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

501–510 of 619 posts

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

#501

Post by PebblesRox »

Centigonal wrote: post #493
PebblesRox wrote: post #434

Earlier quoted context omitted.

"Maybe it's some kind of horrible drug interaction with Rationalism, or maybe EA is fundamentally Rationalist?" I have always associated EA with the rationalist movement, since I came across it via reading rationalist stuff. If you define EA as "figuring out how to deploy a given set of resources so as to do the most good," your conclusions are going to depend on your value system: how do you define and measure good?…

> If you define EA as "figuring out how to deploy a given set of resources so as to do the most good," your conclusions are going to depend on your value system: how do you define and measure good? This is exactly it. Because "good" is subjective to the individual, EA without some kind of consensus building reduces to "maximize your agency as an individual in the world, and do what you think is good," which gets unco…

> EA without some kind of consensus building reduces to "maximize your agency as an individual in the world....

I agree and would say that to the extent that there is a consensus, it becomes “maximize our agency as a coordinated group of people” – hence the focus on promoting and expanding EA as a movement

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

#502

Post by naasking »

polygamous_bat wrote: post #391
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…

This whole thread is very stereotypically EA to me -- and why I see it as a sham. Starts off at the top saying how it's all about donating to reasonable charities, but soon devolved into "what about muh AGI". I am literally doing my Ph.D. in robot learning and I can tell you that every single "longtermist" effective altruist I have talked to has been more interested in getting high on their own stash of Sci Fi fantas…

> rather than interacting with the actual practitioners of the field and understanding what is happening, what are the very real limitations of it, and why their assumptions range from willfully ignorant to often simply maliciously stupid.

On the other hand, I've found that "practitioners in the field" often consistently overestimate how complex and different the human mind is from our computers.

Take the recent "LaMDA is sentient" debate. Machine learning researchers consistently and vehemently denied that LaMDA is sentient. They no doubt have a deep and subtle understanding of how LaMDA works, but apparently they didn't realize that we have no mechanistic understanding of what sentience actually is or how it works, so they have absolutely no basis for such a conclusion.

So on the one hand we have these AGI doomsayers that, per your evaluation, don't understand machine learning subtleties, and on the other hand we have some machine learning researchers skeptical of AGI who consistently overestimate human specialness.

At this time the error bars I would assign to each side are wide enough that there's lots of overlap, but given the track record of human specialness falling to algorithmic progress, I'd say some investment in AGI safety is prudent.

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

#503

Post by yifanl »

FeepingCreature wrote: post #492
danaris wrote: post #488

Earlier quoted context omitted.

Part of the problem with this is that conclusive demonstration that AGI is possible is not even in the same ballpark as a conclusive demonstration that it would have consequences like the EA folks seem to believe. From everything I've read about their views, they seem to believe that either the moment an AGI is created or very, very soon thereafter, it will achieve the Singularity and "ascend" to something near godho…

I agree with your description of the AGI/ASI position. My take on it is the other way around: every step of the argument seems plainly obvious.

Plainly, that is not the case, as you are choosing to not spend your every waking moment to stop the antichrist AI coming into existence.

So either it's less obvious than you're claiming, or you are taking the threat of creating an evil AGI not seriously enough according to your own beliefs.

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

#504

Post by naasking »

CobaltFire wrote: post #500
naasking wrote: post #496

Earlier quoted context omitted.

> Either way: when your logic has established that you can obtain moral superiority by literally buying indulgences, it has officially proven too much, and if you can't recognize that, something has gone terribly wrong in your thinking Buying indulgences was wrong because it was used to cover up moral wrongdoing or reduce ones penance for wrongdoing, not because the concept of money buying moral praise is intrinsical…

Someone who has made millions or billions is, on average, doing this by making life worse for many people than if they were not in the loop taking profits. There are exceptions (Notch, etc.) where they took profits that were freely given rather than those that were skimmed off the top. So you are literally saying that by taking a portion of that money and donating it they are morally superior to the poor they took it…

> Someone who has made millions or billions is, on average, doing this by making life worse for many people than if they were not in the loop taking profits.

See the part where I said "all else being equal".

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

#505

Post by naasking »

simonh wrote: post #247
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…

I don't know anything about the movement, so can't comment on the earlier points. >If AGI has even a 1% chance of happening this century and those donations reduce the risk of it going wrong and killing everyone by 1% that's still money very well spent. Ive seen this argument many times on many different issues and I think it's almost always bunk. The classic example in Fusion research, I had friends when I was a stu…

> With general, or strong AI we only just barely have the beginnings of an idea of how to even do preliminary research on it.

Maybe. Or maybe we're just on small change away from AGI. We could be, because we don't yet really know what's required and we may not ever know until after we've done it. That prospect should be terrifying.

> It's far, far too early to start worrying about controls and management strategies for the end point of that research.

No, because if you create AGI before we have some containment or safety protocols, then it may already be too late.

Consider whether we should have researched biolab safety before we started experimenting on modifying pathogens in various ways. Doesn't that just seem blindingly brain-dead obvious? For some reason this clear and obvious conclusion just goes out the window when it comes to AI.

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

#506

Post by naasking »

bootsmann wrote: post #444
edanm wrote: post #389

Earlier quoted context omitted.

You might totally be right. The problem is, humanity is pretty bad at estimating when a new technology is about to be ready. Even the experts are pretty bad. See e.g. multiple people saying human flight is many years away, including some who were saying it unknowingly after it had been achieved! So I think the question is: when should we worry about ai safety? How early is too early? Plenty of experts think now is a…

> Plenty of experts think now is a good time to be spendning at least some resources on this. And we're not talking nuclear fusion level resources here; I don't know the numbers, but I imagine the amount of resources are actually fairly tiny compared to almost all other scientific spending. (And this was certainly true 10 years ago.) The risk theses experts are talking about isn't existential. Its stuff like GPT3 bei…

> This is entirely orthogonal to any existential risk a supposed AGI might create. One is a testable scientific hypothesis and the other is a futurist horror story which is untestable by definition.

There is no difference in their testability, it's just that the existential risk test means no more humans to care about the results.

The danger of a biolab creating a pathogen that wipes out all humanity is real but remote, and is a simple extension of the argument that the lab could create pathogens that can "only" kill or harm millions. That's why biolab safety is carefully studied, and the same argument applies to AI safety.

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

#507

Post by zdragnar »

dagw wrote: post #214
zdragnar wrote: post #139

Earlier quoted context omitted.

People have a finite amount of money to give, so every dollar is an either/or decision. Every dollar you spend on a theater donation is a dollar less that you value human lives. Of course, the same logic applies to candy bars, internet, television, and all the things that being you joy in life.

But much like with investing in stock I don't know for sure a priori which investment will bring the highest return. If I'm 80% sure that charity A is better than charity B (for my definition of 'better) I'll give 80% of my investment to charity A and 20% to charity B.

When comparing mosquito nets versus theater, I don't really think the stock diversification analogy fits at all.

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

#508

Post by blep_ »

TedDoesntTalk wrote: post #205
yanderekko wrote: post #155

Earlier quoted context omitted.

I don't think EA institutions argue that they have the One True Answer. They just often give recommendations - "eg. if you care about maximizing QALY overall, we recommend charities X, Y, and Z." If I don't care about QALY maximization, fine. But if I do, why would I not listen to the recommendations?

Longtermism and QALY do not seem compatible.

They do if you legitimately think the long-term thing will kill everyone.

I don't think that myself, but it's internally consistent.

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

#509

Post by naasking »

coldtea wrote: post #303
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…

> 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…

> such as that intelligence can be grown to absurd levels (or even just majorly) just by adding hardware to allow for it

This has been the empirically validated trend. The more hardware, the more parallel lines of thought can be explored at once (like in chess).

> that intelligence can just be expanded by itself (thinking itself into being more intelligent)

Of course it can, because humans have already done this. What do you think language, writing, books, school, science and computers are? They are humans augmenting their own effective intelligence by various means. An AI will likely have further means at its disposal, such as access to its own source code (see Godel machines and AIXI for formal models of how this works).

> more importantly intelligence is some kind of superpower that transends very real limitations (like, physical access, or convincing people to do its bidding)

Intelligence is a power, not a superpower. However, money is a power too, and all of our money is digitized these days, and a digital superintelligence might arguably have an advantage in this domain. A superintelligence with lots of money can arguably convince a whole lot of people to do what it wants and get the physical access it needs.

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

#510

Post by pvg »

FeepingCreature wrote: post #326
froggychairs wrote: post #218

Earlier quoted context omitted.

Not OP, but I do work in reducing harms of deployed AI systems at a big tech company When most EA/Rationalist folks discuss AI Safety. They’re talking about AGI, a hypothetical construct that we have no way of proving is even possible Now there’s nothing inherently wrong with funding research of that sort. The problem comes is that many EA/Rationalist folks begin downplaying legitimate tangible risks on the horizon (…

The atom bomb was once a hypothetical construct. The Einstein-Szilard letter was talking about a danger that they had no way of proving was even possible. Proof is irrelevant outside of mathematics; the only relevant question is cost to expected benefit - or expected harm. With dangerous research, when a conclusive demonstration of possibility is achieved, it is usually too late to do much in the way of prevention. T…

The atom bomb was once a hypothetical construct. The Einstein-Szilard letter was talking about a danger that they had no way of proving was even possible.

This isn't an accurate representation, really. The letter was written exactly because it was clear that a fission bomb is very likely within the short-term reach of current technological and industrial capability. This is very much not the case with AGI.

Post reply on HN