domain:questioner.substack.com
4bpp
Now I am become a Helpful, Honest and Harmless Assistant, the destroyer of jobs
15hr ago
"Do not judge" (as stated)/"judge only deniably, or based on a narrow set of acceptable criteria (socks with sandals etc.)" (as implemented) is an American cultural value. You could argue that it serves some purpose on a societal level, in a Chestertonian way, but many societies without it mostly work fine, which puts an upper bound on how important it can be.
To maximise personal advantage, it is rational to always update/"judge" on everything that you can extract a meaningful evidential signal from, which surely includes all of your examples. It seems like a pretty complex question which criteria should be kept to maximise the elusive societal advantage (i.e. what set of judgement taboos maximises social welfare?) - the most obvious advantage of any such taboos is that they facilitate coexistence between different groups with divergent aesthetic values, and thereby also encourage such groups to form to begin with, enabling distributed experimentation on value systems. For example, if it turns out pro-tattoo values actually carry some unexpected advantage (aliens invade and kill everyone without?), the societies which did not suppress pro-tattoo aesthetics because they had a taboo against judging based on tattoos would come out ahead.
thejdizzler
15hr ago
Good observation. I also agree that the hustle-culture memes aren't reflective of how people's efforts can actually be allocated. A common failure mode I see in myself is over-scheduling things in my down-time and not doing any of them and gaming/scrolling instead. I really should be resting during that time.
FCfromSSC
Nuclear levels of sour
15hr ago
You really should finish episode 3. I nope'd out maybe 15 minutes into ep 1 the first time I tried to watch it, and then came back a few months later and decided to give it another go. end of ep 3 is where the preflight checklist is complete and takeoff is acheived.
sarker
It isn't happening, and if it is, it's a bad thing
15hr ago
You're right, and I was mistaken about my state too.
The "Attention Economy" is just BRUTAL, b/c it really is an utterly zero-sum game (you can't produce 'more attention' very easily, only reapportion the amount that currently exists), and thus there is strong incentive to try to drag attention out of people even when it is objectively unhealthy.
"Of course I can watch one more episode, Netflix, how thoughtful of you to queue it right up!" (looks up 3 episodes later to see the clock says "1:38 a.m.")
No, fuck off. Give me the app that values my attention approximately as much as I do, and will actively start discouraging me from expending it too much in one place. "Here, you have time for precisely one (1) episode of Tulsa King, then we're cutting you off. I've already set the lights in the room to dim slowly, and your favorite ambient sleep noises are cued up as soon as you get into the bed."
gattsuru
15hr ago
While your opening argument could be construed as just a quibble about the exact coefficient of the power-law distribution that Trace is alluding to...
That's certainly part of the disagreement, especially in terms of ability to successful policy campaigns, but I think there's a deeper disagreement specific to just the relationship between enthusiast focus and mainstream attention (or even attention among other enthusiasts). It's also is a disagreement about what extent :
The answer to ‘why are people suddenly talking about this?’, in that case, comes down to seven people who looked around, weren't professionals in it, didn't have any background experience, anything like that, but looked and were like, this doesn't make sense, this is an issue that we would like to handle, this is an issue we would like to raise the salience of
is true, both for the Bully XL question and in general.
From my understanding, the theory here is that the CEBRDD, BullyWatch, and Lawrence Newport "raised salience", and that explains why everyone was talking about it. We can actually examine this! Newport first posted on Twitter on the matter in April 2023; BullyWatchUK only created an account in July 2023. Tracking websites is harder, but CEBRDD's first domain name registry is October 2023, and BullyWatch's website probably started early spring 2023.
Okay, that lines up real nice with MP Hayes pushing for a ban in June 2023, if perhaps a little messy. What's the problem? Well...
What else happened, in June 2023 and the preceding months? I don't have Trace's full list of those seven activists, but either he's including names that can not accurately fit as "didn't have any background experience", or he's missing names that were a large part of the drive. Indeed, if you start poking at the history there's actually a lot of salience-raising starting from conventional media in 2022 by orgs unrelated to Newport, and there's a lot of motivation for legislation completely separated from a bunch of technical analysis, and a parallel campaign that started in Ireland.
I'm not saying these CEBRDD guys (and women) didn't matter at all... but even before we get to the question of whether they drove the policy campaign, they clearly couldn't have driven the why is everyone talking about this. To the extent that they did matter or show up after they got into the policy debate, I'm not sure how much it reflects them driving the reporting versus reporters looking for someone available to quote that they'll agree with.
The power-law distribution, now matter what its exponent or dividing line, was not the actual important part driving conversation.
That's a particularly severe case: (oftenly gruesomely) dead kids and women, a very narrow timeline, and a very specific set of proposed Important Unrelated Activists. And can certainly believe there are some matters where this traces the whole path. But it's hard to evaluate them, because for every genuine grassroots operation you'll pretty quickly find several where, on further inspection, it turns out that the 'grassroots' speakers are tots-not-speaking for a large organization they're an employee of which focuses on this topic, or they're very intimately tied to a prominent example of the case, or where the genuine grassroots are just laundering the opinions of the media organization interviewing them.
While his approach is clearly underutilised for areas of discrete policy such as dog control, curriculum changes, or selective regulatory reform, bureaucracies are often immensely complex and not so easily transformed by these sorts of (necessarily top-down) outsider campaigns.
To be fair to Trace, I think he argues that a lot of this particular set of problems is also downstream of the pipeline issues, and is suggesting development of separate programs outside of those tools to provide a sort of outside pressure against that bureaucracy. Demonstrating the bad results of popular policy by contrasting to a good external policy won't solve the whole bureaucracy, but it's a fulcrum to get public attention and undermine the proponents of the bad policy if they don't take it up the new alternative.
To be less fair to Trace, there's a really concrete example of exactly that having happened in the very specific sphere of education... but charter schools are a very awkward fit for all of his recs. And Gramsci's long march worked without presenting much in the way of generally useful things, instead favoring benefits for its own advocacy core.
My wife on a couple different occasions expressed the desire to get a tattoo. Each time I'm like "No. No tattoos. I don't want you to get a tattoo." Naturally she asks me why, and I'm at a loss for words. You just...you don't do that! That's your skin! It's not a piece of paper! Do you want to look like the kind of person who gets tattoos?!
I guess you either grew up in a family with standards* or you didn't.
*My younger brother got a small tattoo, of a line of scripture. Getting a Bible verse tattooed on a discreet part of your body seems like the most innocuous kind of tattoo you could get, but he still hid it from us for years and only admitted it with a lot of sheepishness when he came across a situation where we were bound to see it. This is right and correct.
ThomasdelVasto
Κύριε, ποίησόν με ὄργανον τῆς ἀγάπης σου
15hr ago
I agree with this. I cycled through a lot until I found piano, and singing. I do them because they are beautiful and they open my heart. I don't even necessarily have to have energy or anything I just find myself gravitating towards piano more and more because I genuinely want it.
Stellula
15hr ago
Sure a small segment of the population. The warriors are putting their lives at risk so that the poets can write poetry.
ToaKraka
Dislikes you
15hr ago
There does not appear to be vision-blocking vegetation at this particular intersection.
ToaKraka
Dislikes you
15hr ago
That law was part of the jury charge as well. See p. 16 of the PDF.
Having no interest to get into a pissing context^W contest, I'll only disclose I've contributed to several DL R&D projects of this era.
This is the sort of text I genuinely prefer LLM outputs to, because with them, there are clear patterns of slop to dismiss. Here, I am compelled to wade through it manually. It has the trappings of a sound argument, but amounts to epitemically inept, reductionist, irritated huffing and puffing with an attempt to ride on (irrelevant) credentials and dismiss the body of discourse the author had found beneath his dignity to get familiar with, clearly having deep contempt for people working and publishing in the field (presumably ML researchers don't have degrees in mathematics or CS). Do even you believe you've said anything more substantial than “I don't like LLMs” in the end? A motivated layman definition of intelligence (not even citing Chollet or Hutter? Seriously?), a psychologizing strawman of arguments in favor of LLM intelligence, an infodump on embedding arithmetic (flawed, as already noted), random coquettish sneers and personal history, and arrogant insistence that users are getting "fooled" by LLMs producing the "appearance" of valid outputs, rather than, say, novel functioning programs matching specs (the self-evident utility of LLMs in this niche is completely sidestepped), complete with inane analogies to non-cognitive work or routine one-off tasks like calculation. Then some sloppy musings on current limitations regarding in-context learning and lifelong learning or whatever (believe me, there's a great deal of work in this direction). What was this supposed to achieve?
In 2019, Chollet has published On the Measure of Intelligence, where he has proposed the following definition: “The intelligence of a system is a measure of its skill-acquisition efficiency over a scope of tasks, with respect to priors, experience, and generalization difficulty.” It's not far from yours, because frankly it's intuitive. Starting from this idea and aiming to test fluid thinking specifically, Chollet has also proposed ARC-AGI benchmark, which for the longest time was so impossibly hard for DL systems (and specifically LLMs) that many took that as evidence for the need to do “complete ground-up redesign from first principles” to make any headway. o3 was the first LLM to truly challenge this; Chollet coped by arguing that o3 is doing something beyond DL, some “guided program synthesis” he covets. From what we know, it just autoregressively samples many CoTs in parallel and uses a simple learned function to nominate the best one. As of now, it's clearly going to be saturated within 2 years as is ARC-AGI 2, and we're on ARC-AGI 3, with costs per problem solved plummeting. Neither 1 nor 3 are possible to ace for an orangutan or indeed for a human of below-average intelligence. Similar things are happening to “Humanity's Last Exam”. Let's say it's highly improbable at this point than any “complete ground-up redesign from first principles” will be necessary. Transformer architecture is rather simple and general, making it cheaper to train and inference without deviating from the core idea of “a stack of MLPs + expressive learned mixers” is routine, and virtually all progress is achieved by means of better data – not just “cleaner” or “more”, but procedural data predicting which necessitates learning generally useful mental skills. Self-verification, self-correction, backtracking, iteration, and now tool use, search, soliciting multi-agent assistance (I recommend reading Kimi K2 report, the section 3.1.1, for an small sliver of an idea of what that entails). Assembling necessary cognitive machines in context. This is intelligence, so poorly evidenced in your texts.
In order to align an AI to care about truth and accuracy you first need a means of assessing and encoding truth and it turns out that this is a very difficult problem within the context of LLMs, bordering on mathematically impossible.
We are not in 2013 anymore, nor on LessWrong, to talk of this so abstractly and glibly. "Reptile — legs = snake" just isn't an adequate level of understanding to explain behaviors of LLMs, this fares no better than dismissing hydrology (or neuroscience, for that matter) as mere applied quantum mechanics with marketing buzzwords. Here's an example of a relevant epistemically serious 2025 paper, "The Geometry of Self-Verification in a Task-Specific Reasoning Model":
We apply DeepSeek R1-Zero’s setup with Qwen2.5-3B as our base model (Hyperparams: Appx. A). Our task, CountDown, is a simple testbed frequently used to study recent reasoning models [9, 10, 32, 39 ] – given a set of 3 or 4 operands (e.g., 19, 36, 55, 7) and target number (e.g., 65), the task is to find the right arithmetic combination of the operands to reach the target number (i.e., 55 + 36 - 7 - 19). […] The model is given two rewards: accuracy reward for reaching the correct final answer, and a format reward when it generates its CoT tokens in between “” and “” tokens. […] Once we score each previous-token head using Eq. 8, we incrementally ablate one head at a time until we achieve perfect intervention scores (Section 4.4). Using this approach, we identify as few as three attention heads that can disable model verification. We notate this subset as AVerif. To summarize, we claim that the model has subspace(s) (polytope(s)), SGLUValid , for self-verification. The model’s hidden state enters this subspace when it has verified its solution. In our setting, given the nature of our task, previous-token heads APrev take the hidden-state into this subspace, while for other tasks, different components may be used. This subspace also activates verification-related GLU weights, promoting the likelihood of tokens such as “success” to be predicted (Figure 3). […]For “non-reasoning” models, researchers have studied “truthful” representations before [ 4 ], where steering towards a “truthful” direction has led to improvements in tasks related to factual recall [ 17]. In a similar vein, researchers have shown that the model’s representations can reveal whether they will make errors (e.g., hallucinations) [ 28 ], or when they are unable to recall facts about an entity [ 8 ]. Most recently, concurrent work [37, 41 ] also investigate how models solve reasoning tasks. [ 41 ] find that models know when they have reached a solution, while [ 37 ] decode directions that mediate behaviors such as handling uncertainty or self-corrections. While our work corroborates these findings, we take a deeper dive into how a reasoning model verifies its own reasoning trace. Circuit Analysis. A growing line or work decomposes the forward pass of a neural network as “circuits” [24], or computational graphs. This allows researchers to identify key components and their causal effects for a given forward pass. A common approach to construct computational graphs is to replace model components with dense activations with a sparsely-activating approximation. [ 6] introduces Transcoders to approximate MLP layers, while [ 1 ] further develops Cross-layer Transcoders to handle inter-layer features. [18 ] uses Cross-layer Transcoders to conduct circuit analyses for a wide range of behaviors, such as multi-step reasoning (for factual recall) or addition, and also investigate when a model’s CoT is (un)faithful…
The point of this citation is to drive home that any “first principles” dismissal of LLMs is as ignorant, or indeed more ignorant, than sci-fi speculation of laymen. In short, you suck and you should learn humility to do better to corroborate your very salient claim to authority.
There are good criticisms of LLMs. I don't know if you find Terence Tao's understanding of mathematics sufficiently grounded; he's Chinese after all. He has some skepticism about LLMs contributing to deep, frontier mathematical research. Try to do more of that.
ThomasdelVasto
Κύριε, ποίησόν με ὄργανον τῆς ἀγάπης σου
15hr ago
Hah I had to stop midway through ep 3. Not because of the sexualization necessarily just a sort of 'what am I doing here' type moment. I find it increasingly hard to 'waste' time nowadays.
pigeonburger
15hr ago
To what extent is the current competency crisis in government, academia, etc. caused by an inability to spend time by oneself and actually put in the work?
Almost none of it, because IMO the competency crisis is caused by misaligned incentives. In government, the incentives are aligned with playing up tribal politics, not with competent management. In academia, it's in appealing to grant givers, making sensational claims that get published and cited, not producing solid science or advancing human knowledge. In business and especially for public companies, it's maximising current shareholder value rather than building a sustainable business. And so on...
That said, learning and putting in the work is a skill that I believe we in the West have regressed in. Some people expect to be good at something from the start or else they believe they'll never be good at it. Kids need to gain the specific insight of learning how to learn trained into them to grow into capable adults, and I think we might be currently failing at that.
sarker
It isn't happening, and if it is, it's a bad thing
15hr ago
People in my area run reds all the time, but I never saw people try to beat the green like this except in Puerto Rico.
It's nearly impossible to "make observations" for cross traffic when you're traveling at 55mph and the vegetation blocks line of sight to the traffic in question.
self_made_human
Grippy socks, grippy box
15hr ago
I don't think I've ever seen it applied, certainly not in a healthcare setup. If someone's getting utility out of it, it's not happening where I could see them. Which isn't the same as saying it has no utility, it just doesn't seem to come up.
FlyingLionWithABook
Has a C. S. Lewis quote for that.
15hr ago
7:1 starts with "judge not" but then immediately explains why you would want to do that: "lest ye be judged". It's saying "If you judge people then God will judge you." It's the exact same idea as 7:2 (which makes sense! 7:2 is literally the next sentence of the sermon!).
It's also worth noting that in John 8 at the end even Jesus (i.e., God) declines to judge her, and then says "Go and sin no more." Which means that acknowledging adultery is a sin is not the kind of judgement he is talking about. He's talking about the punishment part of judgment.
SubstantialFrivolity
I'm not even supposed to be here today
15hr ago
Interesting that NJ does have that law. Why did the judge instruct the jury as you indicated in your post, then, rather than saying that the law requires a driver to stop if able to do so safely? That seems like it would be more clear-cut as to wrongdoing.
urquan
Hold! What you are doing to us is wrong! Why do you do this thing?
15hr ago
I guess this is just flyover country being behind on the trends, but my impression is that tattoos, especially lots of them, still do signal criminality or BPD or sluttiness, or at least an attempt to look cool.
They're also just a lot less common here, so maybe they're still a reliable signal of something.
Incorrect (in this state).
NJ Statutes tit. 39 ch. 4 § 105:
Amber, or yellow, when shown alone following green[,] means traffic[ is] to stop before entering the intersection or nearest crosswalk, unless when the amber appears the vehicle or street car is so close to the intersection that with suitable brakes it cannot be stopped in safety.
Nothing is said about exiting the intersection before the light turns red.
§ 67:
No vehicle or street car shall be permitted by the owner or driver thereof to so occupy a street as to interfere with or interrupt the passage of other street cars or vehicles, nor shall the driver of a vehicle or street car drive such vehicle or street car into an intersection if preceding traffic prevents immediate clearance of the intersection.
That means a motorist must exit the intersection before any other light turns green, not before his light turns red. A traffic signal normally will have an all-red clearance interval of two or three seconds, so the difference between these two definitions is far from negligible.
TIRM
15hr ago
In 2025 your baron would have economic advisors who would correctly explain to him that his tax revenue and fiefdom GDP would increase if he allowed more people to enter.
Especially with martial arts, once I no longer had anything to prove to myself that I could do it, I just wasn't feeling it anymore.
Similar for me, but I swapped over to teaching it to others, which is quite rewarding on its own.
And you could always try some amateur fights if you want to challenge yourself (at the risk of injury).
But woodworking, at least for now, is fantastic.
3D printing is giving me a portion of this satisfaction of making something 'from scratch' and having a finished product at the end you can take pride in.
But so far that's mostly for trinkets and trivialities.
I dream of having a sizeable enclosed workspace on my property to tinker with cars and wood and produce fairly complex devices and objects. I am become Boomer, acquirer of hobbies.
I suspect gamification has spoiled our brains to expect more rewards for fake task than they deserve.
I think my only point there is that you're going to encounter the gamified stimuli anyway (unless you are VERY actively avoiding it), and it thus behooves you to let the 'good' stuff grab your attention (and money) or else something wasteful and trivial might, instead.
For instance: I do have Duolingo on my phone and I consider it a better use of my time than, say, Candy Crush or the bazillion basebuilding game clones out there, so its like, I dunno, substituting nicotine gum for actual cigarettes. I rage every time my phone updates and it auto-installs a bunch of the little ADHD time-suck apps on there that I have to remove manually.
And I can also say that there is zero chance I'll ever get 'bored' or feel 'satiated' with having sex with women, but that has run into the endless frustration that is modern dating that I bemoan elsewhere. I'm tempted to start setting aside a 'prostitute budget' for myself if I go another year or two without getting into a relationship, but I damn well know what its like to be intimate with someone you truly know and care about, and cares about you in return, so I don't think I can be truly happy just paying for it.
All these basic activities turn out to be the most fulfilling on a primal level, whoda thunk? (lots of people, it turns out, the modern world just wants to keep you distracted with candy and trinkets).
If you want personality pseudoscience I recommend the Enneagram over Myers-Briggs. It has a lot more depth. Myers-Briggs is focused on being descriptive, while Enneagram is more focused on being prescriptive. As in, "If I have this kind of personality type, what should I do to be a healthier and happier person?" And the advice is very good in my experience! At least for type Fives, I have not tried the advice for other types and can't testify to their accuracy and effectiveness. But if you're the kind of nut who finds categorizing by personality really fun, then you're probably a type Five anyway.
More options
Context Copy link