Here to talk about phytomining for now.
Benjamin M.
Karma: 174
There’s probably something that I’m missing here, but:
Given that the dangerous AI capabilities are generally stated to emerge from general-purpose and agentic AI models, why don’t people try to shift AI investment into narrower AI systems? Or try to specifically regulate those systems?
Possible reasons:
This is harder than it sounds
General-purpose and agentic systems are inevitably going to outcompete other systems
People are trying to do this, and I just haven’t noticed, because I’m not really an AI person
Something else
Which is it?
Open Wallets, Empty Hearts [linkpost]
This is an understandable point to leave out, but one issue with the portfolio analogy is that, as far as I can tell, it assumes all “EA” money is basically the same. However, big donors might have advantages in certain areas, for instance if a project is hard to evaluate without extensive consultation with experts, or if a project can only be successful if it has a large and guaranteed funding stream. As such, I’m not sure it holds that, if somebody thinks Open Phil is underinvesting in longtermism compared to the ideal allocation, then they should give to longtermist charities- the opportunities available to Open Phil might be significantly stronger than the ones available to donors, especially ones who don’t have a technical background in the area.
I’ll try to write a longer comment later, but right now I’m uncertain but lean towards global health because of some combination of the following:
1. I suspect negative lives are either rare or nonexistent, which makes it harder to avoid logic-of-the-larder-type arguments2. I’m more uncertain about this, but I lean towards non-hedonic forms of consequentialism (RP parliament tool confirms that this generally lowers returns to animals)
3. Mostly based on the above, I think many moral weights for animals are too high
I’m also not sure if the 100 million would go to my preferred animal welfare causes or the EA community’s preferred animal welfare causes or maybe the average person’s preferred animal welfare causes. This matters less for my guesses about the impact of health and development funding.
Thank you for providing additional context!
The One Per Cent fund: Predecessors of Effective Giving
Both my state senator and my state representative have responded to say that they’ll take a look at it. It’s non-commital, but it still shows how easy it is to contact these people.
Do you like SB 1047, the California AI bill? Do you live outside the state of California? If you answered “yes” to both these questions, you can e-mail your state legislators and urge them to adopt a similar bill for your state. I’ve done this and am currently awaiting a response; it really wasn’t that difficult. All it takes is a few links to good news articles or opinions about the bill and a paragraph or two summarizing what it does and why you care about it. You don’t have to be an expert on every provision of the bill, nor do you have to have a group of people backing you. It’s not nothing, but at least for me it was a lot easier than it sounded like it would be. I’ll keep y’all updated on if I get a response.
Benjamin M.’s Quick takes
Thanks, that makes a lot of sense!
I’m a bit confused about how you get that children that just had severe malaria cases are a good proxy for lives saved with seasonal malaria chloroprevention.
Your page on SMC says that it works by reducing the number of malaria cases, and consequently reducing the number of deaths, rather than by making malaria cases milder.
Thus we’d expect that, for a child whose life was saved with SMC, the impact of receent malaria cases on their health would be nonexistent.
But this is implying that they have recently had a severe malaria case, which seems like it wouldn’t be true.
Am I misinterpreting something either here or on the SMC page? Or is it really being used as a proxy for sickliness in general? In that case, why are you only looking at estimates from children who suffered severe malaria cases, and not ones with other negative health events?
If I wrote this it would probably mostly be links/summaries/categorization of other people’s arguments against funding forecasting, plus maybe a few reasons of my own.
I want somebody to flesh out some of the negative comments on Open Philanthropy’s announcement about funding forecasting into an actual post.
I don’t have a background in forecasting or any insider knowledge of EA community dynamics, so I’m the wrong person to write this post but I might if nobody steps forward to claim it.
If you’re willing to consider literature, The Glass Bead Game by Hermann Hesse is the book that led me to EA ways of thinking, and also the best book, in my opinion, that I have ever read.
This looks very interesting!
One note: Friends Peace Teams has also been producing ceramic water filters, formerly in Indonesia and more recently in the Philippines, I believe. Unfortunately it’s not well documented on their website (I only found out about it through a talk that I went to). At that talk one of their members implied that they thought that they had a better production method based on training local people to make the filters using local materials in some way and then having them train others; I’m not really sure how this differs from other locally-produced water filter manufacturers but they implied that it was.
Link:
https://friendspeaceteams.org/wp-content/uploads/2018/06/Spring-2010.pdf
Alma “Kins” Aparece was the person who gave most of the talk and, if I remember correctly, helped facilitate the water filter making.
They very much don’t fulfill the idea of a charity focusing on one intervention (or maybe a few interventions), however; they do a wide variety of programs, most of which are focused on mediation and interpersonal training rather than clean water/other more tangible goods.
I’m not the first person to post this, but, if you’re an American, calling your senator or representative is probably a good idea. Here’s roughly how calls have gone when I do them:
Find the DC office number of your two senators and one representative and dial it
If it’s during weekday hours, a staffer will probably answer the phone. If it’s on the evening, holidays, or weekends, you’ll probably be talking to an answering machine
A staffer or an answering machine message will ask you to say who you are and what you’re calling for
I usually give my first name, zip code, and say that I’m calling to express my opinion on an issue or a piece of legislation. If you’re talking to an answering machine, you usually have to press a button about what you’re calling for (constituent services, expressing opinions, and some other stuff)
Then they’ll ask you to say your comment
I begin by briefly saying what PEPFAR is (a bill that funds efforts to fight AIDS around the world)
I mention that it needs reauthorization to continue being funded
I then give my main reason for supporting it (it’s a very effective way that the US saves lives cheaply around the world)
I might also give some side reasons, especially if there’s a person saying things like mm-hmm on the other end of the line (it’s boosted America’s reputation abroad, the program isn’t actually related to abortion, yada yada yada)
I say what I have to say in under a minute
I don’t have a personal or professional connection to the issue, but, if I did (i.e. doctor, nurse, AIDS patient, immigrant from a country with PEPFAR funds), I’d probably try to mention it
Some advice online says it’s good to make it more of a conversation and less of a speech, but that’s never worked well for me
You can adjust it a little for Democrats vs. Republicans as long as you’re saying true statements; usually I emphasize the boosting America’s reputation abroad with Republicans and maybe throw in a George W. Bush mention. But I’d definitely caution against saying things you don’t believe in to signal membership with one party or another
If there’s somebody on the line, they’ll say thanks and let you know that they’ll pass along your concern.
Occasionally they’ll ask a follow-up question (the only one I got about PEPFAR was whether this was a matter needing an individual response, but I’ve received actual questions about details when calling about other issues before).
Then I tell them thanks for listening and hang up.
This takes at most 2 minutes per call in my experience. I try to call my senators and reps about something (lately it’s usually PEPFAR) most weeks; I wouldn’t recommend going over once a week.
All this comes from a mix of reading some online articles, my own experience, talking with people who have been calling about other causes, and a bit of speculation.
Conclusion: Calling senators and representatives is easy and a good way to support PEPFAR reauthorization
Edit: found an earlier comment at https://forum.effectivealtruism.org/posts/ebGwTM2FAQcp8aMNH/francis-s-quick-takes#7cQuwprYi9AiAjjK4 that talks more about effectiveness.
This is a comment because it’s not actually a justification for EA elitism.
There are some okay-ish ways to quantify where students interested in Effective Altruism might end up. If we assume that, for a student to be interested in effective altruism, they need to have independently pursued some kind of extracurricular activity involving a skill of the kind that Effective Altruism might discuss, we can look at where the top competitors for those kinds of extracurriculars are.
One thing to beware is confounding factors. People who would be good for EA might be too busy to participate in these activities (either because they have busy class schedules or are involved in research or because they work outside of school). People might also be doing activities because they are superficially impressive, which probably isn’t a good sign for thinking in a very EA way.
Here are some brief summaries of where top competitors in different American extracurriculars come from:
Ethics Bowl (https://en.wikipedia.org/wiki/Intercollegiate_Ethics_Bowl)- no clear pattern among the universities
Bioethics Bowl (https://en.wikipedia.org/wiki/Bioethics_Bowl)- similar to above
National Debate Tournament (https://en.wikipedia.org/wiki/List_of_National_Debate_Tournament_winners)- often but by no means exclusively prestigious US schools, seems to lean towards private schools a bit also? but I’m just eyeballing it
US Universities Debating Championship (https://en.wikipedia.org/wiki/US_Universities_Debating_Championship)- Mostly Ivy-League or similarly prestigious schools
Putnam Exam (https://en.wikipedia.org/wiki/William_Lowell_Putnam_Mathematical_Competition)- Strongly dominated by MIT
College Model UN (https://bestdelegate.com/2022-2023-north-american-college-model-u-n-final-rankings-world-division/)- no clear pattern besides DC-based schools tending to do well
I’m sure other people can add more to this list.
If you think that Putnam results are a strong predictor of Effective Altruism, that could justify more elitism. Personally, I doubt that.
Thanks for pointing this out; I’ll note that Partners in Health is also available, and GiveWell seems to like them but doesn’t think that they beat the GiveWell charity bar, at least when this was written (https://www.givewell.org/international/charities/PIH#:~:text=Partners%20in%20Health%20provides%20comprehensive,network%20of%20community%20health%20workers.). I’d be interested in seeing anything about whether Partners in Health is a better option than GiveDirectly.
I’m not an expert on most of the evidence in this post, but I’m extremely suspicious of the claim that GPT-4 represents AI that is “~ human level at language”, unless you mean something by this that is very different from what most people would expect.
Technically, GPT-4 is superhuman at language because whatever task you are giving it is in English, and the median human’s English proficiency is roughly nil. But a more commonsense interpretation of this statement is that a prompt-engineered AI and a trained human can do the task roughly as well.
What you link to shows the results of how GPT-4 performs on a bunch of different exams. This doesn’t really show how language is used in the real world, especially since the exams very closely match past exams that were in the training data. It’s good at some of them, but also extremely bad at others (AP English Literature and Codeforces in particular), which is an issue if you’re making a claim that it’s roughly human level.
Furthermore, language isn’t just putting words together in the right order and with the right inflection. It also includes semantic information (what the actual meaning of the sentences is) and pragmatic information (is the language conveying what it is trying to convey, not just the literal meaning). I’m not sure whether pragmatics in particular would be relevant for AI risk, but the fact that anecdotally even GPT-4 is pretty bad at pragmatics prevents a literal interpretation of your statement.
In my opinion, the best evidence for GPT-4 not being human level at language is that, in the real world, GPT-4 is much cheaper than a human but consistently unable to outcompete humans. News organizations have a strong incentive to overhype GPT-caused automation, but the examples that they’ve found are mostly of people saying that either GPT-4 or GPT-3 (it’s not always clear which) did their job much worse than them, but good enough for clients. Take https://www.washingtonpost.com/technology/2023/06/02/ai-taking-jobs/ as a typical story.
Exams aren’t exactly the real world, but the popular example of GPT-4 doing well on exams is https://www.slowboring.com/p/chatgpt-goes-to-harvard. This both ignores citations (which is a very important part of college writing, and one that GPT-3 couldn’t do whatsoever and which GPT-4 still is significantly below what I would expect from a human) and relies on the false belief that Harvard is a hard school to do well at (grade inflation!)
I still agree with two big takeaways of your post, that an AI pause would be good and that we don’t necessarily need AGI for a good future, but that’s more because it’s robust to a lot of different beliefs about AI than because I agree with the evidence provided. Again, a lot of the evidence is stuff that I don’t feel particularly knowledgeable about, I picked this claim because I’ve had to think about it before and because it just feels false from my experience using GPT-4.
My rough sense is that one reason for EA’s historical lack of focus on systemic change is that it’s really hard to convert money to systemic change (difficult to measure effectiveness, hard to coördinate on optimal approach, etc.). On the other hand, I do think that this leads to an undervaluing of careers that work in systemic change (and important considerations that cross cause areas, since they’re also hard to donate to). This might not be true if you have AI timelines too short for systemic changes to come into being.
Not super confident about this, though. Feel free to try to change my mind.