EA Forum Reader

Hey Luke, great work thus far! Props to you and the rest of the GWWC and EA NYC team for making this data available in a user-friendly format.

I know this is a big ask, but I would love to see this data visualized in something like a sankey diagram (broken down by core cause area, sub-case area, and finally individual charity, for example). One of the things I've always been curious about is how under/overfunded a given charity or philanthropic fund is relative to other entities  in the same core or sub- cause area (i.e. donor coordination problem), and I think visualizations like this could provide some really interesting insights to individual donors and fund managers alike.

Alternatively, providing a way to export this data to a CSV or creating an API for accessing the data could enable other people (e.g. me) to develop and share visualizations like the one described above.

Looking forward to seeing where this goes!

I find that this approach undermines one of the major intuitions behind utilitarianism in the first place: what is permissible, obligatory, etc., should not depend on parts of the universe that are independent of (unaffected by) my actions, (a stochastic version of) separability. It is no longer the case that what's best depends only on the ex ante prospects each individual faces, basically one of the assumptions in Harsanyi's argument for utilitarianism (Postulate c in the paper, assumption 3 here [EA · GW]) and this generalization [EA · GW] (Anteriority), because now the statistical dependence between individuals' prospects matters. You could assume separability (independence of unconcerned agents) in uncertainty-free cases and still arrive at utilitarianism, but you've still undermined the intuition. Why use an additive theory at all now?

I agree. It's a very intuitive way to introduce people to EA with something they probably already agree with.

I think it's worth keeping in mind that if the action A's expected value is higher than B's, then B can never stochastically dominate A, and there's (in theory) background uncertainty according to which A dominates B. So, if you have enough deep uncertainty about the background uncertainty and entertain multiple distributions, A might be better according to at least one, so that's a reason to prefer A, breaking the indifference.

On the other hand, you might also have deep uncertainty/complex cluelessness about which has higher expected value, anyway.

if you have extreme uncertainty about how much moral value there is in the universe (which we probably do) then paradoxically the “gamble” is actually better across the board than the supposed “guarantee”

Doesn't this depend more on the particulars? I.e. if it's "sufficiently" Pascalian, neither dominates the other, as you've written in the post and your reply to Rohin, and each is permissible.

You might check out this SEP article: https://plato.stanford.edu/entries/morality-biology/. Haven't read it myself, but looking at the table of contents it seems like it might be helpful for you (SEP is generally pretty high-quality). People have made a lot of different arguments that start from the observation that human morality has likely been shaped by evolutionary pressures, and it's pretty complicated to try to figure out what conclusions to draw from this observation. It's not at all obvious that it implies we should try to "escape the shackles of evolution" as you put it. It may imply that, but it also may not. (In particular, "selective evolutionary debunking arguments" seem to have implications along these lines, but "general evolutionary debunking arguments" seem to lead to almost the opposite conclusion.)

You might also check out this post [LW · GW] by Eliezer.

More technical blog post about this result here.

Yes this 80,000 Hours article has some good ideas

My guess is that AI safety papers are more impactful than longtermist econ, since it is directly targeted at significant near-term risks. Having said that, there are now a hundred or so people working on various aspects of long-term AI safety, which is more than can be said for longtermist econ, so I don't think the impact-difference is huge. Maybe we're talking about a three-fold difference in impact, per unit time invested by an undifferentiated EA researcher - something that could easily be overriden by personal factors. But there's plenty of people who would disagree.

My experience in transitioning from medicine to AI is that it was very costly. I feel I was set back by ~5 years in my intellectual and professional development, in the sense that I had to study a masters degree and do years of research assisting and junior research work to even get back to the level of knowledge and seniority that I previously had. From an impact standpoint, I clearly had to exit medicine, but it's not clear that moving to AI safety had any greater impact than leaving into (for example) biosecurity would have.

For most people in a PhD in any long-term relevant subject (econ, biology, AI, stats), with a chance of a tenure-track position at a top-20 (worldwide) school, I expect it will make sense to push for that for at least ~3 years, and to postpone worries about pivoting after that. Because switching subjects reduces those odds a lot.

More broadly, as a community,  we mostly ought to ask people to pivot their careers when they are young (e.g. in an undergraduate), or when the impact differential is large  (e.g. medicine to biosecurity). Which I don't think it really is when you're contrasting the right parts of econ with AI safety.

Finally, I imagine quant trading is a non-starter for a longtermist who is succeeding in academic research. As a community, suppose we already have significant ongoing funding from 3 or so of the world's 3k billionaires. What good is an extra one-millionaire? Almost anyone's comparative advantage is more likely to lie in spending the money, on academic research, or otherwise.

Please contact me for those, and I'll forward them to JP as appropriate. Thanks!