Alex Darby

Karma: 154

Alex Darby 1 Jun 2026 22:20 UTC
8 points
14
on: Lighthaven East—A Feasibility Study
You appear to be looking for a leader. Why don’t you pick yourself?

Alex Darby 31 May 2026 7:47 UTC
12 points
3
on: Comment on “Banning Said Achmiz”
The revised passage is less engaging writing than the original. (It was a really clever analogy.) But it’s clear, and for that I have Said to thank
LessWrong is, to me, not about building a community of people writing dry, technically correct essays on various topics. It is about actually become less wrong, both as individuals who learn from each other, and as a community. This means “other people reading your writing” is not optional.
In this case, I see an example of Said ignoring the tradeoff between engagement, and clarity of the argument. Engaging writing is good. It means people actually read the post, instead of bouncing off after two paragraphs.
As far as I am concerned, this analogy points to what I consider the ‘true reason’ behind Said’s ban.
Net positive effect due to Said causing local argument improvements in the author by commenting.
vs
Net negative effect due to less posts due to people worried about (e.g.) Said nit-picking a non-central point, arguing an actual flaw using blunt and inconsiderate language, or socially trapping the author in a long comment chain that gets further and further away from the post and into critiquing each other’s comments.
If the moderators think the chilling effect of Said on posts and comments is worse for {rationality/the site/a third thing}, a ban is the right decision. If they don’t, a mistake was made.

In all of the above I have attempted to phrase my views on Said’s ban in neutral language observing the tradeoffs involved, but my n=1 example is “I comment more now that I won’t have Said making me regret it, and my comments, while not perfect, are net positive at making the community less wrong, so the ban has made my experience much better.

Alex Darby 30 May 2026 0:56 UTC
2 points
1
in reply to: johnswentworth’s comment on: Some Dating Stories
I don’t mind the original comment, and think it should exist.
My model is that it’s a kinda of “ultra-strong downvote”, that also appears to the side’s moderators in keeping dating-adjacent topics off of frontpage, as well as discouraging them in general.
I also think you should keep posting whatever you’d like, and we let karma sort out what people see. But I also think the correct number of almost-persuasive comments that are merely ultra-strong downvotes/upvotes is nonzero.

Alex Darby 25 May 2026 15:55 UTC
2 points
0
in reply to: Neel Nanda’s comment on: Sayhan Yalvaçer’s Shortform
Solution: allow people to alias their username to a real name if anyone uses the citation tool on them?
Doesn’t solve the problem for people who actually want to be pseudonymous, but does solve the problem for e.g. habryka

Alex Darby 16 May 2026 22:31 UTC
5 points
0
in reply to: Eliezer Yudkowsky’s comment on: Irretrievability; or, Murphy’s Curse of Oneshotness upon ASI
The point here is one of utility function of France, no?
Even with this massive strategic miscalculation, France suffered less than half as many casualties in WWII as it had during WWI.
I think this is non-central to the post, and doesn’t undermine any of your central points.
However, I see the primary thrust of the Maginot line as “force both sides to pit all their strength against each other”, and the secondary as “actually win”, with the human understanding that losing a war to other humans is usually not an existential threat to the societies involved.

In terms of your point: “France was incorrect about a detail in war games, and therefore they lost on the first critical try according to the utility ‘beat the Germans’” is true.

The counterpoint: “France was succeeding according to the utility: ‘minimise sum of German + French deaths due to F/G conflict in next war, with an additional winning term’” may also be correct.

At this point we’re very non-central though, I think.

Alex Darby 23 Apr 2026 13:27 UTC
1 point
0
on: Annoyingly Principled People, and what befalls them
How do we tell, from the inside, the difference between pushing our friends and our society to uphold norm X in a not-too-annoying way when we’re right about the importance of norm X, and upholding a different norm Y where we’re wrong about its importance? What differentiates an “Alice” from an “Alex” in the language of the post?

(motivating question that I am asking abstractly above because I think it’ll be more helpful: suppose Dave has a norm X that he believes in. What about his situation will mean that if he pushes harder he turns into an Alice (and not an Alex)).

Alex Darby 22 Apr 2026 18:40 UTC
16 points
−2
on: Preventing extinction from ASI on a $50M yearly budget
What happens if you win?

I ask because in some sense, it meets the stated goal of any non-profit if the injustice or issue they work in opposition to were to go away. However, non-profits in practice respond badly to this, and pivot away from the original topic area.
To turn this into a concrete question, suppose we spend $300 million total over the next 5 years, and get a global ASI ban, enforced by some combination of limits on GPUs/compute and limits/reporting on (low-compute) fundamental research. What happens next with your budget, and do you currently have a plan for how to retain the ethos of “let’s not all die to AI” as we pivot from the “JUST STOP IT” phase to the “okay now what” phase.

examples of things that could go wrong (warning: long-run predictions are difficult. these are mostly intuition pumps)
- the political coalition that ends up banning ASI includes a generalised anti-technology faction which continues to gain in power and attempt to ban more technologies. Is your org working against this, or is it somebody else’s job.
- we stop AI, and twenty years and over a billion dollars later, you start to lose funders.

Alex Darby 22 Apr 2026 13:43 UTC
2 points
0
in reply to: Ruby’s comment on: Posts I don’t have time to write
Why is the 15-sentence epistemic status meta-commentary at the start rather than at the end?

Alex Darby 22 Apr 2026 10:02 UTC
10 points
0
on: Evil is bad, actually (Vassar and Olivia Schaefer)
In your claims, how much of these patterns are “Vassar et al. are doing different and new damaging things every year”, and how much is “Vassar et al. have been running the same playbook for approximately the last decade, with differing levels of success over the years, but every year we get more evidence”?

Alex Darby 21 Apr 2026 15:09 UTC
2 points
0
in reply to: Linch’s comment on: Eric Neyman’s Shortform
I don’t think “true if big” is a good claim to make? Plenty of things have gotten big that aren’t true, and the argument ‘move’ of “noticing that someone’s claim proves a lot” is a good one to make without implying the converse.

(epistemic status: yes, I am aware the parent comment is a joke)
(joke status: poor. explaining the joke makes it worse)

Alex Darby 19 Apr 2026 14:50 UTC
4 points
1
in reply to: Olivia Scharfman’s comment on: There are only four skills: design, technical, management and physical
According to my model of what’s going on:

(i) all of this is downstream of general intelligence.
(ii) according to Ollie’s galaxy-brained categorisation, people who are good at management will pick up relationship-building skills very quickly if they need them, whereas people who are good at design will be similarly bad at management and relationship-building for a year or so if they need to pick it up.

I think these categories allow two things to exist within them that have disjoint characteristics (as you say relationship-building and management are different—the crux is whether these skills end up correlated in people, and whether learning many of these disjoint skills helps with the rest.

Alex Darby 19 Apr 2026 14:42 UTC
7 points
3
in reply to: habryka’s comment on: Jan_Kulveit’s Shortform
Inkhaven (and Inkhaven-like) posts are frequently somewhere between ‘low-effort’ and ‘personal blog’ if on LessWrong.
fwiw, I enjoy more of these on LessWrong on the margin. The bar for quality that a post needs to enable marginally more rational thought is not that high, and interacting with lower-effort or more casual posts is psychologically easier, and hence enables me to actually change my mind in practice more often.

Alex Darby 17 Apr 2026 23:21 UTC
1 point
0
in reply to: cousin_it’s comment on: Let goodness conquer all that it can defend
I would encourage you (potentially with moderator help) to split this comment (and its children) into two comments, so we have two discussion threads—both discussions seem worthwhile, but mixing them is probably bad.

Alex Darby 17 Apr 2026 23:20 UTC
3 points
1
in reply to: habryka’s comment on: Let goodness conquer all that it can defend
I would be highly interested in reading an “against postmodernism” follow up.
I think the deeper challenge (which is not particularly relevant in the America example) is the idea that there are (a) some things which we think are good and we do them and they are good, and there are also (b) some things which we think are good that are actually bad, but not in a way where we can tell ‘from the inside’ right now. I think this point is more at home in a hypothetical post that is not about postmodernism, which (i) seems like an interesting post, and (ii) is not the one I am encouraging you to write.
(the third leg of this rule of three is (c) things that we think will lead to good ends but violate basic deontology, which we don’t do; https://www.lesswrong.com/s/AmFb5xWbPWWQyQ244/p/K9ZaZXDnL3SEmYZqB. What happens when multiple people have different deontological frameworks is left as an exercise for the reader).

(epistemic status: typed quickly, mostly encouragement of a future post, that I want to exist, where this encouragement is much stronger as a comment than as an upvote)

Alex Darby 17 Apr 2026 22:30 UTC
20 points
0
in reply to: Zack_M_Davis’s comment on: Do not conquer what you cannot defend
According to me (and potentially nobody else), I view LessWrong as a place where we do argument in the truth-building/philosophical/debate sense, and not in the shouting match sense. I think there is a way where we can do argument that works, but the above was not working for me.
[I felt like the above was getting into “shouting match” territory more than “squishing our different models of the situation together and attempting to do our best to get to Aumann’s Agreement Theorem in real life.
(note this is mostly because I noticed myself getting defensive in my own head—your comments may have worked perfectly well on the same words posted by someone else).]

According to me, this is good. The reason that comes to mind is “we want LessWrong to be a place of repeated idea exchange, and therefore people getting alienated is bad because then they might stop posting—model-sharing leads to much less alienation than bad-tempered argument”, although this may not be cruxy.

(epistemic status—typed quickly. Am interested if you disagree with my central point. My examples almost certainly have non-cruxy holes in them)

Alex Darby 17 Apr 2026 19:47 UTC
13 points
0
in reply to: Zack_M_Davis’s comment on: Do not conquer what you cannot defend
opting out of this conversation.
I feel like I have a thesis about how generating cultural information in the modern world involves writing essays where, if you do well, you impact more people who didn’t read the source material than did (e.g. my Korzybski point), and you are ignoring my central point.
Instead, you are aiming to persuade me of your point by using weaknesses in my analogies.
LessWrong is for learning about each other’s models, not for having an argument. We’re having an argument. I’m deliberately not engaging with your most recent points because I don’t want arguments like this on my favourite website.

Alex Darby 17 Apr 2026 16:41 UTC
6 points
4
in reply to: Zack_M_Davis’s comment on: Do not conquer what you cannot defend
I disagree. Of course the Sequences are about inculcating social norms, as is almost all large-scale human communication. Norms are how humans actually communicate ideas.
An explanation: when humans think in groups, their thinking is shaped by the norms of the group. e.g: the way one would post to obtain social status and respect on 4chan is different from LessWrong 2.0.
LessWrong’s norms (i.e. explain, don’t persuade. avoid insulting people about their knowledge of source material, get curious about other people’s models, etc) have been built (both in deliberate ways by Habryka/Lightcone, and in more nebulous social ways) to make it easier for people to be better at thinking when they communicate using them.
We can also reverse this. Someone who thoroughly understands LessWrong/rationalist/Yudkowskian/etc norms will find most of the content of the sequences obvious. E.g. the norm “offer concrete predictions” is making beliefs pay rent. the norm “offer concrete models” enables crux-finding. And so on.
Yes, the Sequences say they are about giving you the tools to know which computations result in accurate beliefs. However, the average person who benefits from a specific Yudkowsky essay probably has not read that essay (claim not fully justified in this comment, it’s long enough already). To be more specific, the average person who benefits from the idea “the map is not the territory” has not read the original (Korzybski, 1931). Instead, they read/listened to someone who read/listened to… (some intermediate layers) until we get to an original source.
We call the ideas that propagate throughout a community until they become obvious to most members ‘norms’.
(epistemic status: feeling incredibly insulted. I’ve read the Sequences)

Alex Darby 16 Apr 2026 7:46 UTC
8 points
3
on: Do not conquer what you cannot defend
Executive Summary: LessWrong 2.0 as it actually exists runs at Bus Factor Habryka, and this is probably fine.
(epistemic status: I notice my thesis is confused, but want comments on it anyway. Writing a long comment since I don’t have time to write a shorter one.)
If we compare this post directly to LessWrong, things become less clear to me, because I’m not certain which elements of LessWrong are designed to persist.
When we look at LessWrong 1.0 (before my time), then, as described by “what Alex of LessWrong 2.0 believes about history”, it consists of (i) The Sequences, which are widely read (tangent: probably not-required-reading, in that many modern community members pick up the norms without having read the original source material) and also (ii) comments on blog posts, which are archived, but ~nobody reads them. (tangent: I believe archiving the comments is strongly good for ‘it makes the information environment better for future historians and regardless of whether we expect these future historians to actually exist, it’s good for humans to act as if there will be a Future and be pro-social in relation to it’).
LessWrong 2.0 is the site that Oliver is the chief moderator of. This is your walled garden. It’s a good garden. I spend a lot of time here. It is inexorably linked with Lighthaven (technically separate) and Lightcone (umbrella org) and the surrounding community. However, unlike Washington and the US Government, or the French and their revolutions, where the institution is the mechanisms of state, and it is designed to be a legible and predictable monopoly over violence that persists for generations, I don’t see why we have to do the same?
You have short timelines (meaning <20yrs with high probability, <40yrs with very high probability) (epistemic status: if this is wrong I have to throw out a lot of my models of the world).
My Oliver-model believes the current version of LessWrong 2.0 (with “bus factor 1=Ollie”) can persist for as long as is relevant pre-AGI/ASI, unless the world drastically changes in ways he doesn’t expect.
Therefore I’m not sure you need a succession plan? You don’t need to defend the information ecosystem of LessWrong forever. This institution doesn’t need to last forever—I certainly hope “we win AI”, and we can continue with a “LessWrong 3.0″ that is similar and yet also different, but LessWrong 2.0 doesn’t need to be that.

Alex Darby 5 Apr 2026 15:42 UTC
4 points
0
in reply to: Daniel Kokotajlo’s comment on: Q1 2026 Timelines Update
I would say the comment had two purposes:
(i) if my implications were wrong/misleading, I would be corrected (thank you!),
(ii) to encourage future posts on this/related topics to think hard about the sensitivity of the world-model to different types of future information.

I wouldn’t say I want a more quantitative measure—although it’s important, trying to pin them down too much can make the very precise written content into “the thing that’s easier to write in a precise way” and not “what we actually believe”.

Alex Darby 4 Apr 2026 19:47 UTC
4 points
0
on: Q1 2026 Timelines Update
Thank you for posting. In terms of “information not in this post that would be useful”, a measure of “how sensitive are these timelines to various potential events” seems useful.

e.g. post implies
high-ish sensitivity to “general impressiveness of new models” (e.g. Opus 4.7 (or Mythos/Capybara), GPT5.5),
low but non-zero sensitivity to “opinions of people in labs about their future output”.