You see either something special, or nothing special.
Rana Dexsin
Karma: 1,040
That’s also the title of a Paul Graham essay!
I think I semi-agree with your perception, but I did have a recent experience to the contrary: when I did a throwaway post about suddenly noticing a distinction in the reaction UI, I found it very odd that some people marked the central bit “Insightful”. Like, maybe it’s useful to them that I pointed it out, but it’s a piece of UI that was (presumably) specifically designed that way by someone already! There’s no new synthesis going on or anything; it’s not insightful. (Or maybe people wanted to use it as a test of the UI element, but then why not the paperclip or the saw-that eyes?)
Everything2 did this with votes. I think the “votes per day” limit used to be more progressive by user level, but it’s possible that’s me misremembering; looking at it now, it seems like they have a flat 50/day once you reach the level where voting is unlocked at all. Here’s what looks like their current voting/experience system doc. (Note that E2 has been kind of unstable for me, so if you get a server error, try again after a few minutes.)
One good that might be offered is dominance.
“As material needs are increasingly met, more surplus goes into positional goods” seems like a common pattern indeed. Note “positional” and not purely “luxury”. I consider both prestige-status and dominance-status to be associated with position here. Even the former, while it could lead to a competition for loyalty that’s more like a gift economy and cooperates with the underclass, could also lead to elites trying to outcompete each other in pure consumption such that “max comfort” stops being a limit for how much they want compared to the underclass. Indeed I vaguely recall hearing that such a dynamic, where status is associated with how much you can visibly spend, already holds among some elite classes in the current day.
My thoughts are shaped by the cultural waves of the last few decades in the USA, so they lean toward imagining the moral fig leaf as something like “make sure the people benefiting from our stuff aren’t enemies/saboteurs/Problematic” and a gradual expansion of what counts as the excluded people that involves an increasing amount of political and psychological boxing-in of the remainder. That all flows nicely with the “find ways to get potential rebels to turn each other in” sort of approach too. Of course that’s one of many ways that a dominance-motivated persistent class asymmetry could play out.
If you’re familiar with the story “The Metamorphosis of Prime Intellect”, a shorter story that the author wrote in the same universe, “A Casino Odyssey in Cyberspace”, depicts characters with some related tendencies playing out against a backdrop of wild surplus in a way you may find stimulating to the imagination. (Edited to add: both stories are also heavy on sex and violence and such, so be cautious if you find such things disturbing.) Also, Kurt Vonnegut’s novel Player Piano about a world without need of much human labor doesn’t show the harsh version I have in mind, but the way ‘sabotage’ is treated evokes a subtler form of repression (maybe not worth reading all of just for this though).
The “sounds like a bunch of it has definitely been written by ChatGPT” from Garrett’s initial response and the “genuine specific critiques that I can take to improve my writing” you are asking about do not go well together. This has inspired me in the background to try to write something more detailed about why (which may or may not yield anything), but a shorter and more actionable take for you is that if you affirmed unambiguously that you have personally followed the Policy for LLM Writing on LessWrong for this post, it might go a long way toward convincing people to give it more of a shot. (Or if it turns out it’s too late for that for this post, it might convince people not to bypass later posts for slop-reputation reasons.) A key paragraph is quoted below, emphasis mine:
A rough guideline is that if you are using AI for writing assistance, you should spend a minimum of 1 minute per 50 words (enough to read the content several times and perform significant edits), you should not include any information that you can’t verify, haven’t verified, or don’t understand, and you should not use the stereotypical writing style of an AI assistant.
Edited to add: if you find this also unconvincing due to being the “kind of meta discussion that infuriates [you]”, I have an alternate and more directly political variant to try, but I’d rather not go for that one first.
In the “Race” ending of “AI 2027”, the actual destruction of humanity only occurs in 2030, though?
To be clear, that last part isn’t just from here. Occasionally getting a Substack link that insists on surrounding the post with the rest of the feed/home interface and being unable to find how to get the same post without that has frustrated me before, and I think Reddit’s version of something similar has frustrated me, and maybe one or two other sites doing similar things… in general it seems like it’s a really easy source of “75% of it works, have fun whacking your metaphorical limbs on the edges”. Which is sort of still indirectly relevant as an influence on responses to the pattern even though some of it is not in your control. 🙂
I am curious what alternatives you’re looking at now!
I just wound up in some additional super-confusing behavior. My best recollection is that “Angry Atoms” showed up in the feed display, I clicked it, it opened in the overlay view, and then when I tried to navigate from there using its own link to “How An Algorithm Feels From Inside”, seemingly nothing happened, except that I got a weird hunch and clicked the close-overlay left-arrow button, revealing that the front page had been replaced with the second post? And now there was no way to get back to the first one, and the browser Back button did… initially nothing, and then… maybe something? I wasn’t writing it all down while it happened and now I can’t easily reproduce it because the recommendation contents have naturally changed unpredictably.
I seem to be gradually picking up lingering emotional-experiential scrapes around the “yo I heard you liked navigation so I put navigation in your navigation so you can navigate while you navigate” school of site design.
I’ve hovered over them to see the applicable text or lack thereof before, yes, and I was aware that both types of reaction were possible. Overall I don’t have a clear enough memory to say why I didn’t pick up on this connection sooner, but my off-the-cuff guess would be that seeing both inline-portion and whole-comment reactions on the same comment is rare, which would mean there wasn’t a clear juxtaposition to show that it’s only present sometimes, and my visual processing would likely have discarded the cartouche as a decorative separator.
Oh! Is that what the dotted cartouche around a reaction means? That it’s applied to an inline portion of text instead of to the whole post? That wasn’t obvious to me before.
I can confirm that this was true when I tried something very similar with ChatGPT several months ago, and that my recent experiments with image generation in that context involving specific geometric constructions have also generally gone badly despite multiple iterations of prompt tuning (both manually and in separate text conversations with the bot).
The case I’m most curious about is actually the hybrid case: if you want to embed a specific geometry inside a larger image in some way, where the context of the larger image is ‘softer’, much more amenable to the image model and not itself amenable to traditional-code-based generation, what’s the best approach to use?
mruwnik’s answer is correct on the surface level but doesn’t connect it up to the specific use. The specific word “retvrn” references particular Internet waves of rejection-of-modernity sentiment (possibly especially in reactionary politics, though how true this is seems to be contested) that use Roman iconography to represent their imagined idea of traditional Western society. People with such a sentiment often advocate for the reinstatement of stricter social mores that they believe produced better societies in the past—thus the use in the post, where the hypothetical person believes that “our problems started when the fence was removed”.
I approve that someone is trying to write these!
That said: is this actually intended for general audiences who are not already positively disposed to the local memesphere? I feel like some aspects assume otherwise, and I would expect “remember the litany” in particular to give “I am a foreign priest telling you what to do” vibes. Or, I suppose, is this targeted at people who were otherwise about to post their pseudo-breakthroughs on LW? In that case it would make more sense, in a “since you were about to come here, this is important if you want to be accepted here” sort of way.
Basically, by maintaining strict openness to the evidence, it takes away any motivation and justification you might have had for being unfair to me.
Am I missing some context here? Let’s look at this hypothetical conversation, which seems pretty darn plausible to me:
“Wait, you thought that was a rule, not a request to be less messy?”
“What the hell kind of nitpick is that? Stop arguing stupid semantics! Since when should I even have to ask Your Highness for basic decency?”
“Do you actually think that messiness is correlated with thievery, even after conditioning on honesty?”
“What are you even talking about now? Some math shit? That’s what’s really important to you, huh, rather than being a good person who knows when to clean up? Grow up or go live on the street. And don’t take any more of my cookies.”This is admittedly a strained use of the specific quotations, but I think the directional picture should be clear. Extrapolate to your (least) favorite contested-valence social markers to taste.
I think the first thing to note here is that the bar isn’t “Is there zero chance of secure response failing to exonerate me?” but “Is the secure response less likely to exonerate me?”.
Surely this depends on your surroundings?
What is the “secure response”? One where you try outwardly to retain a certain kind of dignity? When you don’t actually have the status security in local social reality, you can’t necessarily get away with that. In the inconvenient world that I’m currently imagining from which I generated the above dialogue, screwing around with things like ‘evidence’, or even acting calm (thus implying that the rules (which every non-evil person can infer from their heart, right?) are not a threat to you or that you think you’re above them—see also, some uses of “god-fearing” as a prerequisite for “acceptable” in religious contexts), is breaking the social script. It’s presumed to be trying to confuse matters or go around the problem (the problem that they have with you; think “skipping out on your court date” as an analogy in a less emotional context), and it gets you the most guaranteed negative judgment because you didn’t even meta-respect what was going on. Your mainline options under that kind of regime can be more like “use a false apology to submit, after which the entire social reality is that You Did It but at least you showed some respect” or “make a counterplay by acting openly defensive, which acts kind of like a double-or-nothing coin flip depending on whether the audience both believes you and believes enough others will believe you to coordinate against the accuser”. (In this context, the audience may culturally share the felt-sense of “don’t try to get all fancy on us” even if their beliefs about your specific guilt may vary.) Naturally, as Kaj_Sotala described above, refusing to say anything at all can be interpreted as a tacit admission, so that doesn’t help either.
“Agitated? Listen to this guy. He’s fucking agitated!” “Well, good. That’s good. You stay that way.”
Maybe you could say that the type of emotional and motivational backing for what “acting defensive” means in that context is substantially different from the type of “defensive insecurity” being described above, but at least when I imagine the experiences and expressions they come out close to indistinguishable. I can also imagine trying to retain a feeling of security on the inside (likely at great mental cost) while play-acting the defensiveness in the above context, but that seems like a very noncentral case.
Now for extra fun, imagine this being simultaneously watched by people whose main experience is in a different cultural regime where (perhaps due to the above type of control being uncommon and frowned upon) they can more reasonably justify defensiveness as evidence in favor of guilt, except you don’t have separate private channels to those people and to the people above—possibly because you don’t even know which subset of people is which—and everything you do is being interpreted by both.
Maybe many rationalists interact with chatbots more as a letter correspondent, but if so this is highly unusual (and not true for me).
I do! Not full ‘letters’, but definitely a paragraph at a time is normal for me. (I also naturally do it in other conversational media, sometimes unfortunately; this is a habit I’ve tried to break among genpop, with partial success. In another life I have sometimes been known for accidentally smothering people’s Twitch chat…) I would guess that my entire communicative and thinking style was heavily influenced by written culture first and oral culture a distant second, so I talk like a book unless I’m trying not to.
Spot check on a recent conversation I had with ChatGPT-5 in which I was trying to solidify my understanding of a few points in mathematics: my message lengths from that conversation, rounded to multiples of 10 and sorted, were (70 310 340 370 400 480 770 820). The lowest one corresponds to my initial question, and you can see the others all fall into a sort of paragraph-y range, with no really short ones at all.
I can easily believe that this would be unusual overall, though I don’t know if it would have occurred to me to think of that if you hadn’t pointed it out. I don’t know how unusual it is among people selected for being active LW users.
That’s a good point, but what that makes me wonder in turn is whether the AI having a speed advantage in writing might be displacing that for this particular type of interaction. Basically: “I would’ve needed to write a long manifesto to truly demonstrate the depth and brilliance of my ideas, and now I can get a really classy-looking one with equations and everything for 10% as much effort!”
(I’m tempted to go look up transcripts and check, but my cognitohazard shielding is on the low side for that right now. I wouldn’t mind if someone else did it.)
More seriously: you might be unrealistic as an example of the type of user who can readily be pulled into a delusional spiral?
A relevant-feeling quotation from off at an angle, from Eric Raymond’s version of the Jargon File, section “Personality Characteristics”:
Although high general intelligence is common among hackers, it is not the sine qua non one might expect. Another trait is probably even more important: the ability to mentally absorb, retain, and reference large amounts of ‘meaningless’ detail, trusting to later experience to give it context and meaning. A person of merely average analytical intelligence who has this trait can become an effective hacker, but a creative genius who lacks it will swiftly find himself outdistanced by people who routinely upload the contents of thick reference manuals into their brains. [During the production of the first book version of this document, for example, I learned most of the rather complex typesetting language TeX over about four working days, mainly by inhaling Knuth’s 477-page manual. My editor’s flabbergasted reaction to this genuinely surprised me, because years of associating with hackers have conditioned me to consider such performances routine and to be expected. —ESR]
Using the space bar to scroll in overlay puts the next line of text underneath the slightly-transparent header so that I have to adjust backward to see it, often after reading a few words from the wrong spot and getting confused. (Edited to add: okay, obviously not so often after I recognized the pattern, but I’m prospectively amused that if/when this gets changed I’ll have to adjust the habit a second time.)
(Is this what frontend Web development is just like? Should I pawn off as much of it as possible on AI in the future? Will that work, or will it drive the AI mad? Or will I be insufficiently inured when I have to dive in and fix a bug that it can’t handle, and then I’ll lose a day gibbering, thus canceling out the time and sanity saved?)
Interesting! Maybe there’s an experiential crux in there (so to speak)? My reader experience of this is that the first-person inner monologue is indeed very different from mine, but I perceive that as increasing the immersion and helping frame the story. To the extent that there’s a group of humans it saliently bears little resemblance to, I might think of that group as something like “humans who are psychologically ‘healthy’ in a certain way which varies across a wide spectrum, where social spheres with concentrated power may disproportionately attract people who are low on that spectrum”. I’m deliberately putting the main adjective in scare quotes there because in my fuzzy mental model, there isn’t really a clear delimiter between treating that trait cluster as a health indicator and treating it as intersubjective values dissonance; it feels consonant with but not directly targeting dark triad traits. But also, I’m not sure if you’re referring to the same thing I am or if it’s some other feature of the first-person description that bothers you.
FWIW, culturally speaking, I’ve been socially adjacent to Bay Area / SV-startup / “mainstream big tech” culture via other people, but not really been immersed in it—I’ve splashed around in the shallow part of that pool long ago, but for “try not to build the Torment Nexus” reasons (plus other unrelated stuff) I historically bounced away a lot as well and wound up in a sort of limbo. So that colors my impression quite a bit.