Live data from Hacker News

Effective altruism is not effective

philosophersbeard.org

181–190 of 248 posts

Re: Effective altruism is not effective

#181

Earlier quoted context omitted.

If you are unclear between AMF (bed nets) and AI Risk, and you find AI Risk dubious, then donate to AMF. It sounds like you're complaining that your choices have been pared down to two, and you don't like one of them.

I think the pared down bit is important: the whole point behind EA is to increase giving to some types of organization at the expense of giving to other types of organization based on success metrics. So if leading EA initiatives are dissuading (say) giving to electrify Africa because of a lack of efficacy data or wealth of existing funding, they might event be having a quantitative negative impact on those particula…

The point of EA is to prioritize E for people who want to be E, and to help people in general to make informed decisions.

If you think raising up your city or your country is unboundedly more important than raising up, say, Africa, then EA supports you making the best decision in that framework. But it's still silly that we expect everyone to independently evaluate charity choice in depth on, basically, the basis of their advertisement material. And people will usually either not care, or want to do the most good they can do, period. It's a rare case of a person who wants to do the most good they can for their country, but stops caring at the border.

> So if leading EA initiatives are dissuading (say) giving to electrify Africa because of a lack of efficacy data or wealth of existing funding, they might event be having a quantitative negative impact on those particular causes.

Well, yes, spending is zero sum. On the other hand, this creates an incentive to gather efficacy data, which I enormously welcome. The point isn't to maximize moneyflow to charity, the point is to maximize good things happening.

> an already exceptionally well-funded AI research project their friends work for whose impact on Microsoft's share price is much easier to quantify than their impact on preventing Hypothetical Cyberpunk Future

I don't understand what this means. At any rate, Microsoft share price is not generally seen as a negative. If a project that buys Microsoft compute nodes has positive value for the world, it would be stupid not to fund it just because doing so also happened to be good for Microsoft.

edit: Oh yeah, inasmuch as AI safety as a cause is well-funded now it's because of Effective Altruism raising the alarm to begin with - and AI capability research still outstrips it a hundred times over.

Frankly, what I'm seeing with EA is that ... like, Effective Altruism started out talking about maximizing the amount of good for the world, and the "weird AI people" were the ones who showed up. And it's not like it's a contradiction either, inasmuch as EA now focuses less on earning to give, it's because people like Bill Gates, once-leader of Microsoft, have Malaria handled. Focusing on tech people seems to have been very good for the Malaria cause. So to come in now, years later, and to say that EA shouldn't have focused on the weird AI thing because it might put off some other people who never gave a damn to begin with is almost insulting.

Re: Effective altruism is not effective

#182

Earlier quoted context omitted.

The focus on AI risk is because it seems to be the #1 long-term existential risk, so it has essentially an infinite payoff. But we have idea how to solve it (it's not tractable) and future concerns should be discounted, so most people in EA I've met don't think it's the best use of effort.

> The focus on AI risk is because it seems to be the #1 long-term existential risk, so it has essentially an infinite payoff. You can read "AI risk" as "not going to Hell" here; rationalism is just a new religion focused on worshipping what SF writers think computers are going to get up to in the future (Roko's basilisk). There's not much reason to believe an evil AI out to conquer the world is a self-consistent idea…

I'm willing to bet you got all your info on this topic from RationalWiki. A note: the author of that page has a bone to pick with rationalism. Don't take it as gospel, read other sources.

(The Basilisk gets discussed about a million times more on pages about LessWrong than the actual LessWrong. It's a decision-theoretic curiosity.)

Re: Effective altruism is not effective

#183

Earlier quoted context omitted.

> The focus on AI risk is because it seems to be the #1 long-term existential risk, so it has essentially an infinite payoff. You can read "AI risk" as "not going to Hell" here; rationalism is just a new religion focused on worshipping what SF writers think computers are going to get up to in the future (Roko's basilisk). There's not much reason to believe an evil AI out to conquer the world is a self-consistent idea…

I'm willing to bet you got all your info on this topic from RationalWiki. A note: the author of that page has a bone to pick with rationalism. Don't take it as gospel, read other sources. (The Basilisk gets discussed about a million times more on pages about LessWrong than the actual LessWrong. It's a decision-theoretic curiosity.)

I don't even know what that is. Most of my information comes from personal observation of their blogs and observing that they are all in polycules.

Re: Effective altruism is not effective

#184

Earlier quoted context omitted.

I'm willing to bet you got all your info on this topic from RationalWiki. A note: the author of that page has a bone to pick with rationalism. Don't take it as gospel, read other sources. (The Basilisk gets discussed about a million times more on pages about LessWrong than the actual LessWrong. It's a decision-theoretic curiosity.)

I don't even know what that is. Most of my information comes from personal observation of their blogs and observing that they are all in polycules.

That's fair enough.

edit: actually in that case

> There's not much reason to believe an evil AI out to conquer the world is a self-consistent idea.

Wanna argue this? The case is mostly based on the notion of basic AI drives ( https://selfawaresystems.files.wordpress.com/2008/01/ai_driv... ), indicating that for nearly any agentic (with general intelligence and capable of plan generation and execution) system driven by nearly any goal, perpetuating its own existence and protecting itself from danger will be instrumentally useful. (For an AI, "danger" = "human with off switch")

Re: Effective altruism is not effective

#185

Earlier quoted context omitted.

I think you are mischaracterising the argument. His argument is that effective altruism can not eliminate poverty, it's inherent in the concept. You also dis not show that he is the one who mischaracterises the "believes" (maybe you mean arguments here?). I think he gave a reasonably accurate description of effective altruism, what in the description do you believe is wrong? You have not actually engaged with the act…

No, the burden of proof lies on the person claiming that something is "obvious". And nitpicking the numbers is incredibly important if said nitpick happens to span orders of magnitude.

It does not fundamentally change the argument, so nitpicking does not change how valid an argument is.

Re: Effective altruism is not effective

#186

Earlier quoted context omitted.

> (e.g., in the trolley problem, what happens if you save the 3 people over the 1 person, but one of those 3 people ends up becoming (or is) a dictator who orchestrates a genocide.) Presumably by this logic, you'd have no opinion on whether a midwife should save a newborn baby's life, as it's unknowable whether the baby will one day orchestrate a genocide?

I'll bite. Probability comes into play. The vast majority of babies do not in future orchestrate a genocide, but instead live lives worth living, and in turn contribute in some way to shared human progress and wellbeing. So the expected value of the baby's life is positive, the baby should be saved. The very low chance of a future genocide does not weigh highly in this decision.

Thanks, but elsewhere Karrot_Kream says "I'm not sure what probabilities have to do with anything here" so I'm really hoping to hear their non-probability-based thinking.

Re: Effective altruism is not effective

#187

Earlier quoted context omitted.

I think you are mischaracterising the argument. His argument is that effective altruism can not eliminate poverty, it's inherent in the concept. You also dis not show that he is the one who mischaracterises the "believes" (maybe you mean arguments here?). I think he gave a reasonably accurate description of effective altruism, what in the description do you believe is wrong? You have not actually engaged with the act…

> I think he gave a reasonably accurate description of effective altruism, what in the description do you believe is wrong? I disagree that this article give an accurate impression of the EA. The main point of effective altruism is that people should use evidence in choosing which charitable causes to devote their time to. I don't feel like the author sufficiently engages with this point; instead he attacks Singer fo…

Well, I disagree the whole premise of Singer is that we can't change things through the political process so instead we should use "effective altruism".

Now if we are donating to charity should we follow the principles of EA? The answer to that is probably yes, but that's a different question and is not the point raised by the author.

Re: Effective altruism is not effective

#188

Earlier quoted context omitted.

The focus on AI risk is because it seems to be the #1 long-term existential risk, so it has essentially an infinite payoff. But we have idea how to solve it (it's not tractable) and future concerns should be discounted, so most people in EA I've met don't think it's the best use of effort.

> The focus on AI risk is because it seems to be the #1 long-term existential risk, so it has essentially an infinite payoff. You can read "AI risk" as "not going to Hell" here; rationalism is just a new religion focused on worshipping what SF writers think computers are going to get up to in the future (Roko's basilisk). There's not much reason to believe an evil AI out to conquer the world is a self-consistent idea…

You're arguing the idea of AI dominance is internally inconsistent? That's a new one.

Re: Effective altruism is not effective

#189

Earlier quoted context omitted.

Countries, not nations. There are many nations living in Argentina and Venezuela, some do have political representation, some do not. e.g.: the Mapuche nation in Argentina lacks political representation. Then, Argentina is working extremely well, except for themselves, that is. They are in a treadmill of unpayable debt that works in the following way: - The left is tasked with buying people's complacency with borrowe…

Have they tried selling themselves back to Spain?

Argentineans are happy with their limbs and they surely are not in need for illegitimate children, so no.

Re: Effective altruism is not effective

#190
post #43

This article illustrates a big issue with people opposed to EA - they severely underestimate the risks associated with political change. Of course, we can just assume everything goes well, we agitate for political change, we get a change for the better, everything is great. But the risk of a different outcome never seems to be taken into consideration. For instance: We believe the conditions in our country are unfair…

I think this argument actually illustrates a big issue with EA: for many people the "worse outcome" is no worse than the status quo.

I'm not going to argue that an authoritarian theocracy or military dictatorship would be an improvement on the corporatist kleptocracy (or however we want to negatively describe the status quo). But political change is mostly not carried by those already doing moderately well under the status quo, even if they may often end up being the one implementing it.

There's always a risk of existing power structures swooping in and opportunistically inserting themselves when the dominant ones are torn down but the problem isn't the tearing down of those structures but the continued existence of those that swoop in.

This is btw why many leftists advocate for intersectionalism rather than hoping racism, sexism etc will go away on their own once the class hierarchy is defeated.

Post reply on HN