@Aaron Bergman says
I wish there was (and there should be) more discussion of “which singular organization is most deserving of money rn” Individual donors should make their best guess public and indicate openness to critique
So let's guess. What is your suggestion?
In the AI governance space (specifically research; I'm not familiar with advocacy), I think GovAI and RP's AI Governance & Strategy [edit: now IAPS] team convert marginal money into research very well. I think this mostly because I like their teams and track records and (last I checked) they say they want to hire more but are funding-constrained.
But I wouldn't be surprised if some alignment or advocacy thing is predictably better.
I wouldn't be shocked to find it's a scrappy piece of anti AI advocacy like the AI pause work. Like maybe they get a big win and then it's been a few $k and very large impact.
https://manifund.org/projects/holly-elmore-organizing-people-for-a-frontier-ai-moratorium
Yes - if your timelines are short, then everything starts to look like it flows through a bottleneck of there actually still being a world in 2030 (2028, 2025..), which requires a global moratorium on AGI (as there is not enough time left for Alignment to be solved otherwise). There are a few people/orgs now working on this. Not sure which is the absolute best in terms of bang for buck though.
All of these are new (post-GPT-4): Centre for AI Policy, Artificial Intelligence Policy Institute, PauseAI, Stop AGI, Campaign for AI Safety, Holly Elmore, Safer AI, ... (read more)
Family Empowerment Media seems more cost-effective than GiveWell. They look especially promising from a cluster thinking approach.
If you agree you can donate.
Hi!
Thanks for this comment. We are interested in potentially funding family planning programs, including Family Empowerment Media. Our research and grant decisions are independent of our outreach and fundraising plans. To clarify and add detail on some of the points above:
GiveWell recommended a $500,000 grant to Family Empowerment Media in March 2023. This funding would support a planned randomized controlled trial (RCT) of its program. We haven't yet published about this grant, but we plan to soon. (Our grant page publication often lags our funding decisions considerably, though this is something we're working to improve!)
We believe that FEM or other types of family planning–related programs may be very cost-effective—our current rough estimate for FEM's program (not the RCT of the program) puts it at 18-28 times as cost-effective as unconditional cash transfers—but we have a high degree of uncertainty about some of our inputs and expect that the results of FEM's RCT would help us address some of our uncertainties. We are also continuing to assess how to prioritize family planning programs among the many other promising programs we're investigating.
I hope that's helpful!
Best,
Miranda Kaplan
GiveWell Communications Associate
"Deserving" feels off-- how about something like "which org turns marginal $ into the most EV"?
Maybe this is a bad question to ask - it sort of implies that our inside view is good and useful. But maybe it isn't. I don't know.
Genuinely makes me feel a bit queasy thinking about this which seems like a bad sign.
I wouldn't be shocked to find it's a scrappy piece of anti AI advocacy like the AI pause work. Like maybe they get a big win and then it's been a few $k and very large impact.
https://manifund.org/projects/holly-elmore-organizing-people-for-a-frontier-ai-moratorium
Yes - if your timelines are short, then everything starts to look like it flows through a bottleneck of there actually still being a world in 2030 (2028, 2025..), which requires a global moratorium on AGI (as there is not enough time left for Alignment to be solved otherwise). There are a few people/orgs now working on this. Not sure which is the absolute best in terms of bang for buck though.
All of these are new (post-GPT-4): Centre for AI Policy, Artificial Intelligence Policy Institute, PauseAI, Stop AGI, Campaign for AI Safety, Holly Elmore, Safer AI, ... (read more)