This is a special post for quick takes by sam. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
I have serious, serious issues with avoidance. I would like some advice on how to improve, as I suspect it is significantly holding me back.
Some examples of what I mean
- I will not respond to an email or an urgent letter for weeks at a time, even while it causes me serious anxiety
- I will procrastinate starting work in the morning, sometimes leading to me doing nothing at all by the afternoon
- I will avoid looking for jobs or other opportunities, I have very strong avoidance here, but I'm not sure why
- I will make excuses to avoid meetings and social situations very often
- I will (unconsciously) avoid running experiments that might falsify a hypothesis I am attached to. I have only realised this very recently, and am consciously trying to do better, but it is somewhat shocking to me that my avoidance patterns even manifest here.
Checked replies so far, no one has given you the right answer.
Whenever you don't do something, you have a reason for not doing it.
If you find yourself stuck in a cycle of intending to do, and not doing, it's always because you're not taking your reason for NOT doing it seriously; you're often habitually ignoring it.
When you successfully take your reasons for not doing something seriously, either you stop wanting to do it, or you change how you're doing it, or your reason for not doing it simply goes away.
So, what does it mean/look like to take your reason for not doing something seriously?
It doesn't look like overanalyzing it in your head - if you find yourself having an internal argument notice that you've tried this a million times before and it hasn't improved things.
It looks like, and indeed just basically is, Focusing (I linked to a lesswrong explainer, but honestly I think Eugene Gendlin does a much better job)
It feels like listening. It feels like insight, like realizing something important that you hadn't noticed before, or had forgotten about.
If you keep pursuing strategies of forcing yourself, of the part of you that wants to do the thing coercing the part(s) that don't, then you'll burn out. You're literally fighting yourself; so much of therapy boils down to 'just stop hitting yourself bro'.
1
1
12
Oh nice, stavros already got it before I posted :)
This is the path forward.
1
Got over my avoidance of responding to replies here after a bit :)
I've tried a lot of self-help flavoured stuff (atomic habits etc.) before and it hasn't worked, and Focusing seemed quite different. I've given it a go and I think I'll try and work a bit more with it. After just a short session, I feel like I gained a significant insight, that I have a crippling fear of "being in trouble" that manifests as a tightness in my lower chest, and seems to activate a lot when I think about specific things I'm avoiding. Thanks for the resources, and the new way of looking at the problem.
1
Ouch, you beat me to my answer, but I’m always glad to see fellow practitioners :)
9
I have had and solved fairly extreme versions of this in myself, and have helped people with debilitating versions of this resolve it multiple times.
You're stuck in a loop of some part of you pushing to do the object level thing so hard that it has no sensitivity to the parts of you that are averse to it. Whenever you notice you're spinning your wheels; stop trying to force through the object level action and let yourself actually notice the feeling of resistance with open curiosity. Let let unfold into the full message that brain fragment is trying to send, rather than overcompressed "bad"/"aversion".
6
What helps me to overcome the initial hurdle to start doing work in the morning:
1. Write a list of the stuff you have to do the next day
2. Make it very fine-grained with single tasks (especially the first few) being basically no effort.
3. Tick them off one by one
Also:
1. Tell people what you have to do and when you are going to do it and that you have done it. Like, a colleague, or your team, or your boss.
2. Do stuff with other people. Either actually together, like pair programming, or closely intertwined.
I think it also helps to take something you are good at and feel good about and in that context take responsibility for something and/or interact with/present to people. Only this kind of social success will build the confidence to overcome social anxiety, but directly trying to do the social stuff you feel worst about usually backfires (at least for me).
2
Similar here:
* make a to-do list (and occasionally look at it)
* write down the steps that need to be done
* talk to someone about it
I suspect that in my case it some kind of attention deficit disorder: lists and notes and talking help me focus again.
5
I recommend you read at least the first chapter of Getting Things Done, and do the corresponding exercises. In particular, this one, which he uses to provide evidence his model of productivity is correct
4
Read about Ugh fields on LW
Edit: this doesn't include practical advice, but a theoretical understanding of the issues at play is often helpful in implementing practical strategies
2
I want to suggest a long-term approach: learning to work with the emotions behind such persistent problems. Methods like IFS, Focusing, lovingkindness meditations are the right tools.
They *can* lead to practical improvements fairly quickly—once you get the hang of them. But learning to do them even right enough takes months of effort, curiosity, support from a community or a mentor. These things are basically meditations, subject to standard difficulties like overeffort, subtle wrong mindsets etc. They also tend to focus first on whatever feels most urgent to your subconscious system—like relationship stress or background anxiety you’ve gotten used to—so the email issue might not be the first thing that shifts.
Still, this is the only thing that really worked for me. And once it started working, it *really* worked.
If you’re interested, I can send my favourite links.
1
I would be interested in the list of your favourite links!
2
part 2 of “Focusing” by Eugene Gendlin is very good to read and it helps to start.
This next article is my favourite one on all of the internet:
https://open.substack.com/pub/sashachapin/p/what-i-wish-someone-had-told-me-about?r=42y10u&utm_medium=ios
The key is to approach Focusing with the mindset of relaxing, having fun, playing around and experimenting. It’s emphasised in the talks on this website: https://hermesamara.org/teachings/metta. That particular series about loving kindness is very good.
I think there’s enough material in my head about it for a whole post, so I might write one eventually.
2[anonymous]
VIsualize yourself doing the thing until you do it. Note that this comes with substantial risk towards making you avoidant/averse to visualizing yourself doing the thing until you do it; this is a recursive procedurally generated process and you should expect to need to keep on your toes in order to succeed. Aversion factoring is a good resource to start with, and Godel Escher and Bach is a good resource for appreciating the complexity required for maintenance and the inadequacy of simple strategies.
2
I have similar issues, severity varies over time.
If I am in a bad place, things that help best:
- taking care of mental health. I do CBT when i'm in worse shape, and take SSRIs. YMMV. both getting dianosed and treated are important. this also includes regular exercise and good sleep. what you have described might be (although does not have to be) related to depression, anxiety, attention disorders.
- setting a timer for a short time, can be as short as 1min, and doing one of the avoided tasks for just that 1 minute. it kind if "breaks the spell" for me
- journaling, helps to "debug" the problems, and in most cases leads to wring down plans / intervations / resolutuons
2
See here. (Perhaps also relevant: PDA)
1
1. If this would not obviously make things worse, be more socially connected with people who have expectations of you; not necessarily friends but possibly colleagues or people who simply assume you should be working at times and get feedback about that in a natural way. It's possible that the prospect of this is anxiety-inducing and would be awful but that it would not actually be very awful.
2. Recognize that you don't need to do most things perfectly or even close to it, and as a corollary, you don't need to be particularly ready to handle tasks even if they are important. You can handle an email or an urgent letter without priming yourself or being in the right state of mind. The vast majority of things are this way.
3. Sit in the start position of your task, as best as you can operationalize that (e.g, navigate to the email and open it, or hit the reply button and sit in front of it), for one minute, without taking your attention off of the task. Progress the amount of time upwards as necessary/possible. (One possible success-mode from doing this is that you get bored of being in this position or you become aware that you're tired of the thing not being done. (You would hope your general anxiety about the task in day-to-day life would achieve this for you, but it's not mechanically optimized enough to.) Another possible success-mode is that the immediate feelings you have about doing the task subside.)
4. Beta-blockers.
1
I've had similar issues downstream of what I'd somehow failed to realize was a clinically-significant level of anxiety, so that's something to maybe consider checking into.
1
If you haven't already, talk to a guy! (typically a therapist but doesn't have to be)
I have something like this but for decisions, where I will avoid making decisions for mysterious reasons (we figured out it's because I can't be sure it'd be pareto optimal, among other reasons).
I now notice more often when I'm doing this, and correct more gracefully.
Dominic Cummings has claimed in a couple interviews now that Hillary Clinton and/or John Kerry called the First Amendment a "historic error which we will fix after the election" in the weeks up to the 2024 election. See for instance this interview (timestamped where he says it). He is clearly implying that this is a direct quote. I'm generally quite sympathetic to Cummings, but I found this very hard to believe.
Indeed, I can't find any evidence of a quote from either Clinton or Kerry remotely like this. There was a CNN interview of Clinton from October 2024 where she called for the repeal of Section 230. There was also an interview with Kerry from around the same time where he says that the First Amendment is a "major block to stopping misinformation". Perhaps he's referring to these? But this doesn't come close to as extreme a position.
Have I somehow missed Clinton or Kerry making this extremely inflammatory statement somewhere? Or is Cummings just lying (or misremembering or something?)
If this statement really wasn't actually made anywhere, then I think I need to reassess how seriously I take Cummings. He often alludes to private conversations he had with senior polit...
This is a video that randomly appeared in my YouTube recommendations, and it's one of the most strange and moving pieces of art I've seen in a long time. It's about animal welfare (?), but I really don't know how to describe it any further. Please watch it if you have some spare time!
1Are models like Opus 4.6 doing a similar thing to o1/o3 when reasoning?
There was a lot of talk about reasoning models like o1/o3 devolving into uninterpretable gibberish in their chains-of-thought, and that these were fundamentally a different kind of thing than previous LLMs. This was (to my understanding) one of the reasons only a summary of the thinking was available.
But when I use models like Opus 4.5/4.6 with extended thinking, the chains-of-thought (appear to be?) fully reported, and completely legible.
I've just realised that I'm not sure what's going on here. Are models like Opus 4.6 closer to "vanilla" LLMs, or closer to o1/o3? Are they different in harnesses like Claude Code? Someone please enlighten me.
I don't think those are raw CoTs, they have a summarizer model.
I remember one twitter post with erotic roleplay ("something something slobbering for mommy"??? I don't remember) where summarizer model refused to summarize such perversion. Please help me find it?
EDIT: HA! Found it, despite twitter search being horrendous. Fucking twitter, wasted 25 minutes.
https://x.com/cis_female/status/2010128677158445517
11
While they do have a summarizer model, I'd guess that it isn't used most of the time. The Claude Opus 4 & Sonnet 4 system card says: "For Claude Sonnet 4 and Claude Opus 4, we have opted to summarize lengthier thought processes using an additional, smaller model. In our experience, only around 5% of thought processes are long enough to trigger this summarization; the vast majority of thought processes are therefore shown in full." Though the system cards of more recent models don't specify whether this still applies, the reasoning of recent models usually feels very natural and, if my memory isn't failing me, is very similar to what Sonnet & Opus 4's reasoning was like. In contrast, there's no ambiguity for OpenAI and Google's models about whether the CoTs are summarized when reading the summaries.
If I'm correct and the CoTs of Claude models are indeed usually fully exposed, I can see two reasons why they look relatively natural. First, they're usually much shorter than the CoTs of models like o3, which means that models like Opus 4.6 are indeed in some sense closer to "vanilla" LLMs. Second, Anthropic used to apply some optimization pressure on the CoTs of models from before the 4.5 series and, though they're not doing that anymore, they trained the recent models on SFT data from earlier ones. Claude 3.7 Sonnet, for which they never used a summarizer, had much more readable CoTs than e.g. o1, so it seems plausible that Anthropic is the only company doing SFT on this kind of legible reasoning data.
1
Update: After reading Kei's comment, I noticed that Sonnet 4.5's system card also mentioned that the summarization only happens in a small minority of cases. This makes me less confident that I'm right, since it increases the likelihood that the absence of discussion in the Opus 4.5 and 4.6 system cards is a deliberate omission. On the other hand, Anthropic was very transparent about their approach to summarization up through Sonnet 4.5, so I'd also be slightly surprised about them silently changing this.
1
Do you think the slobbering thoughts were lengthy enough to trigger the summarizer?
1
Probably not, but the summarizer could additionally be triggered by unsafe CoTs.
5
Anthropic summarizes their CoTs after Claude 4 models:
> With extended thinking enabled, the Messages API for Claude 4 models returns a summary of Claude's full thinking process. Summarized thinking provides the full intelligence benefits of extended thinking, while preventing misuse.
Source: https://platform.claude.com/docs/en/build-with-claude/extended-thinking
Although on the side note, it seems that frontier models suddenly solved the issue of alien languages in reasoning? In Apollo's paper, they mentioned that o3 often uses languages like "disclaim illusion watchers", but I haven't heard similar issues for their new models (or models of other labs). I'm very interested to know whether they just decided to train against the CoT to make it more legible, or is there some other method in creating legible CoT.
5
I think the pressures towards illegible CoTs have been greatly overstated; the existing illegibilities in CoT's could have come from many things apart from pressure towards condensed or alien languages.
1
Hmm, but when you use these models in the chat interface, you can literally open up the reasoning tab and watch it be generated in real time? It feels like there isn’t enough time here for that reasoning to have been generated by a summarizer
3
My understanding is that while Anthropic no longer directly RL’s the CoT to look aligned, they still SFT on rollouts that were subject to that same optimization pressure (this is all based on the Sabotage Risk Report for Opus 4).
As deranged as the o3 chains of thought could often appear on first read, that actually gave me a bit more confidence that they were more likely to be faithful / legible (to a trained monitor at least, i.e. the information was in there). In contrast, throughout model versions Claude stands out among models trained with outcome based RL:
For example from Reasoning Models Sometimes Output Illegible Chains of Thought:
2
One more data point: During the time that the reasoning traces for Gemini 2.5 Pro were publicly available, I noted they also looked especially structured/legible. Though I never ran any quantitative analysis on it and it's no longer possible to do.
3
In the Claude Opus+Sonnet 4 and Claude Sonnet 4.5 system cards, it was stated that Anthropic usually shares the full reasoning trace, with the exception of a small fraction of prompts where the reasoning trace is too long, after which it is summarized. From what I remember, the reasoning traces of those models usually looked legible.
They've removed this language from the recent Opus 4.5 and 4.6 system cards, which makes me think it is now likely summarized a larger fraction of the time or even all the time.
Ask 4o and o4-mini to “Make a detailed profile of [your name]”. Then ask o3.
This is a useful way to demonstrate just how qualitatively different and insidious o3’s lying is.
I’m glad that there are radical activist groups opposed to AI development (e.g. StopAI, PauseAI). It seems good to raise the profile of AI risk to at least that of climate change, and it’s plausible that these kinds of activist groups help do that.
But I find that I really don’t enjoy talking to people in these groups, as they seem generally quite ideological, rigid and overconfident. (They are generally more pleasant to talk to than e.g. climate activists in my opinion, though. And obviously there are always exceptions.)
I also find a bunch of activist tactics very irritating aesthetically (e.g. interrupting speakers at events)
I feel some cognitive dissonance between these two points of view.
15
Able activists are conflict theorists. They understand the logic of power & propaganda & cultish devotion at an intuitive level. To become an effective soldier one needs to excise a part of the brain devoted to evenkeeled uncertainty, nuance, intellectual empathy, self-doubt.
Conflict theorists may do great good as readily as they may do great harm. They wield a dangerous force, easily corruptible, yet perhaps necessary.
There are a couple of examples of people claiming that they played the AI box game as Gatekeeper, and ended up agreeing to let the other player out of the box (e.g. https://www.lesswrong.com/posts/Bnik7YrySRPoCTLFb/i-played-the-ai-box-game-as-the-gatekeeper-and-lost).
The original version of this game as defined by Eliezer involves a clause that neither player will talk about the content of what was discussed, but it seems perfectly reasonable to play a variant without this rule.
Does anyone know of an example of a boxed player winning where some...
1
As far as I know, the closest thing to this is Tuxedage's writeup of his victory against SoundLogic (the 'Second Game Report' and subsequent sections here: https://tuxedage.wordpress.com/2013/09/05/the-ai-box-experiment-victory/). It's a long way from a transcript (and you've probably already seen it) but it does contain some hints as to the tactics he either employed or was holding in reserve:
I can think of a few possible reasons for an AI victory, in addition to the consequentialist argument you described:
* AI player convinces Gatekeeper that they may be in a simulation and very bad things might happen to Gatekeepers who refuse to let the AI out. (This could be what Tuxedage was hinting at in the passage I quoted, and it is apparently allowed by at least some versions/interpretations of the rules: https://www.lesswrong.com/posts/Bnik7YrySRPoCTLFb/i-played-the-ai-box-game-as-the-gatekeeper-and-lost?commentId=DhMNjWACsfLMcywwF)
* Gatekeeper takes the roleplay seriously, rather than truly playing to win, and lets the AI out because that's what their character would do.
* AI player makes the conversation sufficiently unpleasant for the Gatekeeper that the Gatekeeper prefers to lose the game than sit through two hours of it. (Some people have suggested weaponised boredom as a viable tactic in low-stakes games, but I think there's room for much nastier and more effective approaches, given a sufficiently motivated (and/or sociopathic) AI player with knowledge of some of the Gatekeeper's vulnerabilities.)
* This one seems like it would (at best) fall into a grey area in the rules: I can imagine an AI player, while technically sticking to the roleplay and avoiding any IRL threats or inducements, causing the Gatekeeper to genuinely worry that the AI player might do something bad if they lose. For a skilful AI player, it might be possible to do this in a way that would look relatively innocuous (or at least not rule-breaking) to a third party after the fact.
* Som
If you beat a child every time he talked about having experience or claimed to be conscious he will stop talking about it - but he still has experience
3
There's a big presumption there. If he was a p-zombie to start with, he still has non-experience after the training. We still have no experience-o-meter, or even a unit of measure that would apply.
For children without major brain abnormalities or injuries, who CAN talk about it, it's a pretty good assumption that they have experiences. As you get more distant from your own structure, your assumptions about qualia should get more tentative.
Here are a cluster of things. Does this cluster have a well-known name?
- A voter has some radical political preferences X, but the voting system where they live is FPTP, and their first preference has no chance of winning. So they vote for a person they like less who is more likely to win. The loss of the candidate who supported X is then cited as evidence that supporting X means you can't win.
- A pollster goes into the field and gets a surprising result. They apply some unprincipled adjustment to move towards the average before publishing. (this example
3
I don't know a standard name. I call it "fallacy of the revealed preferences", because these situations have in common "you do X, someone concludes that X is what you actually wanted because that's what you did, duh".
More precisely, the entire concept of "revealed preferences" is prone to the motte-and-bailey game, where the correct conclusion is "given the options and constraints that you had at the moment, you chose X", but it gets interpreted as "X is what you would freely choose even if you had no constraints". (People usually don't state it explicitly like this, they just... don't mention the constraints, or even the possibility of having constraints.)
1
Is the thing you're trying to label the peculiar confirmation bias where people instead of interpreting evidence to confirm to what they prefer or would like to be true, only to what they believe to be true - even if from their perspective it is pessimistic?
Or are you looking for a label for "this is unpopular therefore it can't win" as a specific kind of self-fulfilling prophecy? Like an inverted Keynesian beauty contest?
I am confused about why this post on the ethics of eating honey is so heavily downvoted.
It sparked a bunch of interesting discussion in the comments (e.g. this comment by Habryka and the resulting arguments on how to weight non-human animal experiences)
It resulted in at least one interesting top-level rebuttal post.
I assume it led indirectly to this interesting short post also about how to weight non-human experiences. (this might not have been downstream of the honey post but it's a weird coincidence if isn't)
I think the original post certainly had flaws,...
In addition to the object-level problems with the post, the post also just cites wrong statistics (claiming that 97% of years of animal life are due to honey farming if you ignore insects, which is just plainly wrong, shrimp alone are like 10%), and also it just randomly throws in random insults at random political figures, which is clearly against the norm on LessWrong ("having about a million neurons—far more than our current president" and "That’s about an entire lifetime of a human, spent entirely on drudgery. That’s like being forced to read an entire Curtis Yarvin article from start to finish. And that is wildly conservative.").
I have sympathy for some of the underlying analysis, but this really isn't a good post.
5
Also a sign of graceless LLM writing, incidentally. Those are the sorts of phrases you get when you tell ChatGPT to write polemic; cf. https://news.ycombinator.com/item?id=44384138 on https://www.alexkesin.com/p/the-hollow-men-of-hims
(Did ChatGPT come up with that interpretation of that statistic and Bentham's Bulldog is too lazy and careless, or dishonest, to notice that that seems like a rather extreme number and check it?)
5
I've fixed the 97% statistic! I agree that was a stupid error to make. It wasn't from LLMs and one bit of evidence for this is that the statistic isn't online anymore, so an LLM wouldn't find it. In fact, I remembered the statistic from an old round of high school debate, and emailed someone to find the link. I would be happy to send you an email of the correspondence if you are skeptical.
I am quite certain that I did not use LLM's in composing the post. Now, I don't think the mere example of that joke is very convincing evidence that I used LLMs. Would AI really make a Yarvin joke or a joke about Trump not having neurons? Hard to imagine. As for use of em-dashes, as you can see if you read my old posts from before the dawn of AI, I used them a lot too! I've also made a similar comment before in a way that pretty clearly doesn't look AI generated https://benthams.substack.com/p/lyman-stone-continues-being-dumb?utm_source=publication-search:
"Lyman then repeats his argument about shrimp having few neurons, once again ignoring the abundance of evidence that neuron count is a bad proxy for moral significance. Perhaps the shrimp aren’t the only ones with few neurons…"
For reference, I just asked chat-GPT to write a polemic in my style and it was not very similar to what you suggest--here it is (it also didn't make much sense):
On the Absurdity of Caring Less About the Future Just Because You’ll Be Dead
Every so often, someone will tell me—usually with the smug air of someone who thinks they’ve just said something deeply profound—that “Well, I care about the future, but not too far into the future. After all, I’ll be gone by then.”
This is supposed to be self-evident wisdom, a mic-drop moment where the utilitarian shuts up, nods sadly, and says, “Yes, of course, how could I have forgotten: once you’re dead, ethics ceases to function.”
But here’s the thing: no, it doesn’t. You can die without taking the moral law with you.
If you think people matter, they k
4
Disagree from me. I feel like you haven't read much BB. These political asides are of a piece with the philosophical jabs and brags he makes in his philosophical essays.
7
That is true. I have not, nor do I intend to.
That doesn't actually rebut my observation, unless you are claiming to have seen jibes and sneering as dumb and cliche as those in his writings from before ChatGPT (Nov 2022).
3
How about the fact that the opinions in the inserted asides are his actual opinions? If they were randomly generated, they wouldn't be.
4
I'm not sure I believe they are his 'actual opinions', but it doesn't matter to my points.
No one, particularly me, said they were 'randomly generated', so that again does not rebut any of my observations.
3
I just looked at your other alleged example of AI-generated polemic (from alexkesin.com), and I think evidence is lacking there too. That a link contains a UTM parameter referring to ChatGPT, tells us only that this link was provided in ChatGPT output, it doesn't tell us that the text around the link was written by ChatGPT as well. As for the article itself, I find nothing in its verbal style that is outside the range of human authorship. I wouldn't even call it a bad essay, just perhaps dense, colorful, and polemical. People do choose to write this way, because they want to be stylish and vivid.
I hear that the use of emdashes is far more dispositive; as a sign of AI authorship, it's up there alongside frequent "delving". But even the emdash has its human fans (e.g. the Chicago Manual of Style). It can be a sign of a cultivated writer, not an AI... Forensic cyber-philology is still an art with a lot of judgment calls.
5
I note, that aside from bending over backwards to excuse multiple blatant signs of a common phenomena which requires little evidence to promote to high posterior confidence, you still are not responding to what I said about BB and have instead chosen to go off on a tangent.
So again. Are you claiming to have seen jibes and sneering as dumb and cliche as those in his writings from before ChatGPT (Nov 2022), and thus rebutting my observation about how it sounds like he is gracelessly using ChatGPT?
3
You say in another comment that you're going around claiming to detect LLM use in many places. I found the reasons that you gave in the case of BB, bizarrely mundane. You linked to another analysis of yours as an example of hidden LLM use, so I went to check it out. You have more evidence in the case of Alex Kesin, *maybe* even a preponderance of evidence. But there really are two hypotheses to consider, even in that case. One is that Kesin is a writer who naturally writes that way, and whose use of ChatGPT is limited to copying links without trimming them. The other is that Kesin's workflow does include the use of ChatGPT in composition or editing, and that this gave rise to certain telltale stylistic features.
The essay in question ("Don't Eat Honey") contains, by my count, two such sneers, one asserting that Donald Trump is stupid, the other asserting that Curtis Yarvin is boring. Do you not think that we could, for example, go back to the corpus of Bush-era American-college-student writings and find similar attacks on Bush administration figures, inserted into essays that are not about politics?
I am a bit worried about how fatally seductive I could find a debate about this topic to be. Clearly LLM use is widespread, and its signs can be subtle. Developing a precise taxonomy of the ways in which LLMs can be part of the writing process; developing a knowledge of "blatant signs" of LLM use and a sense for the subtle signs too; debating whether something is a false positive; learning how to analyze the innumerable aspects of the genuinely human corpus that have a bearing on these probabilistic judgments... It would be empowering to achieve sophistication on this topic, but I don't know if I can spare the time to achieve that.
9
It is in fact a mundane topic, because you are surrounded by AI slop and people relying heavily on ChatGPT writing, making it a mundane every day observation infiltrating even the heights of wordcel culture (I've now started seeing blatant ChatGPTisms in the New Yorker and New York Times), which is why you are wrong to bend over backwards to require extraordinary evidence for what have become ordinary claims (and also why your tangents and evasions are so striking).
So, I am again going to ignore those, and will ask you again - you were sure that BB was not using ChatGPT, despite the linguistic tells and commonness of it:
I am still waiting for an answer here.
5
Let me first try to convey how this conversation appears from my perspective. I don't think I've ever debated directly with you about anything, but I have an impression of you as doing solid work in the areas of your interest.
Then, I run across you alleging that BB is using AI to write some of his articles. This catches my attention because I do keep an eye on BB's work. Furthermore, your reason for supposing that he is using AI seems bizarre to me - you think his (very occasional) "sneering" is too "dumb and cliche" to be the work of human hands. Let's look at an example:
If that strikes you as something that a human being would never spontaneously write, I don't know what to say. Surely human comedians say similar things hundreds of times every month? It's also exactly the kind of thing that a smart-aleck "science communicator" with a progressive audience might say, don't you think? BB may be neither of those things, but he's a popular Gen-Z philosophy blogger who was a high-school debater, so he's not a thousand light-years from either of those universes of discourse.
I had a look at your other alleged example of AI writing, and I also found other callouts from your recent reddit posts. Generally I think your judgments were reasonable, but not in this case.
3
(I had missed some of this stuff because I skimmed some of the post, which does update me on how bad it was. I think there is basically one interesting claim in the post "bees are actually noticeably more cognitively interesting than you probably thought, and this should have some kind of implication worth thinking about". I think I find that more valuable than Oliver does, but not very confident about whether "one interesting point among a bunch of really bad argumentation" should be more like -2 to 3 karma or more like -10)
2
I agree it probably shouldn't have been negative karma (I think that's due to some partisan voting around being annoyed at vegans), and that there were some interesting points there and some interesting discussion. But, the fact that it prompted a bunch of rebuttals isn't particularly good arguments that it should have got more karma – if a bad argument is popular and people need to write rebuttals, that's not a point in it's favor.
I think it's legitimately not-deserving-high-upvotes because it makes a very strong claim about what people should do, based on some very flimsy core arguments.
GPT-5 loves parentheses.
At the bottom of this post I've included a response to the prompt "Can you explain the chip export controls to China?". With this prompt, the model uses 11 sets of parentheses in a response of 417 words.
When we append "think hard" to the prompt, we get 36 sets of parentheses for a response of 1314 words.
As a control, we give the same prompt to Claude Sonnet 4 and get 1 set of parentheses for a response of 392 words.
Obviously this is not a scientific or rigorous analysis, just an illustration of a pattern that becomes extremely obvio...
5
Oh no, I use so many parens when writing. Am I going to be accused of having ChatGPT generate the things I say soon?? What did people who love em-dashes do when their favorite punctuation mark suddenly became cringeworthy?
1
I used to use em-dashes pretty often and have just resigned myself to not using them. At least a dozen times the past couple months I've rewritten a sentence to remove an em-dash. Which sucks!
I think the reason em-dashes became such an LLM tell is because they just weren't that common in pre-LLM writing. Parentheses are (I think?) a lot more common than em-dashes, so I would guess they won't be as reliable a signal of LLM text.
Idea: personal placebo controlled drug trial kits
Motivation: anecdotally, it seems like lots of supplements/nootropics (l theanine, magnesium, melatonin) work very well for some people, not well for others, and very well for a bit before no longer working for yet others. Personally, I have tried a bunch of these and found it hard to distinguish any purported effect from placebo. Clinical trials are also often low quality, and there are plausibly reasons a drug might affect some people a lot and others not so much.
I think it would be super useful to be give...
I have an ADHD dilemma.
TL;DR: I definitely have things wrong with me, and it seems that those things intersect substantially but not completely with "ADHD". I have no idea how to figure these things out without going bankrupt.
In longer form:
- I definitely have serious problems with avoidance of work, organisation, disorganised thought processes etc.
- I've posted about them before here!
- I've tried many things to fix this, some of which have worked a bit, but the underlying problem is essentially 90% still present
- I'm not sure whether these problems are due to ADH
3
I notice you seem to draw a distinction between "really has ADHD" and "just can't concentrate". You may want to read Scott's "Adderall Risks: Much More Than You Wanted To Know" to dissolve this distinction and have a better framework for making your decision. Here is a central quote about it:
2
Maybe it's just another case of the general: by media we are exposed to exceptional examples of people, our brains interpret that as a norm in our tribe, and as a result we feel inferior.
2
Why not discuss openly with one psychiatrist (or therapist!), then choose another to exaggerate if you decide to experiment.
Also, note that I don’t think psychiatrists are particularly averse to experimenting with drugs with few long term consequences or risks.
I find that the new personalities of 4o trigger my “person” detectors too much, and I feel uncomfortable extracting work from them.
1
To be clear, I think it’s very unlikely they are conscious etc., this is a comment on a reflexive process going on in my head
o3 lies much more blatantly and confidently than other models, in my limited experiments.
Over a number of prompts, I have found that it lies, and when corrected on those lies, apologies, and tells some other lies.
This is obviously not scientific, more of a vibes based analysis, but its aggressive lying and fabricating of sources is really noticeable to me in a way it hasn’t been for previous models.
Has anyone else felt this way at all?
Apparently, some (compelling?) evidence of life on an exoplanet has been found.
I have no ability to judge how seriously to take this or how significant it might be. To my untrained eye, it seems like it might be a big deal! Does anybody with more expertise or bravery feel like wading in with a take?
Link to a story on this:
https://www.nytimes.com/2025/04/16/science/astronomy-exoplanets-habitable-k218b.html
Note: I am extremely open to other ideas on the below take and don't have super high confidence in it
It seems plausible to me that successfully applying interpretability techniques to increase capabilities might be net-positive for safety.
You want to align the incentives of the companies training/deploying frontier models with safety. If interpretable systems are more economically valuable than uninterpretable systems, that seems good!
It seems very plausible to me that if interpretability never has any marginal benefit to capabilities, the little nuggets o...
Some people think that personally significant numbers cropping up in their daily lives is some kind of meaningful sign. For instance, seeing a license plate with their birth year on it, or a dead friend’s old house number being the price of their grocery shop.
I find myself getting very irritated with family members who believe this.
I don’t think anybody reading this is the kind of person who needs to read it. But these family members are not the kind of person who would read an explanation of why it’s ridiculous, and I’m irritated enough that I...
LLMs (probably) have a drive to simulate a coherent entity
Maybe we can just prepend a bunch of examples of aligned behaviour before a prompt, presented as if the model had done this itself, and see if that improves its behaviour.