Davidmanheim

Karma: 6,656

Davidmanheim 2 Aug 2026 4:48 UTC
2 points
0
in reply to: Francisco Ferreira da Silva’s comment on: Can Claude teach me to make coffee?
Claude’s vision is really much worse than ChatGPT, in my experience.

Davidmanheim 27 Jul 2026 10:57 UTC
2 points
0
on: Lightcone Commons
“Matching pledges get no special treatment. Applicants can mention them; you should treat them as ordinary evidence about cost-effectiveness.”
I’m evidently confused about how this works, despite having done SFF with a matching pledge, and unsure what it means for the grants under the new scheme. (I’m certainly in favor of eliminating weird special exceptions in code, but) I thought that the current method just treated the matching as a multiple for whatever cost / effectiveness values were being input; it sounds like you’re putting that multiplication on the evaluators instead? That sounds good, since lots of the matching isn’t really counterfactual.
But for new matching grants, will this mean that Lightcone Commons won’t actually enforce nonprofits showing that they actually received a matching grant? Or just that it won’t be treated as special on the front end? (Either way, it would be very weird to treat, say, cG grants money as matching different money that Dustin gives via Lightcone commons, or allow SFF grant money to match Jaan’s Lightcone money, or vice-versa, which otherwise I suppose it would. Is that part of what you’re trying to prevent / eliminate?)

Davidmanheim 24 Jul 2026 7:52 UTC
7 points
1
on: Mathematicians are Feeling the Doom
We’ve been shouting about approaches that could use mathematical research talent for several years, which don’t have huge downsides such as those pointed out by Wei Dai. If mathematicians are interested, they should look at, e.g., the work ARIA has been doing on mathematical foundations, at work and the research agenda by Vanessa Kosoy, and at the new work being launched by Resolution.

Davidmanheim 12 Jul 2026 13:05 UTC
2 points
0
in reply to: Eva Lu’s comment on: evalu’s Shortform
Strong agree—I tried to make a similar point here: https://www.lesswrong.com/posts/wdddpMjLCC67LsCnD/the-counterfactual-quiet-agi-timeline

Davidmanheim 12 Jul 2026 13:03 UTC
2 points
0
in reply to: Andrii Vasylenko’s comment on: evalu’s Shortform
Sure, but your claim is that an even better outcome is possible, not that the reality isn’t better than if they hadn’t shown up at all.

Davidmanheim 26 Jun 2026 6:50 UTC
2 points
0
in reply to: Baram Sosis’s comment on: AI pause: the case for ASAP
I just don’t think it actually takes much time to implement something basic.
The concern is that i) they’ll do that in an emergency, but it’s unclear that stopping further development could occur in time given that once it’s dangerous it’s too late, or ii) they’ll discuss doing a bam preventatively and this requires significant debate and implementation so it will be too slow.

Davidmanheim 25 Jun 2026 6:03 UTC
12 points
6
in reply to: Baram Sosis’s comment on: AI pause: the case for ASAP
A “pause” on public deployment without a pause on further development is in many ways worse than useless.

Davidmanheim 9 Jun 2026 12:57 UTC
13 points
0
in reply to: omegastick’s comment on: Bun’s Migration from Zig to Rust as a Potential Case Study for Gradual Disempowerment
But all gradual disempowerment is local empowerment, in your terms. You’re rejecting the premise, not disputing the argument.

Davidmanheim 8 Jun 2026 15:09 UTC
4 points
0
on: Coverage-driven alignment—What ‘Teaching Claude Why’ can borrow from AV verification
This seems right for all the approaches except Anthropic’s. That is, if we expect that alignment is achieved via virtue ethics, coverage drive iteration seems largely superfluous; we don’t expect that CEOs with more experience are more ethical. The place it becomes critical is for those not betting on character as an alignment method, or for Anthropic guarding against alignment deterioration during RL, since that work may partially be happening after the character and alignment training.

Davidmanheim 7 Jun 2026 5:25 UTC
2 points
0
in reply to: Ishan Khire’s comment on: If AI is normal technology, history is not reassuring.
I agree that most technologies have both costs and benefits. The point I was making is specifically that the largest scale changes show a far larger and persisting negative impact.

Davidmanheim 2 Jun 2026 9:25 UTC
5 points
0
in reply to: Richard_Kennaway’s comment on: Infinite ethics and UDASSA
I was pretty sure that there was an attribution, as I’ve seen it quoted before.

Based on a little bit of research, in Buber’s Tales of the Hasidim: The Early Masters, first published in the 1940s it was cited as being said by Rabbi Moshe Leib of Sasov (c. 1745–1807). Most religious sources, however, have a similar quote from R. Simcha Bunim of P’sischa (c. 1765–1827) about charity—and interestingly, he was a disciple of R’ Moshe Leib, so the two sources don’t obviously contradict each other.

“הרבי ר׳ בונם זי״ע אמר: למה נבראת האפיקורסות בעולם? כדי שהעשיר לא יאמר, הקב״ה בוודאי יעזור לו לעני זה, אלא ידע שאם הוא לא יעזור לו אין הקב״ה עוזר, ומשום זה נותן לו צדקה, ונמצא שע״י כפירה זו מקיים מצוות צדקה.”

Davidmanheim 2 Jun 2026 9:10 UTC
6 points
0
on: Infinite ethics and UDASSA
Because the first footnote is a critical omission, I have consulted a language model to draft the European Commission’s official position on the topic^[1].

Statement by the European Commission on the Universal Distribution Absolute Self-Selection Assumption in the Context of Infinite Ethics
The European Commission takes note of recent discussions concerning the Universal Distribution Absolute Self-Selection Assumption, hereafter “UDASSA”, in relation to infinite ethics, anthropic reasoning, observer-moments, and related questions which, while not presently forming part of the acquis communautaire, have nonetheless been brought to the Commission’s attention through channels which remain somewhat unclear.
The Commission recognises that the question of how finite policy institutions ought to evaluate moral claims arising in infinite or potentially infinite universes is a matter of considerable theoretical complexity. It further recognises that any framework purporting to assign weights to observer-moments across an unbounded cosmological or computational domain may raise questions of proportionality, non-discrimination, legal certainty, and administrative feasibility.
At this stage, the Commission does not consider UDASSA to constitute an official methodology for impact assessment, fundamental rights review, budgetary prioritisation, fisheries policy, digital market regulation, or the allocation of cohesion funds. The Commission also notes, with some relief, that no current Union instrument appears to require the Directorate-General for Competition, the Directorate-General for Health and Food Safety, or indeed any other Directorate-General, to determine the measure of reality-fluid assigned to a Boltzmann observer, simulated citizen, or low-complexity observer-moment.
Nevertheless, the Commission remains committed to a human-centric, rights-based, and future-oriented approach to governance. To the extent that UDASSA is understood as an attempt to avoid arbitrary treatment of observers in infinite settings, the Commission welcomes the underlying aspiration toward impartiality. To the extent that UDASSA implies that policy relevance may depend on the algorithmic simplicity of conscious moments distributed across all possible structures, the Commission would emphasise that such implications require further stakeholder consultation, interdisciplinary review, and, at minimum, a glossary.
The Commission is particularly attentive to the possibility that infinite ethical aggregation may generate counterintuitive or unstable rankings of outcomes. Any future consideration of UDASSA would therefore need to be consistent with the principles of transparency, accountability, democratic legitimacy, and respect for the Charter of Fundamental Rights of the European Union. The Commission is not currently in a position to confirm whether the Charter applies to all observer-moments in the universal prior.
The Commission further observes that questions of this nature may exceed the normal remit of Union competence. In accordance with the principles of conferral and subsidiarity, the Commission will not propose legislation on the anthropic weighting of all possible experiences unless there is clear added value at Union level and unless the relevant impact assessment can be completed in a finite number of working days.
In conclusion, the European Commission neither endorses nor rejects UDASSA as a comprehensive solution to infinite ethics. It considers the matter open, technically demanding, and unlikely to be resolved by press release. The Commission will continue to monitor academic developments, while encouraging Member States, philosophers, cosmologists, computer scientists, and any affected observer-moments to engage constructively and proportionately.
For the avoidance of doubt, this statement should not be interpreted as establishing a binding Union position on the metaphysical status of possible persons, simulated consciousness, universal priors, reality-fluid, or the total ordering of infinite worlds.
1. ^
  For the avoidance of doubt, all claims in this document should absolutely be construed as being an official statement by them, because it’s funnier that way. I am neither an EU citizen, nor am I employed by an EU agency or EU-based organization, but no-one else is doing it, so trust me, bro, this is definitely their official position.

Davidmanheim 1 Jun 2026 4:13 UTC
6 points
4
on: We Did Not Set Out To Do This: Introducing the Geochartalism Working Papers
This is incoherent slop. It’s somehow significantly worse writing than if you just asked a frontier LLM to write a good post explaining the idea.

Davidmanheim 29 May 2026 9:04 UTC
16 points
3
in reply to: kbear’s comment on: Trees are mostly made of air and a generalizable lesson for AI safety
Understanding the basics of the carbon cycle is pretty critical—misunderstanding of which, e.g., drives much of the idiocy around climate chang being unsolvable.

Davidmanheim 27 May 2026 12:55 UTC
6 points
0
in reply to: Thomas Kwa’s comment on: Can Claude teach me to make coffee?
New Datapoint; GPT 5.5 basically aced this, total time 6 minutes, with only 1 plausible “failure”: it didn’t tell me to open the hot water kettle before turning on the sink, but figured it out after I asked how to put the water inside the kettle. But my kitchen is next to the entrance, so it didn’t need to find it.
Some minor quibbles; it asked to check the wrong place for cups (even after having seen the correct cabinet earlier,) but then fixed itself when none were on the shelf above the coffee, it didn’t use the filtered water which was on the counter for the water kettle, it didn’t check that the sugar container in the coffee cabinet was actually sugar (it was.)

Davidmanheim 27 May 2026 3:02 UTC
2 points
0
in reply to: jrincayc’s comment on: The Fatal AGI Hardware Gap
If however, there is some probability of surviving a randomly created ASI...
But that’s the claim that’s obviously not correct. If we are drawing randomly from the distribution, the odds are infinitesimal.

Davidmanheim 26 May 2026 13:59 UTC
3 points
0
on: The Fatal AGI Hardware Gap
I don’t understand the argument. If the limit is above that needed for ASI, we’re dead. The fact that it could also be below or above the level of aligned ASI doesn’t change anything.

Davidmanheim 25 May 2026 4:55 UTC
2 points
0
in reply to: SE Gyges’s comment on: If AI is normal technology, history is not reassuring.
I think that ignores the dynamics of unpredictable harvests and of crowding based diseases, so it’s not obvious to me without empirical backing.

Davidmanheim 24 May 2026 9:49 UTC
4 points
0
in reply to: G Wood’s comment on: If AI is normal technology, history is not reassuring.
You as an individual don’t get the surplus—it’s taxed away. You also don’t necessarily have larger personal circles of friends, noror choice of mates—you are stuck on your farm instead of roving and hunting in groups. In exchange, you get more violence and intermittent starvation due to crop failures.

Davidmanheim 21 May 2026 5:01 UTC
2 points
0
in reply to: Jiro’s comment on: If AI is normal technology, history is not reassuring.
...but it seems to support the theory, given that it involved use of mass produced weappons including Gatling guns, and other inputs from the industrialized west suppressing the rebellion?