All of Rana Dexsin's Comments + Replies

I don't fully agree with gears, but I think it's worth thinking about. If you're talking about “proportion of people who sincerely think that way”, and if we're in the context of outreach, I doubt that matters as much as “proportion of people who will see someone else point at you and make ‘eww another AI slop spewer’ noises, then decide out of self-preservation that they'd better not say anything positive about you or reveal that they've changed their mind about anything because of you”. Also, “creatives who feel threatened by role displacement or think g... (read more)

Table 2's caption is confusing to read. I think this is because in most of what people write about around here, cross-context fusions are positively valenced by default, and “in the context of” doesn't quite capture the scenario. Something like “misapplying the mindset of one House while working for another” (emphasis on changes) would be much clearer.

I actually think that last one just sounds straightforwardly (hah) right? Note shapes express subdivisions of duration that correspond to common rhythmic structures of music, so if jazz music often uses an uneven subdivision at one level but follows the broad structure otherwise, then skewing the meaning of that level in the note shapes is bending the map toward the logical shape of the territory.

I agree, and if the author also agrees with this or something like it, I think the post would be easier to read if something like that were described in the preface.

The “???” in the row below “Not-so-local modification process” for the corporation case should perhaps be something like “Culture and process”?

Small but repeated error: you mean “Ginkgo Bioworks”, right?

I don't think it's not describable, only that such a description being received by someone whose initial mental state is on “thinking about wanting to get better at switching away from thinking” won't (by default) play the role of effective advice, because for that to work, it needs to be empowered by the recipient processing the message using a version of what it's trying to describe. If you already have the pattern for that, then seeing that part described may act as a signal to flatten the chain, as it were; if you don't, then advice in the usual sense ... (read more)

If I may jump in a bit: I'm not sure ‘advice’ can actually hit the right spot here, for “getting out of the car”-style reasons—in this case, something like “trying to look up ‘how to put down the instruction manual and start operating the machine’ in the instruction manual”. That is, if “receiving advice” is a “thinking”-type activity in mental state, the framing obliterates the message in transit. So in some ways the best available answer would be something like “stop waiting for an answer to that question”, but even that is inherently corruptible once pu... (read more)

Dear people writing in the TeX-based math notation here who want to include full-word variables: putting the word in raw leads to subtly bad formatting. If you just write “cake”, this gets typeset as though it were c times a times k times e, as in this example which admittedly doesn't actually show how awkward it can get depending on the scale: $cak{e}_{birthday}$. It's more coherent if you put a \mathrm{} around each word to typeset it as a single upright word, like so: ${cake}_{birthday}$.

Assuming this is the important distinction, I like something like “isolated”/“integrated” better than either of those.

If you have 3 different hypotheses, it's much more natural to keep generating more hypotheses, and to pivot around in a multiple dimensional space of possibility.

The way I imagine this playing out—though I'm not sure how literal this is—is that three hypotheses plus the starting state generate a three-dimensional vector basis when they're in general position. A corollary would be that you want neither all three nor two alongside the starting state to be collinear.

In case anyone's curious, using \124 in a Lua string literal is a decimal escape for “|” (VERTICAL BAR), which looks to be used as a control sequence introducer here. I assume the subsequent character “c” represents a “color” command followed by a 32-bit color: FFD000 is 24-bit RGB for an orange color that looks like the text depicted on that page, with the preceding FF probably meaning full opacity.

For cross-reference purposes for you and/or future readers, it looks like Erik Søe Sørensen made a similar comment (which I hadn't previously seen) on the post “SolidGoldMagikarp III: Glitch token archaeology” a few years ago.

Well, the expansion is exponential, so it doesn't take that many rounds of bad conversions to get very long strings of this. Any kind of editing or transport process that might be applied multiple times and has mismatched input and output encodings could be the cause; I vaguely remember multiple rounds of “edit this thing I just posted” doing something similar in the 1990s when encoding problems were more the norm, but I don't know what the Internet Archive or users' browsers might have been doing in these particular cases.

Incidentally, the big instance in... (read more)

“Ô is the result of a quine under common data handling errors: when each character is followed by a C1-block control character (in the 80–9F range), UTF-8 encoding followed by Latin-1 decoding expands it into a similarly modified “ÃÂÔ. “Ô by itself is a proquine of that sequence. Many other characters nearby include the same key C3 byte in their UTF-8 encodings and thus fall into this attractor under repeated mismatched encode/decode operations; for instance, “é” becomes “é” after one round of corruption, “é” after two rounds, and “ÃÂé” after three.

(Edited for accuracy; I hadn't properly described the role of the interleaved control characters the first time.)

Re part of the quotation from §3.3.1 of the o1 system card beginning with “Apollo found that o1-preview sometimes instrumentally faked alignment during testing”: I noticed that two of the three Assistant quotes in this section were identical in a way that didn't make sense. This appears to be an error in the original system card; §3.3.1 references §8.3 (which shows three separate examples) but quotes from the third while describing the behavior of the second. Separately, that section includes summaries of resultant actions in angle brackets, which are in t... (read more)

This is awkwardly armchair, but… my impression of Eliezer includes him being just so tired, both specifically from having sacrificed his present energy in the past while pushing to rectify the path of AI development (by his own model thereof, of course!) and maybe for broader zeitgeist reasons that are hard for me to describe. As a result, I expect him to have entered into the natural pattern of having a very low threshold for handing out blocks on Twitter, both because he's beset by a large amount of sneering and crankage in his particular position and be... (read more)

The identifiable code chunks look more specifically like they're meant for ComputerCraft, which is a Minecraft mod that provides Lua-programmable in-game computers. Your link corroborates this: it's within the ComputerCraft repository itself, underneath an asset path that provides files for in-game floppy disks containing Lua programs that players can discover as dungeon loot; GravityScore is a contributor with one associated loot disk, which claims to be an improved Lua code editor. The quoted chunk is slightly different, as the “availableThemes” paragrap... (read more)

I was pretty sad about the ongoing distortion of “I checked” in what's meant to be an epistemics-oriented community. I think the actual meanings are potentially really valuable, but without some way of avoiding them getting eaten, they become a hazard.

My first thought is to put a barrier in the way, but I don't know if that plays well with the reactions system being for lower-overhead responses, and it might also give people unproductive bad feelings unless sold the right way.

Cars and planes and knives and various chemicals can be easily goaded to break the law by the user. No one has yet released a car that only ever follows all applicable laws no matter what the driver does.

Without taking a position on the copyright problem as a whole, there's an important distinction here around how straightforward the user's control is. A typical knife is operated in a way where deliberate, illegal knife-related actions can reasonably be seen as a direct extension of the user's intent (and accidental ones an extension of the user's negli... (read more)

Detached from a comment on Zvi's AI #80 because it's a hazy tangent: the idea of steering an AI early and deeply using synthetic data reminds me distinctly of the idea of steering a human early and deeply using culture-reinforcing mythology. Or, nowadays, children's television, I suppose.

Followup:

How so much artistry had been infused into the creation of Hogwarts was something that still awed Draco every time he thought about it. There must have been some way to do it all at once, no one could have detailed so much piece by piece, the castle changed and every new piece was like that.

Years later, Midjourney happened.

My favorite active use of those is differentially. Wiggling my nose can inspire visceral surprise.

Temporarily taking the post's theory as given, then speculating: managers a few levels above the bottom won't feel much dominance increase from hires at the bottom if they're too organizationally distant for it to register, I'd think; the feeling boost from Nth-level reports would drop sharply with increasing N due to less personal contact. They would then seek to manipulate their set of direct reports. Some would see internal underlings as a threat, want to get them out of the way, and not necessarily have another insider suitable to displace them with. S... (read more)

… at least in the default light theme. (This is arguably a secondary reason not to overuse images.)

Observation of context drift: I was rereading some of HPMOR just now, and Harry's complaint of “The person who made this probably didn't speak Japanese and I don't speak any Hebrew, so it's not using their knowledge, and it's not using my knowledge”, regarding a magic item in chapter 6, hits… differently in the presence of the current generation of language models.

The Review Bot would be much less annoying if it weren't creating a continual stream of effective false positives on the “new comments on post X” indicators, which are currently the main way I keep up with new comments. I briefly looked for a way of suppressing these via its profile page and via the Site Settings screen but didn't see anything.

I haven't worked in an organization that uses microservices extensively, but what I hear from people who use them goes far beyond visibility constraints. As an example, allowing groups to perform deployment cycles without synchronizing seems to be a motivation that's harder to solve by having independently updated parts of a build-level monolith—not impossible, because you could set up to propagate full rebuilds somehow and so forth, but more awkward. Either way, as you probably know, “in theory, people could just … but” is a primary motivator behind all k... (read more)

And microservices are mostly a solution to institutional/management problems, not technical ones.

So this is interesting in context, because management and coordination problems are problems! But they're problems where the distinction between “people think this is a good idea” and “this is actually a good idea” is more bidirectionally porous than the kinds of problems that have more clearly objective solutions. In fact the whole deal with “Worse is Better” is substantially based on observing that if people gravitate toward something, that tends to change... (read more)

Publishing “that ship has sailed” earlier than others actively drives the ship. I notice that this feels terrible, but I don't know where sensible lines are to draw in situations where there's no existing institution that can deliver a more coordinated stop/go signal for the ship. I relatedly notice that allowing speed to make things unstoppable means any beneficial decision-affecting processes that can't be or haven't been adapted to much lower latencies lose all their results to a never-ending stream of irrelevance timeouts. I have no idea what to do her... (read more)

Ted Chiang's Chrystal Nights

Minor corrections: “Crystal Nights” does not have an H in the first word and is by Greg Egan. (The linked copy is on his own website, in fact, which also includes a number of his other works.)

So in the original text, you meant “openness minus conscientiousness”? That was not clear to me at all; a hyphen-minus looks much more like a hyphen in that position. A true minus sign (−) would have been noticeable to me; using the entire word would have been even more obvious.

Could restaurants become better aligned if instead of food we paid them for time?

The “anti-café” concept is like this. I've never been to one myself, but I've seen descriptions on the Web of a few of them existing. They don't provide anything like restaurant-style service that I've heard; instead, there are often cheap or free snacks along the lines of what a office break room might carry, along with other amenities, and you pay for the amount of time you spend there.

How are those staying alive in the first place? I had previously used Nitter for keeping up with some of Eliezer's posts without being logged in, but my understanding was that the workaround they were using to obtain the necessary API keys was closed off several months ago, and indeed the instances I used stopped working for that purpose. Have the linked instances found some alternative method?

Have you met non-serious people who long to be serious?

I am one of those people—modulo some possible definitional skew, of course, especially around to what degree someone who wishes to be different from how they are can be considered to wish for it coherently.

I know that right now I am not acting seriously almost at all, and I feel a strong dislike of this condition. Most of my consciously held desires are oriented in the direction of seriousness. A great deal of me longs to be serious in wholeness, but that desire is also being opposed by a combinatio... (read more)

That description is distinctly reminiscent of the rise of containerization in software.

Given the presence of mood fluctuations and other noise, repeatedly being triggered to re-evaluate a decision on whether or not to take a one-shot action when not much has relevantly changed in the meantime seems subject to a temporal unilateralist's curse: if you at time 1000 choose to do the action even if you at times 0–999 didn't choose it and you at times 1001–1999 wouldn't have chosen it, it still happens. The most well-known example that comes to mind of this being bad is addiction and “falling off the wagon”, but it seems like it generalizes.

Yes, but I've seen “syncopathy” rarely as a coinage meaning something along the lines of “convergence of emotion”, which is just-about within plausibility range.

Is the repeated use of “syncopathy” here a misspelling of “sycophancy”, or does it have a domain-specific meaning I'm not familiar with?

In less serious (but not fully unserious) citation of that particular site, it also contains an earlier depiction of literally pulling up ladders (as part of a comic based on treating LOTR as though it were a D&D campaign) that shows off what can sometimes result: a disruptive shock from the ones stuck on the lower side, in this case via a leap in technology level.

feature proposal: when someone is rate limited, they can still write comments. their comments are auto-delayed until the next time they'd be unratelimited. they can queue up to k comments before it behaves the same as it does now. I suggest k be 1. I expect this would reduce the emotional banneyness-feeling by around 10%.

If (as I suspect is the case) one of the in-practice purposes or benefits of a limit is to make it harder for an escalation spiral to continue via comments written in a heated emotional state, delaying the reading destroys that effect c... (read more)

![[Pasted image 20240404223131.png]]

Oops? Would love to see the actual image here.

Not LLMs yet, but McDonalds is rolling out automated order kiosks,

That Delish article is from 2018! (And tangentially, I've been using those as my preferred way to order things at McDonald's for a long while now, mostly because I find digital input far crisper than human contact through voice in a noisy environment.)

The subsequent “Ingrid Jacques” link goes to a separate tweet that links to the Delish article, but it's not the Ingrid Jacques tweet, which itself is from 2024. I think the “tyson brody” tweet it links to instead might be a reply to the ... (read more)

My (optimistic?) expectation is that it ends up (long run) a bit like baking.

Home/local music performance also transitioned into more of a niche already with the rise of recorded music as a default compared to live performance being the only option, didn't it?

At that scale its more "celebrity-following", but that is also something the AI would not have - I don't know how big a deal that is.

While I doubt it will be the same thing for a transformer-era generative model due to the balance of workflow and results (and the resultant social linkages) bein... (read more)

I don't see it as an elephant overall, but I can see how you could push it to be the head of one: the head is facing to the right, the rightmost curve outlines the trunk, the upper left part of the main ‘object’ is an ear, and some of the vertical white shapes in the bottom left quadrant can be interpreted as tusks.

While I appreciate the attempt to bring in additional viewpoints, the “Sign-in Required” is currently an obstacle.

I claim that very few people actually understand what they are using and what it effects it has on their mind.

How would you compare your generative-AI focus to the “toddlers being given iPads” transition, which seems to have already happened?

This SMBC from a few years ago including an “entropic libertarian” probably isn't pointing at what people call “e/acc”… right? My immediate impression is that it rhymes though. I'm not sure how to feel about that.

The first sentence here is very confusing and I think inverts a comparison—I think you mean “would make the world enough worse off”.

The first somewhat contrary thing that comes to mind here is whether visible spending that looks like a status grab or is class-dissonant would also impact your social capital in terms of being able to source (loaned or gifted) money from your networks in case of a crunch or shock. If your friends will feel “well I sure would've liked to have X, but I was the ‘responsible’ one and you weren't, so now I'm not going to put money in when you're down” and that's what you rely on as a safety net, then maybe you do need to pay attention to that kind of self-poli... (read more)