AI" struggles with reliability, and probably will continue to, though less and less. that's a big problem for some robotics tasks, but I have the feeling there are a lot where it's less important than self-driving cars. the unitree G1 robot is about 4 feet tall and 77 pounds, it's a lot less dangerous if that fucks up catastrophically every million operating-days than if a car does. of course, depending on what it's doing...

I've been meaning to read stud's terkel's "Working" for years now, have an old copy on the floor bookstack under my desk since I'm out of bookshelf space. in part to get a broader view of the world and the people in it, but more recently to get a broader view of just what specifically people do in their jobs. of course it'll have changed a lot since that book was written. I'd also like to review some economic data on what sorts of jobs people do. to get some idea of what's more or less automatable with middling dexterity and human oversight

yesterday evening I was more confident about the robotics thing than I should have been I think. at least more confident than I am now. but it feels plausible. it also doesn't need to be as successful, relatively speaking, to be a bigger deal than self-driving cars. though I don't have the best handle on that personally because I've never owned a car and lived most of my life in relatively car-light environments. so I'm probably underestimating the potential significance there

I spent the last ~7 years working in a robotics lab and I while I think things might get somewhat better over the next 10 years, we're absolutely not gonna have generalist robotics. If we're lucky, we'll have robots for automated pick-and-place tasks like Amazon warehouses. To my mind, the basic issues are these:

  1. The big thing LLMs have showed us is that you can brute force a lot of problems with enough data + compute. The problem is that generalist robotics do not have internet-scale data and probably never will. We got internet-scale datasets because investors subsidized people messinging around on the internet leaving lots of data for 20-25 years, which then turned into LLM training sets. You're never going get millions of people messing around with home robots enough to produce this scale of data unless the robots are already cheap and useful, and they're not going to be cheap and useful unless you already have internet-scale data to train them on. People are trying to build robot training datasets from video, but the fundamental problem is that you need internet-scale datasets of of people controlling robots in real physical environments, which just isn't going to happen.
  2. The main way to bypass this would be to use reinforcement for training, which lets you skip a lot of the data requirements of traditional ML if you do training in simulation. However, the big reason LLMs got smarter all of a sudden is that internet data is inclusive by default. Anything people talk about, the LLM will learn about, unless you make a special effort to exclude this from the dataset. This is *not* true of RL training in simulation. The simulation is something that a grad student wrote up and includes only the elements of the environment that are relevant to one specific task. So you only get data for things that you've explicitly thought to include, which is the same basic problem as the Good Old Fashioned AI of the 50s-70s -- there are fundamentally too many facts about the world, and those facts are too implicit in our minds, for it to be possible to articulate them explicitly. In the same way we would never expect a video game to ever provide a realistic simulation of daily life, we also can't expect training simulations to do this. So RL training in simulation can help us get data for specific tasks, but it won't help us with generalist robotics.
  3. This leaves online learning/reinforcement learning in the real world as the one way to get inclusive-by-default data about robotic control. Unfortunately, the algorithms that we use are far, far too inefficient to solve realistic problems, especially general ones. Basically all of the commonly used algorithms take at least exponential time in the diameter (max number of steps to navigate between states) of an environment. AlphaZero can take double exponential, PPO is probably similar, just harder to analyze. So you can't just throw money at the problem, you have to find solutions that are faster than this. Techniques that make this polynomial exist, but they're not really scalable yet.

The one case where these barriers aren't really in place is self-driving cars. Tesla has a huge fleet that gathers tons of data that they can use to train driving models, and that data is from real environments so it's inclusive by default. It's in many ways the easy-mode version of robotics, one that's missing all the fundamental barriers that make general robotics so hard. The fact that they still haven't figured out the much more limited problem of self-driving is to me a very bad sign.

Personally, I think that after we solve self-driving, it'll be another 15-20 years minimum to sort out all the additional problems generalist robotics brings. But honestly I think it could be much more than that, it's totally possible that the cost for household robots doesn't drop to the level that allows large dataset collection for the forseeable future. I think robotics may just be in the same camp as space exploration, where it's cool but economically does not make sense unless something changes the cost analysis in an unforeseeable way at some point in the distant future.

Re: Point 3, what about DayDreamer? It learned to walk in an hour!

Sponsored

You are using an unsupported browser and things might not work as intended. Please make sure you're using the latest version of Chrome, Firefox, Safari, or Edge.