“Charitable purpose” can be pretty broad under U.S. law, for instance, and could probably encompass such funding. Or do you mean that funders are not interested in it?
Midtermist12
Karma: 340
A lot of the specific things you’ve mentioned make a lot of sense.
Generally, I would be cautious about asking for help in contexts where review and/or supervision will be necessary or your reliance on the volunteer could be detrimental. People are often excited to help and overstate what they actually can do. Often the value of the time involved with dealing with volunteers is far more valuable than what they produce.
Another virtue of the houses as a way of doing it is would be a pretty strong signal that someone using the resource is in it for the right reasons. There are a lot more constraints than someone just receiving a check and hoping that the person is trying to make an impact.
Teaching counterfactual reasoning in economics education
A crucial EA concept for high school economics is counterfactual reasoning – systematically asking “what would have happened if agent X had not done action Y?” This is essential for understanding the actual impact of interventions.
Why it matters:
Many interventions don’t create as much value as they appear because something similar would have happened anyway
The true impact is only the additional change caused by the intervention – the difference between what actually happened and what would have happened in the counterfactual scenario
It’s counterintuitive – our brains naturally credit actions without considering what would have occurred otherwise
Methods to evaluate counterfactual impact:
Randomized controlled trials (RCTs): Randomly assign some groups to receive an intervention and others not, then compare outcomes. The control group approximates what would have happened without the intervention.
Before-and-after with comparison groups: Compare changes in a treated group to changes in a similar untreated group over the same period. This helps account for broader trends that would have occurred anyway.
Trend analysis: Plot pre-intervention trends and project them forward. If post-intervention outcomes match the projected trend, the intervention may have had little counterfactual impact.
Natural experiments: Find situations where an intervention occurred in one place but not another similar place due to arbitrary reasons, allowing comparison.
Classroom applications:
Analyze case studies using these methods (e.g., evaluating a job training program’s effectiveness)
Have students design simple evaluation plans for school or community interventions
Critique news articles that claim causation without proper counterfactual analysis
This teaches students both to think counterfactually and to evaluate causal claims empirically.
(Comment made in collaboration with generative AI)
I have thought for a long time that the EA power centers have a lack of curiosity or appropriate respect for the potential of the EA community and has a pretty specific set of markers that suggest that people should be given the time of day. I have the impression that there is a pretty small “nerve center” that sets priorities and considers how the EA community might be helpful to address these priorities along the paths that it sees.
This seems to me to limit the power of EA significantly: if we could be getting more perspectives and ideas taken seriously and dedicating resources in these direction, rather than just to relatively narrow agents of the “nerve center”, we might be able to accomplish a lot more. Right now it seems pretty sad that EA is often just identified with the people who have power in it, rather than its more basic and important idea of maximizing good with the resources that we have.
I suppose there have probably been posts along these lines, but I guess a “democratizing EA funding and power as maximizing epistemic hygiene and reach” would be appreciated.
Yeah, I understand the need for credibility with the animal rights community, but it probably would be helpful if there were more prominent omnivores who emphatically identified as animal advocates. Probably one of the reasons factory farming can be so successful is that there’s a perceived barrier to entry to fighting it as becoming vegan. The more that vegans reinforce the narrative that “to be on our side, you need to be vegan”, the more they are alienating potential allies and making it easier for the monstrous system to persist. I think what might be the most important in broadening the movement would be prominent animal rights activists who are omnivores.
I don’t really understand why I am getting downvoted/disagreevoted. I was just pointing out the contrast between humans—who also have the traits discussed in this post—and AI, which is basically that human beings have intrinsic motivations and virtues that AI does not have. I thought that this was a critical piece that was not really emphasized. It is pretty dispiriting to read through an article, point out something you think might be helpful, and have this happen.
Thank you for this detailed analysis. I found the human analogy initially confusing, but I think there’s an important argument here that could be made more explicit.
The essay documents extensive DSL (deceitful, sycophantic, lazy) behavior in humans, which initially seems to undermine the claim that AI will be different. However, you do address why humans accomplish difficult things despite these traits:
“People do not appreciate how much of science relies on the majority of people practicing it to be rigorous, skilled, and ideologically dedicated to truth-seeking.”
If I understand correctly, your core argument is:
Humans have DSL traits BUT also possess intrinsic motivation, professional pride, and ideological commitment to truth/excellence that counterbalances these tendencies. AI systems, trained purely through reward optimization, will lack these counterbalancing motivations and therefore DSL traits will dominate more completely.
This is actually quite a sophisticated claim about the limits of instrumental training, but it’s somewhat buried in the “vibe physics” section. Making it more prominent might strengthen the essay, as it directly addresses the apparent paradox of “if humans are DSL, why do they succeed?”
Does this capture your argument correctly?
(note the above comment was generated with the assistance of AI)
Thanks for sharing this. While I think there are strong reasons to invest heavily in AI safety, I’m concerned this particular cost-benefit framing may not be as compelling as it initially appears.
The paper uses a $10 million value of statistical life (VSL) to justify spending $100,000 per person to avoid a 1% mortality risk. However, if we’re being consistent with cost-effectiveness reasoning, we should note that GiveWell-recommended charities save lives in the developing world for approximately $5,000 each—roughly 2,000 times cheaper per life saved.
By this logic, the same funding directed toward global health interventions would save orders of magnitude more lives with near-certainty, compared to reducing AI x-risk with uncertain probability.
This doesn’t mean AI safety is a bad investment—there are strong arguments based on:
The value of preserving future generations (which Jones notes would increase spending estimates)
Diminishing returns or bottlenecks in scaling proven global health interventions
The categorical importance of preventing existential catastrophe
Portfolio diversification across different types of risk
(Note: comment generated in collaboration with AI)
As someone who endorses offsetting (or donating to animal charities in excess of offset) as a form of being an ally to animals, would not being an omnivore who donates far in excess of the offset make you more credible regarding this position?
I don’t know that the contrast between climate and animal offsets is so strong. The harm caused by consuming animal products is also indirect in most cases in relation to the consumer: the animal consumed is already dead and has already suffered whatever harms the factory farming system inflicted on it, so your action harms it no further. The harm you are actually doing are increasing the demand for the product.
The set of people eating meat that choose to offset and those that choose not to probably have a very different psychological environment regarding the animals consumed, I would think.
I haven’t read more than the abstract of the referenced article, as access is restricted, but the abstract suggests that the reason that people deny the existence of animal minds is due to their dissonance between not wanting to cause harm to things that have minds and enjoying eating meat. Imposing offsets on oneself would seem to reconcile this dissonance: you enjoy an activity that causes harm and because you have this aversion to harm you couple that with an activity that has commensurate harm reduction.
Although it is possible that meat consumption with offsets could have a corrosive effect on one’s perception of animals as moral patients, I would not infer this from the linked article, based on the psychological mechanisms described in the abstract.
I think that the use of an LLM here embodies what they are prioritizing: speed and results-orientation.
The blog post pattern-matches to AI-speak, but clearly communicates what they are looking for. If anything, I would update positively for the prudent use of AI here.
In the case of the author with the history of fraud, you are applying prejudice, albeit perhaps appropriately so.
You raise stronger points than I’ve yet heard on this subject, though I still think that if you read some kind of content and find it compelling on its merits, there is still a strong case to apply at least similar scrutiny regardless of whether there are signs of AI use. Although I still think there is too much knee-jerk sentiment on the matter, you’ve given me something to think about.
I’ve made a pretty clear distinction here that you seem to be eliding:
Identifying AI content and deciding on that basis it’s not worth your time
Identifying AI content and judging that content differently simply because it is AI generated (where that judgment has consequences)
The first is a reasonable way to protect your time based on a reliable proxy for quality. The second is unfair and poisoning of the epistemic commons.
See my response to titotal.
Identifying signs of AI and using this as a reason not to spend further time assessing is rational for the reasons you and titotal state. But such identification should not effect one’s evaluations of content (allocating karma, up voting, or more extremely, taking moderation actions) except insofar as it otherwise actually lowers the quality of the content.
If AI as source effects your evaluation process (in assessing the content itself, not in deciding whether to spend time on it) this is essentially pure prejudice. It’s similar to the difference between cops incorporating crime statistics in choosing whether to investigate a young black male for homicide and a judge deciding to lower the standard of proof on that basis. Prejudice in the ultimate evaluation process is simply unjust and erodes the epistemic commons.
Thanks for sharing these other posts.
We have fairly different beliefs regarding replaceability of staff by orgs with funding (depending, of course on the scarcity of the labor supply), but you can certainly apply the framework this post endorses with a wide range of discounting job impact due to replaceability.
Wish I could take total credit for that turn of the phrase, but that was courtesy of Claude.
I agree that using AI flavor as a means to determine whether content is worth further consideration makes sense. If AI flavor is a strong proxy for garbage, than it makes sense to consider stop reading after you detect AI flavor
What does not make sense is, after you have decided to read and evaluate the content, to designate it inferior simply because of the process that generated it. This is essentially pure prejudice.
An example would be: I evaluated X content and determined Y deficiencies. Because I believe the content was entirely human-generated, I assess these deficiencies as relatively minor. Alternatively, I evaluated A content and determined B deficiencies which are substantially similar to Y, but because I suspect that the content was AI generated, I place greater weight on said deficiencies.
It’s interesting to think about the potential upsides of AGI from the perspective of people who struggle with suicidal thoughts. It seems like there are significant chances of an extremely long, happy future that probably is not balanced by the S-risk (it seems more likely misaligned AGI would annihilate us than perpetually torture us).
This has made suicidal thoughts much more compelling in the past than after recent developments. Thinking about losing the chance of an unimaginably good future (even just like 5-10%) chance that could be missed forecloses thoughts of further consideration of methods by which it could be achieved.
Maybe disseminating this line of thinking could be helpful for suicide prevention?