Hide table of contents
This is a Draft Amnesty Week draft. It may not be polished, up to my usual standards, fully thought through, or fully fact-checked. 

Commenting and feedback guidelines: 

Keep one and delete the rest (or write your own):

  1. I'm posting this to get it out there. I'd love to see comments that take the ideas forward, but criticism of my argument won't be as useful at this time. 
  2. This draft lacks the polish of a full post, but the content is almost there. The kind of constructive feedback you would normally put on a Forum post is very welcome. 
  3. This is a Forum post that I wouldn't have posted without the nudge of Draft Amnesty Week. Fire away! (But be nice, as usual)

Epistemic status: musings and fuzzy ideas. There aren't any strong endorsements here, but rather explorations of a topic.

I occasionally wonder if we who care about EA are sometimes a bit too strict about applying principles. I mean that in the sense of cause neutrality and applying EA ideas/principles to areas that generally do not meet the bar for being EA. For example, I am guessing that a variety of interventions in low-income countries will be more cost effective (more bang for buck) than analogous interventions in the city of Cincinnati.[1] But if an organization has a mission that is limited to a helping people in Cincinnati, then they should probably find the most effective local maximum

I'm thinking of this as something like bounded effectiveness: it isn't cost competitive at a level of 10 times direct cash transfers, but within the boundaries that exist in this context, we want to be as effective as we can.

If John Doe runs a charity in Cincinnati and the donors want to keep the money local, he probably can't really do the "typical" EA stuff.[2] But he could certainly find the most cost-effective options within the restrictions. In a simplistic scenario, maybe the options are to knit hats for local homeless people to stay warm in the winter, or some kind of long-term housing assistance for local homeless.[3] If John Doe has good reason to believe that long-term housing assistance is more effective and efficient at the goal of helping the local homeless population in Cincinnati (it is his local maximum), then he should probably shift the programming away from knitting warm winter hats. John Doe can find the most effective way to use the resources for the benefit of a target population within Cincinnati, because that is the boundary he is working within.[4] At the same time, he is aware that he could use those resources to do much more good if he didn't have these restrictions. To me, it seems like this is still following EA principles of being cause neutral, having a scientific mindset, not being seduced too much by warm fuzzy feelings, but it is functioning within the boundaries/limits/restrictions of the scenario.

I think that to a certain extent we all do this,[5] and occasionally there are questions that come up on the EA forum about "If I want to donate money for this particular cause, what is the most cost-effective way to do it"? and these questions probably have the subtext of "I know this cause probably isn't cost-competitive with the top EA causes, and I am okay with that. I am looking for a local optimum, not a global optimum."

  1. ^

    I just chose a random city in the USA, but feel free to view this as a sort of stand in for any limit or restriction that a charitable organization has: geography, cause, target population, etc.

  2. ^

    Bednets, pandemic preparedness, alternative proteins, AI safety research, legislation, career advice for relatively wealthy and privilidge young people, etc.

  3. ^

    To be clear, this is very simplistic. I know practically nothing about homelessness interventions, so don't take the specific details seriously. This is only for an example.

  1. ^

    Or if you prefer something other than geography to be the restriction, think about orphanages in China: if you want to help cute babies, some organizations would probably be better to donate to than others. I don't know what research exists on this, but someone could certainly follow an EA-inspired path and figure out what the answer to "What is the most cost effective way to use my dollars to reduce suffering and cause happiness (within this particular scope that I am choosing to limit myself to)?" and then do it.

  2. ^

    In the broad sense, this strikes me as part of "personal fit." I interpret "personal fit" to partially mean "what would you be good at and happy doing," but also to mean "how flexible are you and what would you be willing to sacrifice." Maybe you really like animals in general and care a lot about chickens specifically, so you do the most effective things within that particular area/domain, even as you know that there might be other areas where your hours or dollars could have a bigger impact.

Show all footnotes
Comments1


Sorted by Click to highlight new comments since:

I think the way an EA would view this would still be in terms of the most utility-effective use of their time, however, the opportunity for leverage may significantly impact the calculation, and may enable cost-effective uses of time outside of typical cause areas.

For instance, there might be an EA endorsed charity for which marginal donations would generate utility at a rate of 10 utils/dollar. There might be an organization in the developed world that generates utility at an average rate of 1 util per dollar, and has an average annual budget of $10 million.

Suppose an EA sees an opportunity to dramatically increase the effectiveness of the non-EA charity, by about 50%, increasing its utility to 1.5 utils per dollar, and taking the EA about a full-time of work of its time. Alternatively, the EA could Earn to Give, generating a $120k salary, being able to donate $40k to the EA endorsed charity.

The EA working for non-EA charity and increasing its average utility/$ from 1 to 1.5 generates $5 million utility, assuming the same budget. On the other hand, by ETG and generating 40k to the EA charity, only 400k utility is generated.

In this circumstance, the EA generates far more utility by working for the non-Ea charity and rendering it more efficient than ETG for the EA charity. 

I like the idea of enabling different domains (which may not, themselves, be the most cost effective marginal recipient) to reach local maximums as a particularly effective opportunity. There may not be as many opportunities to increase the effectiveness of the most effective charities as there are some of the less effective charities that are still significant recipients of funds. 

One might say that it is better not to support such charities and let them die. This logic may be more applicable in the for-profit world, where failure to generate sufficient returns is often a death knell. However, survival in the nonprofit world can be more tied to being able to make donors happy than it is to demonstrate QALYs/dollar. 

Curated and popular this week
 ·  · 38m read
 · 
In recent months, the CEOs of leading AI companies have grown increasingly confident about rapid progress: * OpenAI's Sam Altman: Shifted from saying in November "the rate of progress continues" to declaring in January "we are now confident we know how to build AGI" * Anthropic's Dario Amodei: Stated in January "I'm more confident than I've ever been that we're close to powerful capabilities... in the next 2-3 years" * Google DeepMind's Demis Hassabis: Changed from "as soon as 10 years" in autumn to "probably three to five years away" by January. What explains the shift? Is it just hype? Or could we really have Artificial General Intelligence (AGI)[1] by 2028? In this article, I look at what's driven recent progress, estimate how far those drivers can continue, and explain why they're likely to continue for at least four more years. In particular, while in 2024 progress in LLM chatbots seemed to slow, a new approach started to work: teaching the models to reason using reinforcement learning. In just a year, this let them surpass human PhDs at answering difficult scientific reasoning questions, and achieve expert-level performance on one-hour coding tasks. We don't know how capable AGI will become, but extrapolating the recent rate of progress suggests that, by 2028, we could reach AI models with beyond-human reasoning abilities, expert-level knowledge in every domain, and that can autonomously complete multi-week projects, and progress would likely continue from there.  On this set of software engineering & computer use tasks, in 2020 AI was only able to do tasks that would typically take a human expert a couple of seconds. By 2024, that had risen to almost an hour. If the trend continues, by 2028 it'll reach several weeks.  No longer mere chatbots, these 'agent' models might soon satisfy many people's definitions of AGI — roughly, AI systems that match human performance at most knowledge work (see definition in footnote). This means that, while the compa
 ·  · 4m read
 · 
SUMMARY:  ALLFED is launching an emergency appeal on the EA Forum due to a serious funding shortfall. Without new support, ALLFED will be forced to cut half our budget in the coming months, drastically reducing our capacity to help build global food system resilience for catastrophic scenarios like nuclear winter, a severe pandemic, or infrastructure breakdown. ALLFED is seeking $800,000 over the course of 2025 to sustain its team, continue policy-relevant research, and move forward with pilot projects that could save lives in a catastrophe. As funding priorities shift toward AI safety, we believe resilient food solutions remain a highly cost-effective way to protect the future. If you’re able to support or share this appeal, please visit allfed.info/donate. Donate to ALLFED FULL ARTICLE: I (David Denkenberger) am writing alongside two of my team-mates, as ALLFED’s co-founder, to ask for your support. This is the first time in Alliance to Feed the Earth in Disaster’s (ALLFED’s) 8 year existence that we have reached out on the EA Forum with a direct funding appeal outside of Marginal Funding Week/our annual updates. I am doing so because ALLFED’s funding situation is serious, and because so much of ALLFED’s progress to date has been made possible through the support, feedback, and collaboration of the EA community.  Read our funding appeal At ALLFED, we are deeply grateful to all our supporters, including the Survival and Flourishing Fund, which has provided the majority of our funding for years. At the end of 2024, we learned we would be receiving far less support than expected due to a shift in SFF’s strategic priorities toward AI safety. Without additional funding, ALLFED will need to shrink. I believe the marginal cost effectiveness for improving the future and saving lives of resilience is competitive with AI Safety, even if timelines are short, because of potential AI-induced catastrophes. That is why we are asking people to donate to this emergency appeal
 ·  · 1m read
 · 
We’ve written a new report on the threat of AI-enabled coups.  I think this is a very serious risk – comparable in importance to AI takeover but much more neglected.  In fact, AI-enabled coups and AI takeover have pretty similar threat models. To see this, here’s a very basic threat model for AI takeover: 1. Humanity develops superhuman AI 2. Superhuman AI is misaligned and power-seeking 3. Superhuman AI seizes power for itself And now here’s a closely analogous threat model for AI-enabled coups: 1. Humanity develops superhuman AI 2. Superhuman AI is controlled by a small group 3. Superhuman AI seizes power for the small group While the report focuses on the risk that someone seizes power over a country, I think that similar dynamics could allow someone to take over the world. In fact, if someone wanted to take over the world, their best strategy might well be to first stage an AI-enabled coup in the United States (or whichever country leads on superhuman AI), and then go from there to world domination. A single person taking over the world would be really bad. I’ve previously argued that it might even be worse than AI takeover. [1] The concrete threat models for AI-enabled coups that we discuss largely translate like-for-like over to the risk of AI takeover.[2] Similarly, there’s a lot of overlap in the mitigations that help with AI-enabled coups and AI takeover risk — e.g. alignment audits to ensure no human has made AI secretly loyal to them, transparency about AI capabilities, monitoring AI activities for suspicious behaviour, and infosecurity to prevent insiders from tampering with training.  If the world won't slow down AI development based on AI takeover risk (e.g. because there’s isn’t strong evidence for misalignment), then advocating for a slow down based on the risk of AI-enabled coups might be more convincing and achieve many of the same goals.  I really want to encourage readers — especially those at labs or governments — to do something
Recent opportunities in Career choice
46
· · 3m read