Hacker News Reader

Comment by notahacker | Hacker News Reader

But in practice, this list of questions reflects less the principles of Effective Altruism and more the overlap between people calling themselves Effective Altruists and other personal and career preoccupations. That's how you get four of the six "most impactful" questions in politics and international relations being about AI and one of the others involving representation for 'sentient nonhumans'. (Suffice to say this is not a list a political scientist, or even your average consequentialist utilitarian cosmopolitan humanist who read more newspapers than AI papers would propose as research priorities in politics and IR. Probably great research questions to get you a popular blog and a job in a Silicon Valley research institute though)

View on HN

Although all organizations are influenced by their particular constituency, the guys at 80k have been led to these causes by particular philosophical and empirical arguments, which they explain in the website at great length, not (say) because they all used to be ML researchers. (They do not have technical backgrounds.) You can certainly disagree with the arguments, and they would be first to agree that their conclusions (although not premises) are radical.

But more importantly, I think if you just went and talked to random political scientists, you would find they actually had not thought very hard about what the most important causes are. If you ask them, they would come up with something on the spot. Indeed, political scientists don't at all seem to be the correct experts for this question.

> But more importantly, I think if you just went and talked to random political scientists, you would find they actually had not thought very hard about what the most important causes are. If you ask them, they would come up with something on the spot. Indeed, political scientists don't at all seem to be the correct experts for this question.

Did you just come up with this on the spot? ; )

But, I agree! Personally, I'd ask the historians! ;)

No, the questions you're raising have been asked (reasonably) by many. I really do think the EA folks are doing something novel, and there is a reasonable but incorrect intuition that all of this must have been sorted out before (so that if they're getting unconventional answers they must have made a mistake in their assumptions or process).

Step 1 in finding the highest impact thing to do is being serious about establishing cause and effect, and historians by their own admission rarely try to do that. (Cliometrics is a bit better.)

Interesting. If you have time could you share some articles / sources / examples / something to lend some solidity to what you're arguing, and also so that I can follow up and learn more?

TBH I'd make an argument for the elimination of war being a huge way to better the world. Imagine:

- No invasion of Iraq.

- No bungling of the invasion of Afghanistan (or no invasion)

- No Syrian war and millions of refugees.

War has huge negative ripple effects in addition to all of the deaths it causes; societies and communities are destroyed, whose effects echo for generations. It consumes insane amounts of resources. If we want to think big, that's what I'd think about. Maybe even focus on exclusively. I mean, who do you think is going to buck against checks and balances on scary intelligent AI research? Defense-funded research.

Sure, EA talks about international relations, etc, but, where's the study on the cost/benefits of world-wide pacifism, if we're theorizing here?

Part of my skepticism of all this is because there is so much thought put into bettering the world over time:

- All the religions, which still undergird the various systems of morals/laws we have.

- To name a few philosophers / social thinkers who directly discussed making things "better":

    - Plato

    - Hobbes

    - Marx

    - Habermas (I haven't read him, but I've read about him; big on liberal democracy and the post WWII European system, I think)

That's a pretty hefty list right there. That's not even a full accounting of the Western philosophers/thinkers. Not to even mention the non-Western philosophers/thinkers.

And, tbh, I think it's easy for non-historians, or scholars who don't focus on political science or political philosophy/theory to just write the field off; and, IMHO the "hard" sciences and tech/CS folk in particular have a tendency to do this. So I'm always a bit skeptical.

I'm going to go out on a limb and say that political scientists that spend their entire working life studying political science and teaching students the most important questions in political science (including ones which aren't their specialism!), might have put more thought into which political science questions have impact and which aren't studied enough than was required to pick three questions from Allan Defoe's paper on AI, once from a CNAS paper on AI and one from a list already on their website, before moving onto the next field in the alphabet. The list covers 19 academic disciplines and a third of their sources are papers mainly about AI!

To be fair to the authors, they openly state right at the beginning of the article that the scope of the questions they're listing is based entirely on questions being asked within their community. I'm not saying the questions aren't worth answering, but a community with less overlap with LessWrong's 'politics is the mind killer' singularity-believers would perhaps choose a set of 'important questions in politics' somewhat less narrowly focused on AI. And this community overlap clearly has more of an impact on the authors' conclusion that the most impactful questions in politics nearly all concern AI than pure philosophical commitment to maximising the utilitarian value of their time.

> I'm going to go out on a limb and say that political scientists that spend their entire working life studying political science and teaching students the most important questions in political science

Political scientists will know the most important questions in political science as judged by intellectual interest, but that's just not the question being asked here. They are asking which questions in political science will have the biggest impact on the world as judged by a utilitarian and long-termist framework, and I don't see why political scientists would have a confident answer to that.

Likewise, I am a physicist with expertise in what questions are of intellectual interest to physicists, but I don't think that I or my colleagues have good ready answers to which physics questions will have the biggest impact on the world.

> And this community overlap clearly has more of an impact on ...

But the community overlap wasn't a random event that is just now influencing this investigation. The members of the community were attracted to each other because they were convinced by certain abstract arguments. (LessWrong is concentrated in Berkeley while 80k Hours is UK based and the members mostly hail from there and Australia. They found each other through the internet and through the Oxford philosophy department.) You can certainly disagree with the arguments, but chalking this up to having overlap with some dorky community is a cheap ad hominem.

> They are asking which questions in political science will have the biggest impact on the world as judged by a utilitarian and long-termist framework, and I don't see why political scientists would have a confident answer to that.

> Likewise, I am a physicist with expertise in what questions are of intellectual interest to physicists, but I don't think that I or my colleagues have good ready answers to which physics questions will have the biggest impact on the world.

While answering the question involves predicting the future in ways no one should be overly confident in, it's worth noting that social impact questions of that type are, in fact, within the domain of political science in a way they are not within the domain of physics, so practitioners within the two fields aren't exactly similarly situated with regard to the question.

Sort of. Some political scientists are certainly more likely to estimate the impact of their policy suggestions than physicsts, mostly because physicists rarely make policy suggestions. But I don't think they try to survey all political science questions and systematically compare them along some measure of impact. I expect the hypothetically disagreement between Dafoe and a random political scientist is based on a disagreement outside the expertise (the importance and long-term impact of AI in general). Likewise, physicists wouldn't tell you much about the impact of their work because it hinges on things (e.g., econ) outside their expertise.

And is everything you are saying just sort of, your own take on things?

Or do you have some sort of special expertise?

Because sure, what you're saying sounds good, but it's far from a piece of ironclad logic, so, color me un-convinced : ).

And you think AI researchers will be better?

So. This seems like the perfect question for folks with deep knowledge in history, politics/current events (on a global scale), and philosophy.

Luckily, that's kind of what a lot of political philosophy / political science / political theory (the name varies) is!

Or social theory. Or philosophy. Even literature, to be honest. You laugh? Consider how often literature has been a driver of social reform.

It's also how you get a list at all. I don't like EA's "Jeremy Bentham solved philosophy forever" worldview, but at least it gets them thinking big, you know? Is anyone else making these lists?

Reply on news.ycombinator.com