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.
1
1That'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.)
1More seriously: you might be unrealistic as an example of the type of user who can readily be pulled into a delusional spiral?
1A 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]
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.