Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The good delusion: has effective altruism broken bad?

economist.com

Post reply on HN
Print view View original

351–360 of 619 posts

Re: The good delusion: has effective altruism broken bad?

#351

Post by q-big »

navane wrote: post #345

Their quality-of-life points combined with long-term-ism reminds me of that simcity build that maximized the population to absurd levels. If you get 'points' for each life, past and future, than maximizing the number of future life is the easiest way to maximize your high score. > Bostrom calculates that 10E29 potential lives are lost for every second that we fail to colonise the supercluster of galaxies containing t…

There exists another solution that is "obvious" to "everybody" who has worked in life insurance (or long-term investments in finance): the concept of present value (PV):

> https://en.wikipedia.org/wiki/Present_value

Without giving a derivation (which you can find easily on the web), the idea is that a gain of 1 USD that you get in 1 year has to be discounted with a factor of (1+i), where i is the (long-term stable) interest rate. For 3 years, the discount factor is, of course, (1+i)^3, i.e.

- a gain of 1 USD that materializes in 1 year is worth 1/(1+i) USD today

- a gain of 1 USD that materializes in 3 years is worth 1/(1+i)^3 USD today

Thus, including this discount factor for a future monetary flow (as in the PV calculation) makes gains that materialize in, say, 1000 years worth less today than those of the same monetary value that materialize (much) earlier.

Re: The good delusion: has effective altruism broken bad?

#352

Post by tim333 »

jmull wrote: post #340
concordDance wrote: post #225

Earlier quoted context omitted.

> I've dug into this stuff over the last couple days, because I had previously understood Effective Altruism to be something akin to like a pledge drive for Charity Navigator's best-reviewed charities or something, and it turns out it's nothing of the sort. As someone with multiple EA friends: this is precisely what it is. > they have their own weird language, there's a hierarchy, and people on the inside operate und…

Both those 1%s are very unlikely to be anywhere close to being true. The sci-fi version of AGI that people are imaging now will never exist. No one has any idea how to create AGI now, and therefore, no one knows what it might actually look like if it ever does. Any “solutions” people come up with now will have nothing to do with the problems a real AGI might pose. If people want to get together and talk about this st…

Well, opinions differ. I'd guess the probability of there being AGI as in computers able to think as well as a human in most ways by 2300 as close to 100%.

Things we have now are getting closer eg Deepmind's Gato and GPT-3.

Re: The good delusion: has effective altruism broken bad?

#353

Post by notahacker »

FeepingCreature wrote: post #339
notahacker wrote: post #336

Earlier quoted context omitted.

The position that the virtue of SBF's actions are irrelevant, and he should be judged better than all of us because he could afford to give more money to give away to good outcomes sounds exactly like "buying indulgences" to me. Especially in Scott Aaronson's phrasing where it's the SBF's gross donation spend being ahead of all of ours that we should be impressed by, not the net impact of his actions (which I find di…

Sure, but then the argument is "SBF was bad actually". I agree with this take - this is going to lead to bad outcomes. But note that saying so already embraces a consequentialist framework. This is a common argument for "rules utilitarianism", which is that given human psychology, getting humans to pursue (particular) virtues leads to better outcomes in the real world than getting humans to engage in first-order outc…

Well yes, nobody in this subthread is arguing that outcomes have no relevance.

We just object to Aaronson's rather extreme position that the correct measure of SBF's goodness is the single dimension of the absolute amounts he donated to notionally philanthropic causes (which seems like the reductio ad absurdum version of EA utilitarianism). And yes, I'm also unconvinced of the argument that behaviour and intent plays no role in ethics. It may be a good in terms of outcome that a sociopathic hereditary billionaire gives away 1%what would other people have done in the same position become relevant to casting judgement on his benevolence, never mind his other vices. And certainly the courts will consider intent and not just the brute fact that SBF lost a lot of people's money when deciding whether to prosecute him, convictions and sentencing; I imagine most of his defence will be structured around that...

Re: The good delusion: has effective altruism broken bad?

#354

Post by coldtea »

ethanbond wrote: post #319
coldtea wrote: post #303

Earlier quoted context omitted.

> Not weird at all. When (probably when, not if) we create it, we will quickly be interacting with an entity with a similar power imbalance as we have to ants. That's the sci-fi idea, which handwaves away several very big assumptions and improbabilities, such as that intelligence can be grown to absurd levels (or even just majorly) just by adding hardware to allow for it, that intelligence can just be expanded by its…

1) Why do we have a reason to believe there’s a cap to intelligence? 2) No, you don’t need to assume it happens via adding hardware 3) No, you don’t need to assume it can expand by itself 4) We’ve already seen people connecting AI systems to open internet and open internet to command and control of physical systems (cars, utilities, soon weapons systems) - turns out there was no convincing necessary 5) Nanobots also…

>1) Why do we have a reason to believe there’s a cap to intelligence?

There's a cap to all things. Diminishing returns is a thing. Both physical substrates have limits, and there's a utility limit to thinking or having more data. Limitless intelligence is fairy-tale magical thinking - especially in the way it's portrayed as some magical ability to build and control everything. Like, invent working cold-fusion - as if you don't need the experiments, iterative process, random breakthroughs, and such, but just "smarter" thinking.

> 2) No, you don’t need to assume it happens via adding hardware

Yes, you do, else it's just the same capped-capability hardware, thinking itself into being smarter.

> 3) No, you don’t need to assume it can expand by itself

Well, I, personally, don't need to assume anything. But people talking about AGI, do assume it.

> 4) We’ve already seen people connecting AI systems to open internet and open internet to command and control of physical systems (cars, utilities, soon weapons systems) - turns out there was no convincing necessary

That's neither here nor there, though, because I didn't say that connecting AGI to the open internet is impossible or can't happen. Just that it doesn't mean much. We still control the kind of control we give an AI, and any physical systems any AGI would be connected will still have limitations. It's not some magical portal into conquering the world.

>5) Nanobots also not a necessary assumption

Didn't say it's necessary, just pointed out that it has been made - along with many other non necessary, and far-fetched assumptions and hand waving...

>Also, incentives, not charisma, are what steer human behavior.

Both do. With charisma you can do it without incentives - and even instill incentives into people. Otherwise you're stuck with innate incentives like people wanting to be rich - which you don't need to be a AGI to exploit, just rich or a con-man.

Re: The good delusion: has effective altruism broken bad?

#355

Post by ethanbond »

olalonde wrote: post #342
llamaimperative wrote: post #291

Earlier quoted context omitted.

Not weird at all. When (probably when , not if ) we create it, we will quickly be interacting with an entity with a similar power imbalance as we have to ants. No one knows exactly what the outcome will be, even less so what to do about it, but surely the kool-aid is watching accelerating progress toward this state and believing either it won’t happen or it won’t be interesting/surprising. Disclaimer: I don’t think t…

> I don’t think this is coming soon Personally, I feel it might become a risk one day, but how soon is an important detail. If the threat is still far away, its nature is very blurry and there is no use to worry too much about it today (or advocate for a ban on AI research). There are still many safe intermediary steps we have to go through before the threat concretizes and becomes worth worrying about.

These are definitely important considerations, but it’s not clear 1) we’ll see those intermediary steps, or 2) we’ll be able to stop at them. In reality we’ve already blown through many red lines e.g. not connecting AI to the open internet, not connecting it to physical systems. We didn’t even blink!

Agreed a ban on AI research is unworkable, which is part of why this threat is so insidious.

Re: The good delusion: has effective altruism broken bad?

#356

Post by medstrom »

jopsen wrote: post #231

Wow, I took EA to mean try donating using analysis like givewell does. Judging by the article and comments, it looks as if people are living by a creed to be effectively altruistic. In general any philosophy where you think there is an optimal answer suffers from the risk that the answer could be wrong. In the case of givewell, one limitation is that you only support causes that demonstrate impact -- you don't suppor…

You weren't wrong, EA is basically the mindset of donating using analysis. As you fear, at the start the movement suffered from overemphasis on causes that can demonstrate impact. But it's not incompatible with risk-taking, if you personally deem the potential payoff as more than enough. Lately the movement is more nuanced, allowing room for harder-to-measure, "softer" causes like moral circle expansion, even if they're riskier, and even if you outsource part of your judgment to people closer to the cause: https://guerrillafoundation.org/additional-thoughts-on-effec...

Re: The good delusion: has effective altruism broken bad?

#357

Post by blitzar »

FeepingCreature wrote: post #320
blitzar wrote: post #289

Earlier quoted context omitted.

> The problem (and I think roughly the point being made) is that if you follow that logic through, then surely almost everyone here should have the vast majority of their possessions & wealth taken, to fund the equalising up-lifting of others. And if not software developers why not? Who gets to chose which grift is a grift and which is an honest living? > I find it really weird that people are so glib about the "hund…

Humans have very strong instincts against exploitation by peers. I think the appropriate answer here is: "Sure, let's seize the wealth of approximately everyone in a western country." Singling out a particular group, even though it's less seizure in total, raises people's hackles for reasons entirely unrelated to whether redistribution is a good idea.

Like the around 88% of American drivers considered themselves to be above average at driving - everyone also believes that they are the exception. As for redistribution of wealth, even many multi millionaires consider themselves to be on the poor side of the wealth divide.

Re: The good delusion: has effective altruism broken bad?

#358

Post by VieEnCode »

Apart from the already financially secure, who can possibly afford to be an effective altruist in the current climate (in the sense of donating a significant proportion of one's income). Similarly, who can afford to work for these EA foundations and charities on lowish salaries based in major cities but trustafarians?

Re: The good delusion: has effective altruism broken bad?

#359

Post by FeepingCreature »

notahacker wrote: post #353
FeepingCreature wrote: post #339

Earlier quoted context omitted.

Sure, but then the argument is "SBF was bad actually". I agree with this take - this is going to lead to bad outcomes. But note that saying so already embraces a consequentialist framework. This is a common argument for "rules utilitarianism", which is that given human psychology, getting humans to pursue (particular) virtues leads to better outcomes in the real world than getting humans to engage in first-order outc…

Well yes, nobody in this subthread is arguing that outcomes have no relevance. We just object to Aaronson's rather extreme position that the correct measure of SBF's goodness is the single dimension of the absolute amounts he donated to notionally philanthropic causes (which seems like the reductio ad absurdum version of EA utilitarianism). And yes, I'm also unconvinced of the argument that behaviour and intent plays…

I don't fully agree with Aaronson's take, I just think it's underappreciated. I am a sorta-rules utilitarian; in contested questions, I think the moral obligation is to behave in a fashion so that the standards by which we act, if universalized, maximize beneficial outcomes. And I think there is an argument to be made that getting rich people to donate is a thing that a moral system that tries to lead to good outcomes, should focus attention on.

Like it or not, we have set up our society so that economic decisionmaking power is to a large part concentrated in the hands of a small fraction of the population. This population is a moral fulcrum; the things they decide to do have a significant effect on actual events, in some ways even outstripping governments (which generally hold obligations on par with their power). I think an effective morality needs to acknowledge this and make special dispensation for these people if they decide to bend their power towards creating good outcomes, whether or not one approves of their existence in the abstract.

Assume you are trying to invent a morality to sell to the one percent. The pass-or-fail condition is that this morality must absolve their sins. It seems to me that the default option here is laissez-faire libertarianism, which has cultural bias against doing work to improve the worse off; in my view, being able to make inroads in this moral market is a significant victory for EA and good outcomes in general.

Re: The good delusion: has effective altruism broken bad?

#360

Post by imgabe »

FeepingCreature wrote: post #347
imgabe wrote: post #346

Earlier quoted context omitted.

By all means speculate all you like about your own actions and act accordingly. The original comment was about how effective altruism is intentionally looking down on others (as they are implicitly ineffective) and holding themselves up as the ideal that everyone should follow.

Aren't you making a general point here against public arguments that certain things lead to better outcomes than others? Do you reserve the same rancor for these little warning labels printed on packs of cigarettes about cancer risk? (Look at these non-smokers looking down on others...) Public debate is, to a first approximation, nothing but arguing about which actions lead to better outcomes. I could see an argument…

I draw a distinction between a particular charitable organization being effective, i.e. the amount spent on overhead vs going to cause they claim to support and a particular charitable cause itself being effective (mosquito nets vs. community theater).

For the former I agree, seek out charities that most effectively support what you want to support.

For the latter, I don't think it's worthwhile to deride people for supporting a cause just because you think there's some other cause that might technically result in more living humans. Raw population is not a worthwhile metric to optimize for. People should be free to support things they care about and believe make the world a better place as long as it's not harming anyone.

Post reply on HN