Can “effective altruism” maximise the bang for each charitable buck? • Hacker News Board

Re: Can “effective altruism” maximise the bang for each charitable buck?

#61

Post by monadgonad » Fri, Jun 01, 2018, 3:30 PM UTC

skybrian wrote: post #58

monadgonad wrote: post #45

I think the Effective Altruism movement really belies its own values and cause with the fact that one of its own funds is for supporting: > organizations that work on improving long-term outcomes for humanity. Grants will likely go to organizations that seek to reduce global catastrophic risks, especially those relating to advanced artificial intelligence. [1] Yes, an argument can be made that it's important to fund…

Note that GiveWell is sticking to its original mission of evaluating charities based on scientific studies and isn't recommending AI funding. However I'll push back on this. If it's okay for the government to fund basic scientific research where the benefits are long-term and unpredictable, why not private individuals? Or are you saying research shouldn't be funded at all?

I think it's fine for private individuals to do it (though I disagree with their choice), but not to claim that it's evidence-based Effective Altruism while doing so - it is an emotionally-led value judgment, just like donating to my local dog shelter would be (which I understand others may disgree with). That said, I do recognise now that effectivealtruism.org does not represent the entirety of the movement; I have no fault with what GiveWell are doing.

Re: Can “effective altruism” maximise the bang for each charitable buck?

#62

Post by monadgonad » Fri, Jun 01, 2018, 3:32 PM UTC

ogennadi wrote: post #57

monadgonad wrote: post #45

I think the Effective Altruism movement really belies its own values and cause with the fact that one of its own funds is for supporting: > organizations that work on improving long-term outcomes for humanity. Grants will likely go to organizations that seek to reduce global catastrophic risks, especially those relating to advanced artificial intelligence. [1] Yes, an argument can be made that it's important to fund…

I used to feel similarly until I realized that the EA movement isn't a hierarchical organization: it's just a bunch of totally separate orgs who have a common philosophy about how to do good in the world. Sure, OpenPhil funds AI research. But Givewell (who, last I checked, share the same office) has in their Top Charities list, only those that work in poor countries https://www.givewell.org/charities/top-charities [2…

Thank you; this has opened my eyes

Re: Can “effective altruism” maximise the bang for each charitable buck?

#63

Post by brownbat » Fri, Jun 01, 2018, 3:42 PM UTC

thatfrenchguy wrote: post #39

GiveWell’s CEO’s compensation is 200k: http://files.givewell.org/files/ClearFund/Meeting_2017_06_06... That, after working at a investment firm and having a BA in religion. Not sure that’s the skills and salary I would trust the use of the word « reason » and « science » with. Having a 200k salary in San Francisco definitely skews your vision of the world.

If something goes wrong with an entity dealing with over a hundred million dollars, the CEO is expected to respond relatively immediately, as in, on call 24/7, because a hundred million represents the interests of a lot of people most likely.

The variety of things that can go wrong is large, and the necessary responses are nuanced, potentially requiring sophisticated knowledge in several domains, from legal, accounting, personnel, leadership, to communication and reputation management.

I have no idea what such a role should pay, but 0.2% of the money moved by the organization doesn't sound crazy, insofar as money moved correlates somewhat to the risk of lawsuits or the death of the organization after an emergency is handled poorly.

Aside from the day to day management, look at it like an insurance policy on the whole enterprise. In those terms it's actually a pretty reasonable rate.

This theory doesn't justify any price whatsoever. Some CEO compensations are surely still too large. I just don't think this is the best case study for that point.

Re: Can “effective altruism” maximise the bang for each charitable buck?

#64

Post by mlthoughts2018 » Fri, Jun 01, 2018, 3:43 PM UTC

monadgonad wrote: post #59

mlthoughts2018 wrote: post #55

Earlier quoted context omitted.

Isn’t this just a statement that you disagree either with the relative probability of an AI-related disaster or the relative degree of harm it could cause? It seems harsh to say it’s “morally repugnant” when in the end you’re just saying the way you assign probabilities and degrees of negative outcomes would lead you to invest in a different portfolio of charities or efforts than what this other group would do. (I’m…

> Isn’t this just a statement that you disagree either with the relative probability of an AI-related disaster or the relative degree of harm it could cause? Yes, but the point of this movement is that they're supposed to be evidenced-based and maximising ROI. There is evidence for the existence and threat of global warming and drug-resistant bacterial. When you move away from charitable efforts whose effectiveness w…

That's a fair point about moving into areas in which the effectiveness is hard to measure. However, even apart from AI threats, I'd argue that there are many areas like this that pose potentially great harm. For example, I think that we ought to spend more effort to prevent regulatory capture in corporate-friendly legislation. But whether we can find a metric we'd all agree on to target this from an empirical point of view is a very difficult question.

It reminds me of Gilb's Law, "Anything can be measured in a way that is superior to not measuring it at all."

So if your sincere beliefs were such that the threat posed is high enough, and total harm would be astronomical, then you might believe that accepting a present-day metric that has a lot of variance and which is hotly debated might still be ok. Empirical progress along that rough metric might be more valuable, still in an expected value sense, than progress in other possibly less-harmful areas even if they have more clearly defined metrics.

Again, not arguing for or against, just trying to represent why someone might be sincere about investing in AI safety and why, under their particular beliefs and preferences, it could still be rational from an EA point of view, despite more ambiguity.

There are always risks that a certain model's loss function does not correctly correspond to the goals they wish to optimize towards, or that there is a large amount of idiosyncratic noise in the observation of the loss criteria. But you can still account for the risks of these sources of error in an overall method that is still empirical.

Re: Can “effective altruism” maximise the bang for each charitable buck?

#65

Post by justinpombrio » Fri, Jun 01, 2018, 3:43 PM UTC

monadgonad wrote: post #45

I think the Effective Altruism movement really belies its own values and cause with the fact that one of its own funds is for supporting: > organizations that work on improving long-term outcomes for humanity. Grants will likely go to organizations that seek to reduce global catastrophic risks, especially those relating to advanced artificial intelligence. [1] Yes, an argument can be made that it's important to fund…

> However, to consider artificial intelligence to be a potential global catastrophe at all, let alone the single one requiring extra funding, is mostly unfounded.

As another commenter points out, EA is not a single unified group, but rather a disparate set of people who are united by the belief that we should put money where it has the largest effect. Some, but not all, of them are convinced by the arguments in favor of doing AI alignment research. If you aren't, then don't donate money for that, donate it to what you think is a more effective cause!

I'm curious, though, how people come to the conclusion that AI risk is nonsense? Is it a gut reaction, or does it come from thinking or reading about the problem? Every popsci article I've seen on the topic has been atrocious, so kudos if you thought "nonsense!" after reading such an article.

Scott Alexander has a pretty readable introduction to AI risk: http://slatestarcodex.com/superintelligence-faq/

Re: Can “effective altruism” maximise the bang for each charitable buck?

#66

Post by savanaly » Fri, Jun 01, 2018, 3:43 PM UTC

monadgonad wrote: post #59

mlthoughts2018 wrote: post #55

Earlier quoted context omitted.

Isn’t this just a statement that you disagree either with the relative probability of an AI-related disaster or the relative degree of harm it could cause? It seems harsh to say it’s “morally repugnant” when in the end you’re just saying the way you assign probabilities and degrees of negative outcomes would lead you to invest in a different portfolio of charities or efforts than what this other group would do. (I’m…

> Isn’t this just a statement that you disagree either with the relative probability of an AI-related disaster or the relative degree of harm it could cause? Yes, but the point of this movement is that they're supposed to be evidenced-based and maximising ROI. There is evidence for the existence and threat of global warming and drug-resistant bacterial. When you move away from charitable efforts whose effectiveness w…

It's like calling someone morally repugnant because they agreed to give you money equal to the sum of two and two and then gave you three dollars, insisting that the math actually does come out that way. If they honestly do believe that, they are simply mistaken, not morally repugnant.

Re: Can “effective altruism” maximise the bang for each charitable buck?

#67

Post by Seipaho8 » Fri, Jun 01, 2018, 3:49 PM UTC

castlecrasher2 wrote: post #14

Seipaho8 wrote: post #6

Earlier quoted context omitted.

> who value things differently Can you try to formulate those values? > I would add that the Make-A-Wish foundation has a very significant quality of life impact on the kids AND their support network But given the costs per instance that impact is only on a small fraction of all cases. A smaller impact for everyone or one that only materializes in the future would likely generate more utility if you add them up.

> Can you try to formulate those values? Perhaps your question is phrased poorly, but it appears you're asking OP to rationalize every value a donor has when they donate to an organization like Make-A-Wish. Is that what you meant? If so, where would one even start to answer this question?

Well, I wanted a list of values that, for them, rank higher than human lives. I mean every donation has an opportunity cost, and that cost can be lives not saved. People are implicitly willing to make such tradeoffs. What I am asking is to think about it and make those preferences explicit.

Re: Can “effective altruism” maximise the bang for each charitable buck?

#68

Post by wadfginoi » Fri, Jun 01, 2018, 3:50 PM UTC

castlecrasher2 wrote: post #52

Symmetry wrote: post #28

Earlier quoted context omitted.

There's nothing shameful about donating to the Make-A-Wish foundation. It's certianly better to donate there than to spend the money on booze, video games, fancy clothes, etc or the vast majority of things competing for your money. But you still get far less moral credit for donating there than donating somewhere more effective. And yes, you can count the effects of the Foundation on the kids support network but reme…

> But you still get far less moral credit for donating there than donating somewhere more effective. I take issue with this attitude. How is this anything but degrading and shaming those who give their time and money to a noble cause? I frankly don't see how your opinion on efficacy matters, it appears you want to make people feel bad for doing good-but-not-good-enough things.

The Make-A-Wish Foundation is mostly useless. People have a moral obligation to set aside their feelings in order to do the right thing. Giving to the Make-A-Wish Foundation is motivated mostly by a desire to feel good.

If you want me to stop saying that, prove me wrong. Calling me mean isn't an argument. If what I say makes people feel bad, the solution isn't to start lying to make them feel better.

Re: Can “effective altruism” maximise the bang for each charitable buck?

#69

Post by monadgonad » Fri, Jun 01, 2018, 3:52 PM UTC

savanaly wrote: post #66

monadgonad wrote: post #59

Earlier quoted context omitted.

> Isn’t this just a statement that you disagree either with the relative probability of an AI-related disaster or the relative degree of harm it could cause? Yes, but the point of this movement is that they're supposed to be evidenced-based and maximising ROI. There is evidence for the existence and threat of global warming and drug-resistant bacterial. When you move away from charitable efforts whose effectiveness w…

It's like calling someone morally repugnant because they agreed to give you money equal to the sum of two and two and then gave you three dollars, insisting that the math actually does come out that way. If they honestly do believe that, they are simply mistaken, not morally repugnant.

It's not being "simply mistaken", it's a calculated belief. Someone might think it's a moral act to fund, say, ecoterrorists to bomb the headquarters of a major oil company, believing that the loss of life is worth it for helping the world by destablising the company. We might disagree, and consider that highly immoral.

That's an extreme example of course, and in this case the worst that can happen is some loss of money, rather than loss of lives, but I hope it illustrates my point: them choosing to pay people to think about AI, rather than giving that money to other causes, is a moral choice with which I disagree.

Re: Can “effective altruism” maximise the bang for each charitable buck?

#70

Post by _fol8 » Fri, Jun 01, 2018, 3:53 PM UTC

monadgonad wrote: post #45

I think the Effective Altruism movement really belies its own values and cause with the fact that one of its own funds is for supporting: > organizations that work on improving long-term outcomes for humanity. Grants will likely go to organizations that seek to reduce global catastrophic risks, especially those relating to advanced artificial intelligence. [1] Yes, an argument can be made that it's important to fund…

To provide the counter argument I imagine the Effective Altruism folks may give (not that I necessarily support this position):

A number of specialist AIs could potentially eliminate 50%+ of all jobs in the next 50 years (a generation). What does society look like when we don't have the safety nets to support a massive unemployed population? What could it mean for geo-politics if the rich countries population is simply rich because they have the most advanced algorithms, but most people don't work? We're talking about massive social unrest worldwide potentially - capitalism cannot necessarily survive this paradigm shift and since the fall of the Berlin Wall the world has no other social organization in modern times, unless we'd like to revert to totalitarian rule.

This all presumes of course, that the first AGI they turn on doesn't just turn the universe into a pile of paperclips (ducks)