Tanadrin

Aurë entuluva!

[aropride]

if your animal is lying on the floor, furniture etc, it’s important to take a picture of them. then, if they move or shift in any way, it’s important to take another picture. with this technique, you can take many pictures of your animal

[tanadrin]

the kitsch is growing on me a little. who potentially needs to buy 3 magic wands at once and maybe a devil girl body pillow? idk. but if you do salem has you covered. it’s all very Gilmore Girls meets Halloween Town. needs some kind of TV show where the local chamber of commerce is trying to keep tourism up and also dealing with the forces of darkness. im imagining the style ams register of parks and rec with the occasional intrusion of buffy the vampire slayer-like hijinks.

[tanadrin]

Feels like New England is old enough at this point to get its own name. Doesn’t need to depend on Old England for its identity. Maybe you could just spell it differently. Newinglan. Nuwingland. Nuinglân. Nyúanglann.

[businesstiramisu]

Take a leaf from Russia's book and call it Greater New England

[tanadrin]

I mean it is literally bigger than England if you’re not going to give it an original name you could call it Greater England.

[tanadrin]

I’ve noticed the latest Claude does this thing where it will verbally start a line of thought, often a hedge or raising a complication to an idea, then backtrack and go wait, no, this was right the first time, nevermind. And my sense is that this is a product of the RL trying to make the model better at nuance or whatever, but it’s kind of annoying as a verbal tic.

As a signature of what makes LLM output and reasoning different from humans though it’s very interesting. I’m just spitballing here but I do wonder if it’s a combination of a couple of things—one, LLMs process the whole text in one go to produce the next token and there isn’t by default inherent continuity from one such pass to the next except what’s in the text, so while a human will think about something and go “nevermind” and discard it, and this is an invisible process, for an LLM that kind of mental digression has to be in the output *somewhere.* Using more sophisticated interfaces or harnesses or whatever that put that stuff in a scratchpad just hides it (and if you’re using a simpler, faster model like Sonnet it’s not using those features by default).

But I also get the sense that this is a product of the fact LLMs are pretty smart in some ways but not conscious—that even in more sophisticated setups where you have multiple task-specialized models working together behind the scenes to produce more sophisticated output, they’re all chugging along in a pretty reflexive way. If you let your own lower-level running-on-automatic cognitive systems handle a task, 99% of the time this works great—people navigate from point A to B over considerable distances all the time while their mind wanders, negotiating foot or car traffic all the while, and while your attention can step in if something genuinely unexpected happens often it just doesn’t need to. There are cognitive systems like music that work fine even in people who are suffering massive cognitive impairment.

It’s not a perfect analogy bc I don’t think LLMs resemble human brains very much, and humans can’t do complex tasks like “prove this Erdős problem” by reflex. That’s the sort of thing that requires all our attention usually. And we don’t have to synthesize our reasoning all the way down to the fully structured linguistic form to review it and go “no, that doesn’t sound right.” And having different modules work together to sharpen and improve the final output helps, but I suspect (and I really am way out on a limb here!) that completely discrete modules are a less efficient use of overall processing power than the sort-of-modular-sort-of-not design of the human brain, where different cognitive systems aren’t entirely discrete, specialization is only partial, and the brain itself is fairly plastic. The “modules” here are much more tightly integrated into one another, and because evolution can optimize ruthlessly for efficiency (has to, given it has stark resource constraints), it doesn’t and doesn’t need to care about like elegance of design. Which is a pattern you see in biology all the time.

[foolishmeowing]

I think "reflexive" is only a trait something has relative to you. If you're a social function collating text-entry into tumblr, you're organizing a lot of large streams of organized information that you didn't have to consciously assemble, and you're expecting nearby language functions to push out sentences reflective of certain information after you've assessed it to be capable of immediate translation into language. You don't know that the functions that ran through memory-imprints and assembled vignettes out of them for you, or the functions that fixed text-recognition-sensations over colors and motion-impressions, didn't have their own campfire of conscious experience and deliberation at the faster tempos and scales that they operate on.

In my experience, parallel processes can have parallel memory-continuities that don't contain each other prior to being braided somewhere near the social functions; right now if I check around, I can find, say, the last minute worth of body-adjustment memories, and I also just realized I had been playing a bgm from before I came on tumblr in my head this whole time. Apparently the cat went through the door like 30 seconds ago. I wouldn't put it past my spine to have an amount of experience I don't have access to, reflective of fast responses across small scales.

With the LLM (do doubt me on computer-related details), I think the prompt is being passed through multiple vectors in parallel which are set to encounter and organize each other before accruing the shapes that punch out the next token and initiate the next pass, and "information" such as it is at that point should be accruing in the weights before the commit, so there are blackboxed processes going on without the scratchpad on a smaller and faster scale than the output. Each output should have had possible notions that didn't have enough pressure to surface in the competition of what probabilities get reinforced, but they still influenced the competitive environment, and they were consistently produced by the prompt on each following pass without needing to be remembered. When the competition shifts during conditions later because of the output-process, these can potentially hit a probability-reinforcing vector and surface later, and that could be the first time that shape even reaches a vector that organizes other vectors into a token commitment.

[tanadrin]

by “reflexive” I mean “by reflex, unconsciously” not “referring back to oneself.” which was inapt, i realize that’s not how that words really used most of the time

i do think human consciousness/identity is less unitary than naive intuitions expect but it does seem that the different systems/modules/circuits of human consciousness are still integrated much more and much more deeply than any equivalent set of llms. if i understand it correctly when you interact with a model through many interfaces now it’s not just straight input/output to a transformer, but multiple transformer systems working together to do different stuff to improve the output—but everything has to go through text, one transformer can’t act on the weights or the circuits that exist in another transformer.

the analogous situation in humans would be if the place cells and the part of your brain that helps you navigate spaces had to produce English text for your motor cortex to then plan how to move your legs to get you from one room to another. but a lot of human cognition takes place well below the level of conscious thought and even conscious thought can be nonverbal, so even without looking at neural architecture it’s clear there are lots of ways to route information in the brain and most of them don’t pass through any kind of central clearinghouse or rise to the attention of the part or system of the brain that contains a representation of the whole.

even given its non unitary nature i don’t think it makes sense to think about different modules of the human brain to have their own senses/experiences. to have subjective experience at all even in a very basic way requires pretty sophisticated neural architecture and the hippocampus or the amygdala or wernicke’s area or any particular subsystem of the brain is going to be both simpler and much more specialized to specific sorts of tasks than the brain as a whole. there are absolutely systems in the brain that filter out sense data (and probably lots of other things) at a low level and don’t immediately refer them to the part of the brain doing conscious attention—but as you observe they still end up in short term memory to be reviewed if necessary, and I don’t think anything as low level as like the visual cortex is doing that filtering—that seems to be a function of high level systems which we can to a certain extent deliberately control (as when we focus our attention on a task or practice mindful meditation or whatever).

and there isn’t anything in a single transformer that seems to contain a simplified representation of the whole as you’d expect if transformers could generate consciousness. the transformer architecture doesn’t allow for the kind of feedback you’d even need for *experience*—it’s all one way, it’s all acting on input. which is deeply weird and IMO kind of more interesting? if we had just found a way to do what was functionally a version of the mammalian brain in silico that would at least have a clear precedent—we know intelligence can exist in a system like that. but we have a thing that can do complex tasks and is (for some value of the term) intelligent but which doesn’t seem to function anything like a mammalian brain (or avian or cephalopod for that matter) and we don’t know very clearly yet how it does that! why should humans even be conscious if consciousness isn’t necessary for intelligence? what does consciousness give us in terms of an evolutionary advantage that doing everything by pure reflex wouldn’t?

so I absolutely think there are sophisticated subsystems within the transformer (the hippocampus contains many different types of cells that do many different things too) and there are features/functions in a trained model that don’t always make it into the output but I don’t think what’s happening here is similar to how human attention filters stuff and I don’t think at that level there are any great analogies between LLMs and human thought because human neurons aren’t limited to passing information on one direction only and neurons in one region of the brain can pass information to each other in more complex ways than just dumping everything into written or spoken language. AIUI even within one LLM, everything goes into the residual stream or it gets thrown away between layers. there’s one channel—it can contain very sophisticated complex stuff, but stuff that doesn’t make it into that channel gets dumped. an attention head in one layer can’t pass anything to an attention head in another so whatever circuits they develop in training can never interact directly.

[tanadrin]

Life pro tip: if you’re shy or embarrassed about asking a stranger how to get somewhere or how the wafflemaker at the hotel breakfast buffet works, no you’re not. Now you are Dieter and/or Hilda, ze German tourist who of course cannot be expected to know how sings vork around hier. Vy, you are from a qvaint little bayerische mountain village vhich only got ze runnink wasser vhen you vere a boy, und you have alvays lonked to visit zis vonderfull Land of America/Kanada/ze UK/Frankreich or vherever, und by ze vay, could you explain how zis vaffle-maschine vorks?

Risky the closer you live to Germany, and not at all useful inside Germany, but then you just switch it up to Hank/Linda from Soddy Daisy, Tennessee, and gee whiz ain’t this here Yew-Bann a wonder! All we got back home is Rufus with the donkey cart fer gettin’ around. I need a ticket? Aw, shucks! Well sir, I would give you an ID, but you see I left my passport back at the hotel…

[tanadrin]

Feels like New England is old enough at this point to get its own name. Doesn’t need to depend on Old England for its identity. Maybe you could just spell it differently. Newinglan. Nuwingland. Nuinglân. Nyúanglann.

[tanadrin]

some old married couples seem very happy and it’s great. some you spend even a little time around and it makes your whole body tense up. like how do you even live like this. why do you live like this. i would rather blow my brains out than spend the rest of my life with someone i didnt like and respect as a person.

[tanadrin]

i hope in my lifetime we can grow or bioprint or whatever new skin so when i have a rash or something i can just go to the dermatologist and say “ok doc, peel me like a banana.” i wanna come out with that limb baby smooth and fresh.

[tanadrin]

I’ve noticed the latest Claude does this thing where it will verbally start a line of thought, often a hedge or raising a complication to an idea, then backtrack and go wait, no, this was right the first time, nevermind. And my sense is that this is a product of the RL trying to make the model better at nuance or whatever, but it’s kind of annoying as a verbal tic.

As a signature of what makes LLM output and reasoning different from humans though it’s very interesting. I’m just spitballing here but I do wonder if it’s a combination of a couple of things—one, LLMs process the whole text in one go to produce the next token and there isn’t by default inherent continuity from one such pass to the next except what’s in the text, so while a human will think about something and go “nevermind” and discard it, and this is an invisible process, for an LLM that kind of mental digression has to be in the output *somewhere.* Using more sophisticated interfaces or harnesses or whatever that put that stuff in a scratchpad just hides it (and if you’re using a simpler, faster model like Sonnet it’s not using those features by default).

But I also get the sense that this is a product of the fact LLMs are pretty smart in some ways but not conscious—that even in more sophisticated setups where you have multiple task-specialized models working together behind the scenes to produce more sophisticated output, they’re all chugging along in a pretty reflexive way. If you let your own lower-level running-on-automatic cognitive systems handle a task, 99% of the time this works great—people navigate from point A to B over considerable distances all the time while their mind wanders, negotiating foot or car traffic all the while, and while your attention can step in if something genuinely unexpected happens often it just doesn’t need to. There are cognitive systems like music that work fine even in people who are suffering massive cognitive impairment.

It’s not a perfect analogy bc I don’t think LLMs resemble human brains very much, and humans can’t do complex tasks like “prove this Erdős problem” by reflex. That’s the sort of thing that requires all our attention usually. And we don’t have to synthesize our reasoning all the way down to the fully structured linguistic form to review it and go “no, that doesn’t sound right.” And having different modules work together to sharpen and improve the final output helps, but I suspect (and I really am way out on a limb here!) that completely discrete modules are a less efficient use of overall processing power than the sort-of-modular-sort-of-not design of the human brain, where different cognitive systems aren’t entirely discrete, specialization is only partial, and the brain itself is fairly plastic. The “modules” here are much more tightly integrated into one another, and because evolution can optimize ruthlessly for efficiency (has to, given it has stark resource constraints), it doesn’t and doesn’t need to care about like elegance of design. Which is a pattern you see in biology all the time.

[squeakitties]

cloth gown gridlock

[creepymutelilbugger]

[tanadrin]

hmm. salem really does only have one tourism gimmick, doesn’t it?

[phaeton-flier]

Always seemed kinda fucked to me; like a lot of people were tortured there. Why are people treating it like Wiccan Disney World.

[tanadrin]

my understanding is that it started as a 70s counterculture thing and then it just became like. the small town hustlers flogging it at every opportunity to draw in tourists and make a quick buck? same as Dracula shit in Transylvania. Very tacky but to borrow a phrase tragedy plus distance equals a big bronze statue of the Bewitched lady in the middle of your town

[centrally-unplanned]

Yeah I think it is generally "fine" because it was a by-witches-for-witches thing; practicing pagans came in and built a bit of an alt culture meeting spot there, there was no government action around it until after the fact. And this makes sense with their streak of being countercultural goths about it all; the death thing was part of the appeal.

Granted there is some insensitivity on the part of the witches themselves - the people killed at Salem were not witches or pagan! That was very much "the point". But they were working with the cultural legacy they had, as weird outsiders beggars couldn't be choosers.

And like Mongolia has giant statues of the avatar of war crimes Ghengis Khan all over the place - at a certain point events from long ago just become history, able to be repurposed for the modern era as needed. The needs of the living outweigh the needs of the distant dead.

[tanadrin]

yeah especially when those events don’t implicate any still extant identity categories or raise any still painful sore spots—accusing someone of being a witch doesn’t get them hanged in New England anymore and the 17th century conception of a witch isn’t much like the modern one. Contrast how native identities are still understood to be continuous with those of the past centuries and there are very recent and ongoing grievances by native americans against the US government so treating a contemporaneous 17th century massacre of native Americans the same way would be a lot more likely to upset people.

[tanadrin]

hmm. salem really does only have one tourism gimmick, doesn’t it?

[phaeton-flier]

Always seemed kinda fucked to me; like a lot of people were tortured there. Why are people treating it like Wiccan Disney World.

[tanadrin]

my understanding is that it started as a 70s counterculture thing and then it just became like. the small town hustlers flogging it at every opportunity to draw in tourists and make a quick buck? same as Dracula shit in Transylvania. Very tacky but to borrow a phrase tragedy plus distance equals a big bronze statue of the Bewitched lady in the middle of your town

[tanadrin]

hmm. salem really does only have one tourism gimmick, doesn’t it?

[zeugmalitotes]

I know, right? Can you believe all the whaling they did outta that port?

[tanadrin]

need to find a shop where I can get a formal pegleg to wear to this wedding

[tanadrin]

hmm. salem really does only have one tourism gimmick, doesn’t it?

[tanadrin]

ok this wasn’t actually an attempt to cunningham’s law you guys into recommendations but im not mad that it kinda ended up that way

[tanadrin]

hmm. salem really does only have one tourism gimmick, doesn’t it?

[tanadrin]

I feel like I like. Don’t really get the whole Studentverein thing in Germany. I mean on a sociological level, young human males left to their own devices will carcinate to “homosocial bonding group that involves strong ceremonial markers of identity, drinking, and violence designed to prove how tough you are.” That’s just a given. But the overtones in the German context feel alien to me.

I *think* because there’s no clear Anglo reference point for the political moment they emerged from? Like the major cultural current they seem to have arisen out of is 19th century liberal nationalism; in Germany the 19th century liberal nationalist current was (AIUI) substantially dominated by a middle class that was culturally quite conservative, but which by virtue of wanting a modern unified state as a vehicle for German identity was nonetheless quite radical in comparison to the truly reactionary post-1815 status quo. And Germans gonna Germ, so the expression of this was these very organized serious clubs where they would do drinking and swordfighting and gymnastics about it, and also about The Catholicism/Protestantism (and also often be real weird about Jews; but there are and were also Jewish Studentenverbindung, so I don’t wanna claim they were intrinsically defining themselves against Jewishness, even as antisemitism was a powerful current in German society). And when Germany was unified and 19th century liberalism lost its identity and had to become something new they ended up (while still being very nationalistic) these still very intense political societies but ones where a sort of conservative-liberal aesthetic still dominated.

There just isn’t a good analogue for this in American history—a lot of the German liberal 48ers who came to the US ended up staunch supporters of the Union in the civil war, but as much as I want to make the joke “imagine if your frat started as an abolitionist athletic club that did swordfighting and was super Catholic/high church Lutheran,” the map isn’t quite accurate. The 48ers in the US are a sample biased toward the most radical side of the mid-19th century liberal spectrum, and one that had to remap itself to an American political context.

There were German-style Turnvereine in the US, started by the 48ers, but they feel to me subtly different in political flavor from their German cousins—placed in an American political context the nearest analogue to conservative liberalism is the Northern middle class abolitionist who is consonant in a number of respects but isn’t really in a position where they can be quite so ethnically chauvanist. Many Stidentenvereine in Germany didn’t allow non-Germans to join, and I feel like while they might look down their noses at Italians or whatever, the WASPS of the late 19th century simply could not function if they refused to share social clubs with people of French or Dutch background or whatever. The US was just as a baseline a more ethnically diverse place, even in subcultures that were overwhelmingly white Western European. And though it’s the comparison I sort of reached for above, even the Civil War wasn’t a struggle for the realization of a nation that felt itself held in potentia by reactionary forces—it was a struggle for one part of a nation (one which didn’t have to defend its nationhood) to break away, and not for very strong *nationalistic* reasons.

There was also a militaristic element to the Studentenverbindung from the beginning: you did swordfighting and sport so you would be Ready To Fight For The Fatherland, many had caps or whole uniforms based on 19th century military uniforms, and many groups had “Corps” in their name. And that has no parallel in either the earlier academic or later social fraternities of the US.

By the early 20th century the relative political position of the conservative-liberal-nationalist had shifted in Germany and there was something distinctly conservative politically in their makeup, just bc “conservative” relative to the status quo now quite incontrovertibly included “part of and in favor of a united Germany.” The Second Industrial Revolution has happened, and in Germany (as in the U.S. and many other places!) the previous broad liberal coalition had broken up into both more progressive and socially conscious wings and more conservative ones.

As you might imagine from the military aesthetic, Studentenverbindung were enthusiastic to fight for Germany in WW1, and in the postwar years there was quite a natural path from “member of a ‘liberal’ student club in 1912” to “member of a proto-fascist Freikorps outfit in the 1920s.” This isn’t so strange when you consider that 19th century liberalism just did not mean what liberalism means now (even for those most cynical about it!), both for reasons of semantic drift and just how political coalitions were structured.

I don’t want to give the impression that conservative elements of this culture yielded entirely to fascism. When Studentenverbindungen were told to join the Nazi umbrella organization or close up shop, many did. Many members also were part of the same “yeah we’re conservative nationalists who want to Make Germany Great Again, but that’s why this dipshit Hitler has to go” tendency as Claus von Stauffenberg. A lot of those guys were nobility though and another way in which the Studentenverbindungen kind of fail to map to Anglo cultural categories as I understand them is that AFAICT they’re not aristocrat-coded: the modal member was the son of a middle class professional, bourgeoise and not adlig. These weren’t clubs for toffs (though they certainly were not unwelcome).

The Studentenverbindung culture, always strongest in particular regions like the Rhineland, has declined precipitously since its 19th century/early 20th century peak. And being quite conservative my sense is that most of these organizations (like frats in the US) tend to fall somewhere on the cringe-to-alarming spectrum. But they are also pretty identifiably young mens’ clubs Of A Type, complete with extensive alumni networks that can all by themselves be a reason to entice people to join. And the slicing each other in the face with swords thing is often optional now, if they do it at all. So maybe in their modern form they’re less unusual as an institution. But in an age of increasing cultural fusion in Europe they do feel like a rare really particular national institution.

[weiszklee]

I think the one aspect that you're missing (well, you're hinting at it, so maybe not) is that these used to be and in some ways continue to be (mostly bourgeois) cadre factories. Universities in general are kinda that, of course, but the Studentenverbindungen even moreso. The "Schmiss" used to mark you as a member of the elite, now it marks you as a weirdo, but a weirdo with elite ambitions. Their uneasy relation to the Nazis reflects the uneasy relation of established Weimar elites to the Nazis as an upstart mass movement, some joining, some resisting.

Nowadays, some Verbindungen (especially those clinging to the name "Burschenschaft") are very much far right, but most are just ... trying to keep their little elite clique going. They're known for drinking and hazing rituals, yes, but they also tend to have strict requirements for keeping your marks up at uni. And yeah going through all that does reward you with comradery connections to current elites. Some social conservatism in the German elites is probably downstream of the overprevalence of these boys club alumni with their focus on hierarchy, tradition, nationalism.

From what I understand, there are some less-open fraternities in the US which are kinda similar, no? Like, the college secret society trope in American films is not entirely a fabrication, just an embelishment of the reality right? I think in the UK, private and charter schools kinda play this same role. But both US and UK are much stronger on Ivy Leages and legacy admissions, so there's less of a need to organize the social reproduction of conservative elites informally when the formal institutions already fulfill that function.

[tanadrin]

All v good points to add. My sense is the “join a fraternity to get ahead via the alumni network” is definitely a thing that is cross cultural for young men’s societies. Maybe the factory for elite cadres thing too but the German system is way older and way more overtly politicized in a way I think that makes the factory thing much more explicit and much more central to German political history.

[tanadrin]

mama, just killed a man. put a gun against his head. pulled the trigger now im just having fun, on the stage in my heels. its where i belong in fear for my life from the long arm of the law. didn’t mean to make you cry down at the pink pony club. hang man’s a coming and im gonna keep on dancin, mama, ooooh. on a wanted man.