lc

Karma: 10,729

lc Aug 19, 2025, 6:13 AM
11 points
8
in reply to: cousin_it’s comment on: Underdog bias rules everything around me

I don’t think it’s due to evolution or material conditions. I think it’s cultural and goes back to the rise of Christianity...

I think basically every time someone has a story like this it’s wrong. I don’t understand why people seem so eager to blame cultural forces for ubiquitous behavior in this fashion. I guess it makes humans seem more interesting.

lc Aug 17, 2025, 7:54 AM
3 points
0
in reply to: O O’s comment on: METR Research Update: Algorithmic vs. Holistic Evaluation

I think the extra effort required to go from algorithmically to holistically qualifying scales linearly with task difficulty. Dense reward model scaling on hard to verify tasks seems to have cracked this. Deepminds polished holistically passing IMO solutions probably required the same order of magnitude of compute/effort as the technically correct but less polished OpenAI IMO solutions. (They used similar levels of models, compute, and time to get their respective results)

Wondering why you think this

lc Aug 14, 2025, 10:38 PM
5 points
0
in reply to: tdko’s comment on: METR Research Update: Algorithmic vs. Holistic Evaluation
Yup, this is one of the largest problems with using AI solutions today.

lc Aug 11, 2025, 3:58 AM
15 points
8
on: Having children is not the most effective way to improve the world. Have them because you want them, not “for impact”.
If you’re worried about “idiocracies” and think your genes would make the difference, donate your sperm/eggs, and call it a day.
Sperm donation in particular doesn’t produce a new child; it displaces another donor, because there are many more people who apply to donate than get accepted. And the donor you displace is already likely to be above average in intelligence, health, and professional success, because those are the people who pass screening and get selected by prospects.

lc Aug 10, 2025, 4:19 AM
3 points
0
in reply to: Lao Mein’s comment on: Lao Mein’s Shortform
I don’t know what the screenshot you posted in the OP is supposed to be of, or where it came from, so I have no idea what there might be to explain. Is there evidence that OpenAI is using this tokenizer in GPT-5?

lc Aug 10, 2025, 12:07 AM
14 points
1
in reply to: Lao Mein’s comment on: Lao Mein’s Shortform
o200k_base looks to be some shared tokenizer, not a base model. Please don’t bring Twitter epistemic standards to LessWrong.

lc Aug 4, 2025, 12:33 PM
10 points
2
in reply to: Daniel Kokotajlo’s comment on: Saying Goodbye
Can anyone point to any prominent EA promoting a cryptocurrency, full stop? I can’t ever remember this happening.

lc Aug 4, 2025, 12:05 PM
21 points
4
on: Saying Goodbye
A sane response would be to slow down the race and build a trustworthy international framework that ensures everyone can benefit from AGI. Promises would not be enough; you would need actual institutions with real authority. That’s hard, but possible. Instead, we’ve chosen a strategy of cutting China off from GPUs and hoping we can beat them to AGI before they scale up domestic production.
Isn’t this… The entire “Rat/EA” platform?

lc Aug 3, 2025, 8:48 PM
5 points
3
in reply to: Seth Herd’s comment on: lc’s Shortform
I mean I agree, but I just think this is insufficient worldview hedging. What if AGI takes another sixty years?

lc Aug 3, 2025, 12:22 PM
8 points
−3
on: lc’s Shortform
Something that “feels” worth doing to me, even if timelines eventually make it irrelevant, would be starting an independent org to research/verify the claims of embryo selection companies. I think by default there’s going to be a lot of bullshit, and people are going to use that bullshit as an excuse to call for regulation or a shutdown. An independent org might also encourage people who were otherwise skeptical of this technology to use it.

lc Aug 2, 2025, 12:45 PM
9 points
2
in reply to: Stephen Martin’s comment on: lc’s Shortform
IQ doesn’t have a heritability of “1”. If two geniuses have a child, the expected IQ of the child isn’t the midpoint between the mother and father’s IQ. It’s lower.

lc Aug 2, 2025, 1:07 AM
8 points
7
on: lc’s Shortform
The calculators on the herasight website’s landing page are… Quite inaccurate:

lc Jul 24, 2025, 1:22 AM
27 points
26
on: lc’s Shortform
Much like how all crashes involving self-driving cars get widely publicized, regardless of rarity, for a while people will probably overhype instances of AIs destroying production databases or mismanaging accounting, even after those catastrophies become less common than human mistakes.

lc Jul 22, 2025, 12:29 PM
12 points
19
in reply to: JuliaHP’s comment on: JuliaHP’s Shortform
Sounds like you’re depressed.

lc Jul 21, 2025, 9:02 PM
5 points
0
in reply to: Mo Putera’s comment on: Recent AI model progress feels mostly like bullshit
The XBOW PR is a quintessential example of what I’m talking about. Suffice it to mention that:
- XBOW topped the HackerOne ‘leaderboard’ that measures upvotes, not money earned on the platform.
- XBOW almost entirely submitted bugs for the free, non-paid bug bounties! A primary reason they were able to find these bugs was because they weren’t actually competing with anyone!

lc Jul 19, 2025, 10:22 PM
3 points
0
in reply to: habryka’s comment on: A non-magical explanation of Jeffrey Epstein
Two years after the post was written! Although, still, good on Lukas.

lc Jul 19, 2025, 10:18 PM
2 points
0
in reply to: Mitchell_Porter’s comment on: A non-magical explanation of Jeffrey Epstein
None of those factoids are much evidence of anything, but I suppose I should be more specific: Epstein’s intelligence work (if he conducted any) probably had nothing to do with the lenient sentence Epstein received in 2008, which is what the OP claims.

lc Jul 19, 2025, 4:45 PM
22 points
13
on: A non-magical explanation of Jeffrey Epstein
A few years later, I basically no longer endorse a large section of this post, and I think it should have gotten a much more critical reception than it did.
The key things you need to understand about the Epstein case are:
- Epstein almost certainly never trafficked any underage victims to associates, and almost none of his victims have ever claimed this. Pretty much all of the salacious coverage about this case stems from Virginia Guiffre’s accusations, and in the case of Alan Dershowitz these were so effectively refuted that she ended up withdrawing and then recanting one of the only two civil suits she actually brought.
- Epstein killed himself, probably without the assistance of coconspirators (who as I mentioned earlier don’t exist, at least in the case of his sex crimes).
- I was probably wrong and overexcited about possible “intelligence connections”. People have since casted doubt about the “belonged to intelligence” quote from Acosta by pointing out the fact that only once source, the Daily Wire, has ever reported it. I also did not know this at the time, but for his part Acosta himself denied that this reporting was accurate to OPR in their followup investigations in 2020. And there would be no reason for D.A. Acosta to lie about this—on the contrary, he would have screamed about it at the top of his lungs to everyone who accused him of misconduct, if Epstein being a cooperator was actually something that factored into it.
I think some part of me knew that the “Epstein’s intelligence connections got him out of his first conviction” was a ridiculous thing to believe at the time, but wanted to make the post more interesting and so put it in anyways. I apologize for letting my excitement get the better of me that way.
I still endorse everything in the addendum; Epstein having coordinated with or manipulated correctional staff, on his own, to give himself the opportunity to commit suicide is the most plausible theory I’ve heard for the suspicious circumstances surrounding his death. And of course don’t take the organizational chart literally is good advice.

lc Jul 15, 2025, 6:44 AM
−4 points
−5
in reply to: Stephen Martin’s comment on: Is the political right becoming actively, explicitly antisemitic?
https://en.wikipedia.org/wiki/Antisemitism

lc Jul 15, 2025, 6:41 AM
2 points
−1
in reply to: Shankar Sivarajan’s comment on: Is the political right becoming actively, explicitly antisemitic?

The Groypers remain a fringe group of internet trolls without any real influence

I really disagree, I think they’re already starting to become very hard to ignore. Candace Owens gave Nick Fuentes a two hour interview literally just last week.