gjm

Karma: 31,570

Hi. I’m Gareth McCaughan. I’ve been a consistent reader and occasional commenter since the Overcoming Bias days. My LW username is “gjm” (not “Gjm” despite the wiki software’s preference for that capitalization). Elsewehere I generally go by one of “g”, “gjm”, or “gjm11”. The URL listed here is for my website and blog, neither of which has been substantially updated for several years. I live near Cambridge (UK) and work for Hewlett-Packard (who acquired the company that acquired what remained of the small company I used to work for, after they were acquired by someone else). My business cards say “mathematician” but in practice my work is a mixture of simulation, data analysis, algorithm design, software development, problem-solving, and whatever random engineering no one else is doing. I am married and have a daughter born in mid-2006. The best way to contact me is by email: firstname dot lastname at pobox dot com. I am happy to be emailed out of the blue by interesting people. If you are an LW regular you are probably an interesting person in the relevant sense even if you think you aren’t.

If you’re wondering why some of my very old posts and comments are at surprisingly negative scores, it’s because for some time I was the favourite target of old-LW’s resident neoreactionary troll, sockpuppeteer and mass-downvoter.

gjm 8 Sep 2025 13:12 UTC
2 points
0
in reply to: jacob_cannell’s comment on: Hyperbolic model fits METR capabilities estimate worse than exponential model
I agree. (Or perhaps we should plot in log space when measuring/minimizing errors in log space, and plot in linear space when measuring/minimizing errors in linear space. The former is nearer to the Right Thing here than the latter.)
There are several things I find unsatisfactory in my own analysis here and either I will write a followup or Valentin2026 and I will write one together. (We’re still discussing.)

gjm 4 Sep 2025 20:44 UTC
2 points
0
on: Models vs beliefs
Related: Philosopher Peter van Inwagen on the way that one philosopher can disagree with another while accepting that the other guy is smarter and better informed. (Which he points out both because he finds it interesting in itself, and to argue that there’s something inconsistent in complaining about religious believers—of whom he is one—believing things when the balance of the evidence is against them.)

gjm 31 Aug 2025 22:15 UTC
2 points
3
in reply to: dr_s’s comment on: A quantum equivalent to Bayes’ rule
Fair enough! I think the article would be improved by making this a bit more explicit somehow.

gjm 31 Aug 2025 15:53 UTC
4 points
0
on: A quantum equivalent to Bayes’ rule
The title advertises a quantum version of Bayes’ rule, but so far as I can tell the actual post never explicitly presents one. Am I missing something?

gjm 27 Aug 2025 21:18 UTC
10 points
1
in reply to: cousin_it’s comment on: Banning Said Achmiz (and broader thoughts on moderation)
FWIW my personal intention—only time will tell whether I actually stick to it—is to be a little more vigorous in disagreeing with things that I think likely to be anti-rational, precisely because Said will no longer be doing it.

gjm 25 Aug 2025 0:00 UTC
1 point
11
in reply to: Said Achmiz’s comment on: Banning Said Achmiz (and broader thoughts on moderation)
Well, right now my comment saying what I think “top author” means to most LW readers is on +12/+4 while yours saying what you think it means to most readers is on −18/-10. LW karma is a pretty poor measure of quality, but it does give some indication of what LW readers think, no?
And no, it does not clearly mean “person whom Oliver Habryka likes”. You can get it to mean that if you assume that all subjective evaluations collapse into “liking”. I do not make that assumption, and I don’t think you should either.

gjm 24 Aug 2025 21:28 UTC
15 points
5
in reply to: Said Achmiz’s comment on: Banning Said Achmiz (and broader thoughts on moderation)
I took “top author” to mean something like “person whose writing’s overall influence on LW has been one of the most positive”. I would not expect that to be equivalent to anything mechanically quantifiable (e.g., any combination of karma, upvotes, number of links, number of comments, proportion of replies classified as positive-sentiment by an LLM, etc.), though I would expect various quantifiable things to correlate quite well with it. I would not take it to mean “person whom Oliver Habryka likes” but I would expect that Oliver’s judgement of who is and isn’t a “top author” to be somewhat opaque and not to come down to some clear-cut precisely-stated criterion. I would not expect it to mean something objective; I would expect it to be somewhat intersubjective, in that I would e.g. expect a lot of commonality between different LW participants’ assessment of who is and who isn’t a “top author”.
There is a lot of space between “completely meaningless, nothing but vibes, just Oliver’s opinion” and “answerable by looking at some sort of data somewhere”. I would take “top author” to live somewhere in that space, and my guess (for which I have no concrete evidence to offer, any more than you apparently do for what you are “quite sure most readers would assume”) is that the majority of LW readers would broadly agree with me about this.

gjm 23 Aug 2025 16:58 UTC
10 points
1
in reply to: sunwillrise’s comment on: Banning Said Achmiz (and broader thoughts on moderation)
Let me join the chorus: please do not leave in protest; your comments here do some of the same positive things that Said’s comments do, and your leaving would have a bunch of the negative consequences of Said’s banning without the positive ones (because, at least so it seems to me, you are much less annoying than Said).
(For the avoidance of doubt, I find you a net-positive commenter here for reasons other than that you do some of the useful things Said has done, but that particular aspect seems the most relevant on this occasion.)

gjm 21 Aug 2025 14:35 UTC
5 points
0
on: Critiques of FDT Often Stem From Confusion About Newcomblike Problems
In the “Bomb” scenario, suppose we delete the words “by running a simulation of you and seeing what the simulation did” and replace them with something like “by carefully analysing your brain in order to deduce what you would do”.
I am not an FDT expert and so maybe I’m missing something, but it seems to me that FDT still says you should pick Left (because, a priori, if your algorithm picks Right then the predictor will put the bomb in Left and you’ll have to pay, whereas if your algorithm picks Left then the predictor will leave the bomb out and you won’t, and the latter is the better outcome).
But in this version of the “Bomb” scenario, you have no particular reason to think that in addition to the real-world you making the choice there’s also a faithfully-simulated you that you-right-now might equally plausibly turn out to be. You might perhaps worry that that might be how the predictor is doing her analysis, but unless you think she almost certainly ran a faithful simulation it seems obvious that you do better to pick Right contrary to FDT’s recommendation.
(Once again, I’m not an expert and there’s a good chance I’m missing something important. If so, I hope someone will enlighten me.)

gjm 20 Aug 2025 11:27 UTC
4 points
0
in reply to: dr_s’s comment on: Hyperbolic model fits METR capabilities estimate worse than exponential model
For sure nothing in the real world can really grow exponentially for ever. I don’t know how consistently the failure to grow exponentially for ever looks like a sigmoid rather than, say, an exponential that abruptly runs into a wall, or an exponential that gets “abandoned” before it turns sigmoid because some other thing comes along to take its place.
I’ll tweak the wording in the OP to be clearer about this. [EDITED to add:] Now done.

gjm 19 Aug 2025 21:41 UTC
5 points
0
in reply to: sunwillrise’s comment on: Hyperbolic model fits METR capabilities estimate worse than exponential model
Thanks for the kind words!

gjm 19 Aug 2025 20:26 UTC
6 points
0
in reply to: Valentin2026’s comment on: Hyperbolic trend with upcoming singularity fits METR capabilities estimates.
Thanks for checking, confirming, and editing accordingly. I reiterate that despite my criticisms of the content, this is what epistemic virtue looks like: write things in such a way that others can check them, and correct appropriately if they turn out to have been wrong.
(Strong-downvote undone since with the corrections at the top the OP is no longer likely to mislead anyone much.)

gjm 19 Aug 2025 16:16 UTC
5 points
3
in reply to: NickH’s comment on: Church Planting: When Venture Capital Finds Jesus
Even if this is 100% correct it’s also close to 100% irrelevant to this post.

gjm 19 Aug 2025 16:00 UTC
4 points
1
in reply to: ryan_greenblatt’s comment on: Hyperbolic trend with upcoming singularity fits METR capabilities estimates.
It’s in linear space for the hyperbolic model and in log space for the exponential one, which unfortunately means all the comparisons are invalid. See my comment (and the longer post linked from it).

gjm 19 Aug 2025 13:51 UTC
17 points
9
on: Hyperbolic trend with upcoming singularity fits METR capabilities estimates.
The code in your notebook does its hyperbolic fitting by (nonlinear) least squares on the original data, but its exponential fitting by (linear) least squares on the log of the original data.
This means (1) that your exponential fit will look worse “on the right” where the numbers are larger than an actual least-squares fit of an exponential, and (2) it will have worse mean squared error (on the original data) than an actual least-squares fit of an exponential.
When I do an exponential fit using the same machinery as you used for your hyperbolic fit, I get this graph
and a “total error” of 0.107 versus 0.141 for the q=2 hyperbolic fit.
[EDITED to add:] The specific exponential I get, in case anyone wants to check, is horizon = exp b (t-t0) where b = 1.8421 and t0 is 2025-02-27.
[EDITED to add:] I’ve strong-downvoted the post, not because I think there’s anything very terrible about making the mistake I believe you made, but because at first glance it looks like evidence for something important and I think it isn’t in fact evidence for that important thing and I don’t want people to be misled.
[EDITED again to add:] Full post at https://www.lesswrong.com/posts/ZEuDH2W3XdRaTwpjD/hyperbolic-model-fits-metr-capabilities-estimate-worse-than

gjm 18 Aug 2025 12:36 UTC
13 points
7
on: Agent foundations: not really math, not really science
When Alan Turing figured out computability theory, he was not doing pure math for math’s sake; he was trying to grok the nature of computation so that we could actually build better computers.
Are you sure this is true? I am not a Turing expert but my impression is that his practical work on building computing machinery was downstream of his theoretical work on the nature of computation, not vice versa.

gjm 13 Aug 2025 15:28 UTC
3 points
0
in reply to: Noah Birnbaum’s comment on: Mech Interp Wiki Page and Why You Should Edit Wikipedia
I’m glad that the post’s first paragraph is now more consistent with the rest of it, but it does feel like it might need more on the old #3 to avoid the failure mode where a bunch of enthusiastic LessWrongers descend on Wikipedia, edit it in ways that existing Wikipedia editors disapprove of, and get blowback that ends up making things worse. At the very least it seems like there should be links to specific sets of rules that it would be easy to fall foul of, things like WP:COI and WP:NOTE and WP:RS.

gjm 13 Aug 2025 14:33 UTC
3 points
1
on: Mech Interp Wiki Page and Why You Should Edit Wikipedia
The opening paragraph says
Below, we (1) flag the most important upgrades the Mech Interp page still needs, (2) list other alignment / EA topics that are currently stubs or red-links, and (3) share quick heuristics for editing without tripping Wikipedia’s notability or neutrality rules.
but so far as I can see the text that actually follows doesn’t do (1) at all, does kinda do (2) though e.g. I don’t think it’s accurate to describe Wikipedia’s “Less Wrong” article as a stub, and does give advice on editing Wikipedia but doesn’t do the specific things claimed under (3) at all.
I think (3) is pretty important here. If a lot of people reading this go off to edit Wikipedia to make it say more about AI alignment, or make it say nicer things about Less Wrong, or the like, then (a) many of those people may be violating Wikipedia’s rules designed to avoid distortion from edits-by-interested-parties and (b) even if they aren’t it may well look that way, and in either case the end result could be the opposite of what those new Wikipedia editors are hoping for.

gjm 5 Aug 2025 23:06 UTC
6 points
0
in reply to: TAG’s comment on: Two Dogmas of LessWrong
You don’t need to be a moral realist to be swayed by arguments from logical consistency. If I say “I regard animals as moral patients and therefore won’t eat them, but I do eat fish and chicken” then pointing out my inconsistency has much the same force whether I’m a moral realist who thinks “animals have moral importance” is a fact about the universe or a moral nonrealist for whom it’s a fact about my values.
Arguments that actually work are the kind that are based on widely accepted principles.
What matters isn’t whether a principle is widely accepted, it’s whether it’s accepted by the person you’re arguing with. And, again, what matters is whether you can find such principles to base your argument on, not whether the person you’re arguing with thinks those principles are matters of objective fact.
(“You shouldn’t eat that because it’s got lots of beetroot in it and you hate beetroot” is a strong argument even if the person you’re making it at isn’t a gastronomic realist. Similarly, “You shouldn’t do that because it would mean X and you regard X as morally bad” is a strong argument even if the person you’re making it at isn’t a moral realist.)
You can be a utilitarian and a realist, as well.
Of course! I thought that was obvious and I wasn’t in the least intending to suggest otherwise.
There’s very little difference in that distinction [sc. between dismissing something and disagreeing with it].
I completely disagree. Dismissing something is justified much less often than disagreeing with it; it’s often true to say “you can’t just dismiss X” but false to say “you can’t disagree with X”.
There can be truths about human constructs, e.g. Engineering, Economics.
Perhaps I misunderstood you, before. I thought you were talking about constructing an ethical system. The equivalent of that in engineering would be constructing a set of physical laws, but that’s not what anyone does when determining truths about human constructs in the realm of engineering; when an engineer says “if you build the bridge that way then it will fall down”, they mean it will fall down because that’s how the universe actually works, not only that we have invented a system of physical laws according to which the bridge will fall down.
Lack of universalism does not imply lack of realism.
Fair enough. But what you wrote does seem to me flagrantly non-realist, not merely non-universal.
It is not nihilism, or anything-goes subjectivism either.
Indeed it isn’t. But I thought you were making a claim stronger than “nihilism is bad”, and saying specifically that moral realism is necessary in order for there to be any point in engaging in moral arguments. Maybe you weren’t. You’ve talked about “relativism” and “subjectivism” and (non)”realism” and not been terribly clear about what distinctions, if any, you make between those terms; perhaps all along you were saying only “nihilism is bad” and this whole argument has been a waste of time.
If you think there is an actual disagreement here, then I beg you to do the thing I asked you to three years ago upthread: describe an actual concrete situation in which some moral argument should move a moral realist but there’s no closely-equivalent argument that should have the same effect on someone who has the same moral values but isn’t realist about them.
I think that if a moral argument moves someone it’s by appealing to specific moral values they have, and that this will work equally well whether they are realist about those values or not.

gjm 5 Aug 2025 22:31 UTC
2 points
0
in reply to: Said Achmiz’s comment on: Several questions about Zen koans
I took “demonstrate that it is as you say, and not as I say” to be asking for an explanation, especially given that you consider responses like slapping you in the face to be irrelevant. But perhaps I misunderstood, and for sure you didn’t use the word “explanation”. What other sort of demonstration did you have in mind?
As for the face-slapping, I think it’s possible that in this situation a Zen teacher would slap you by way of showing that the answer is X not Y, and possible that in this situation a Zen teacher would slap you to encourage you to stop asking (what he regards as) unproductive questions, and possible that in this situation a Zen teacher would slap you just because he feels like slapping you and the Zen tradition gives him some cover to do it.
And my point was simply that that first possibility seems like a real one, which means that “he might slap you in the face” is not quite like “he might tell you to stop interrupting his lecture”.