On twitter people sometimes say “but what has EA done”. This annoys me, so I am trying to get a list of actual soundbites, which I will tweet after editing.
Please comment with other lists, but:
On twitter, in cases like this, defensibility is king. We want clear sources. Who did what, where was the money from, how many X were saved
Be concise. A paid $X0000 for Y. This achieved Z
Let’s be honest about our failures too
Have played a substantial role in founding all three of the top leading AI capability companies (cry emoji).
What is the substantial role EA played in the founding of Deepmind? I find this claim surprising since Deepmind was founded in 2010. Or is Deepmind not a top three capability company according to you?
Shane Legg started DeepMind together with Demis after he heard about about Superintelligence concerns by Eliezer and Kurzweil, and was active for many months on LessWrong reading Eliezer’s arguments and commenting on the site.
See this tweet thread: https://twitter.com/ShaneLegg/status/1598047654159978496
And this old LessWrong account from Shane: https://www.lesswrong.com/users/shane_legg
...which predates EA? EAs don’t generally take credit/responsibility for intellectual influences pre-2011, even if those intellectuals then become active in the EA community. If we’re not taking credit for Peter Singer’s work pre-2011, I don’t think we should take responsibility for Eliezer or Kurzweil’s work pre-2011.
Sure, you can slice things here however you want. I do think it’s important to be at least consistent with both positives and negatives here. I’ve definitely seen people claim credit for GiveWell as part of EA, which of course was also founded pre-2011. I don’t think think there is a clear answer on how to handle this, and it seems most important to just be consistent.
IMO it does also feel really weird to say the sentence “we’ve had a substantial influence on 2 out of the 3 top AI capability companies”, when like, the same people also had a substantial influence on the third.
Yeah I think this is a legitimate type of entry to include on the list, though I’d like one for each and an actual description.
Anthropic: Founded due to xrisk concern, most of the staff are EA, most of the initial funding was EA, EA encouraged people to work there etc OpenAI: Founded on the back of superintelligence concerns, the FLI Open Letter (the last one!)and a behind closed doors conference organised by EAs. Much of the staff was EAs, and OpenAI was publicised and legitimised by EAs, including capabilities jobs on the job board etc. OpenPhil money got it going (maybe counterfactual would have done anyway)
I sense that the OpenPhil money was quite a small and easily replaced part of the raise. Is that not your sense?
That’s certainly what OpenPhil says about it. I don’t know enough to comment, beyond saying it clearly still legitimised the company and in general as said the EA energy and ideology definitely counterfactual I would say led to their foundation
Made their foundation maybe 5% more likely, perhaps. Led to it seems way too strong, right?
I’d say without ideas around AGI doom coming out of EA/Longtermist spheres and being popularised by people in that sphere ( Stuart Russell, Nick Bostrom, FLI etc), openAI probably doesn’t happen. So I’d say makes their foundation over 50% more likely. In this sense ‘led to’ seems pretty adequate
Isn’t Bostrom publishing that stuff before EA? Though would we take responsibility if it were positive?
With the AI XRisk stuff it’s hard to say what counts as ‘EA’ vs ‘proto-EA’; you are right though, it’s mostly the pre-EA bostrom-miri- rationalist cluster. I’ve lump them in with EA, maybe you don’t want to
I guess if we would lump if positive we should lump if negative
Has anyone written a more detailed account of this?
I also think to strengthen it ‘entirely founding Anthropic’ and as well as founding the other companies, continually work to legitimise them and their influence; only very recently did substantial amounts of EA start working against these companies
There are 6 good examples on the front page of effectivealtruism.org and there are more on its page about impact (which was updated last on January 7, 2021, so maybe it could benefit from an update)
I may break these out, but I think they are not great examples of what I want. Most require me to already trust EA sources or give quite vague impact.
Yes I see what you mean, thanks for taking a look
EA motivated donations - $142 Million to the Against Malaria Foundation[1]
GiveWell estimates it costs about $5 to deliver a bednet.
[2]
That’s about 30 million bednets.
GiveWell estimates that for every 1000 nets delivered about 1 counterfactual life is saved.[3]
So about 30,000 lives have been saved, by EA donations[4].
Thanks to @Holly Morgan for help here.
https://openbook.fyi/org/Against%20Malaria%20Foundation
https://www.givewell.org/charities/amf
https://www.givewell.org/cost-to-save-a-life
These bed net deliveries will have been partly supported by other organisations, eg the Global Fund, which often covers shipping costs—https://www.givewell.org/charities/amf#Does_it_work
Moreover, GiveWell estimate that “GiveWell-directed donations to our recommended charities between 2009 and 2021 will avert over 159,000 deaths” (accounting for all top charities, not just AMF).
This feels a bit insider baseball to me. If someone doesn’t trust GiveWell why will they trust this number.
If they don’t trust GW, why would they trust your calculation, which also rests on GW’s analysis? Here’s a spreadsheet with GW’s analysis: I think that the 159,000 figure is just them doing a pretty similar BOTEC to yours, but across all of their top charities (and they seem to have a higher figure for total donations to AMF, not sure what’s driving the difference there).
Because in my case the GW analysis is one step further back in the calculation and attached to things that one can more easily grasp.
In your comment, the reader can go “oh GW is an EA thing of course they’d say that”
In mine they have to dispute that nets save 1 in 1000 lives or that nets cost $5 each. While GiveWell is backing up those numbers the concepts are much easier to grasp so the trust to GiveWell is much less.
Now, this is really optimising for the small chance that someone will dispute, but in the case that they do,
Yours “Why should we trust givewell” seems hard to answer
Mine “Why should we beleive that 1000 nets save a life” seems trivial “well do you think that number seems way off .. seems about right to me”
Note to the reader. I think the concept I’m trying to pin down here is worth noting. How well does your argument deal with the reflexive responses of someone who was always going to disagree. Here, without dubious epistemics I railroad them into disagreeing with pretty concrete facts. This is not a trick, AMF just does really clear work—you either have to disagree with the costs or the effectiveness of bednets
~30,000 lives at ~$4,500 a life? (https://www.givewell.org/cost-to-save-a-life)
Thanks!
“The estimated the size of the movement to fight factory farming in 2020 was ~$200 million USD” (https://farmedanimalfunders.org/wp-content/uploads/2021/12/External-FAF-State-of-the-Movement-Report-2021.pdf)
Open Phil granted ~$25 million to farm animal welfare in 2020 (https://www.openphilanthropy.org/grants/?q=&focus-area%5B%5D=farm-animal-welfare&yr%5B%5D=2020)
“A decade ago, most of the world’s largest food corporations lacked even a basic farm animal welfare policy. Today, they almost all have one. That’s thanks to advocates, who won about 3,000 new corporate policies in the last ten years” with 88% following through. “[A]lmost every new farm animal welfare law and corporate policy globally has focused on chickens” e.g. “advocates secured cage-free pledges from almost all of the largest American and European retailers, fast food chains, and foodservice companies.” (https://us14.campaign-archive.com/?u=66df320da8400b581cbc1b539&id=06c9567d81)
(Not very soundbite-y I know, but I think it conveys that when it comes to farm animal welfare, change is happening and a decent chunk of that is “EA.”)
From abier:
I didn’t know this
From Ula Zarosa:
Defrauded N FTX users of a total of $X.
Or
Stole $X from N FTX users.
(Fill in N and X).
So I think “encouraged/justified a fraudster who used $8bn of customer funds fraudulently. Currently it looks like about 2/3rds of that will be returned”
FTX claims look to be selling for about 20 cents on the dollar, so 2⁄3 seems optimistic
e.g., https://www.x-claim.com/
Open Phil granted the Humane League over $45M, whose support for CA’s Prop 12 reduced the suffering of 40M animals.
I probably wouldn’t single out THL this way. Many orgs were involved in supporting Prop 12, and I don’t know that THL was leading the efforts (I would guess HSUS, if I had to guess one org, but with low confidence). I’m also not sure THL would claim their support in particular reduced the suffering of 40M animals. Instead, they supported Prop 12, and 40M animals were affected by Prop 12 so far (at the time), at 40 million hens/year (Vox, quoted here by THL).[1]
Also, little of that $45M would have gone towards Prop 12 work, as the vast majority of that was probably for corporate campaigns/outreach. Open Phil granted $4M for work on Prop 12 specifically leading up to it, and another $250K more recently to defend it at the Supreme Court, neither specifically to THL. Open Phil seems to have provided less than 1/3rd of the funds in support of Prop 12 (at least directly earmarked for Prop 12).
On the other hand, they might not mean it counterfactually, and could be comparing to the world without them and where no one replaced their work, and they really do mean 40 million this way. If someone needs CPR, and there are two people available, ready and willing to perform CPR, but only one does, then the one who does perform CPR gets to say they saved the person needing CPR, even if this person would have been saved by the other anyway.
Or, THL might actually mean it counterfactually and believe it probably wouldn’t have passed without them, even if others would have done more to make up for (some of) what THL did.
I haven’t seen any attempts to justify either claim, and it seems more likely they’re just using Vox’s 40M animals number directly.
Thanks for your strong caveats, supported by deeper research!
There’s probably a much more rigorous statement to be made regarding THL’s impact and EA funding which led to accomplish it, but my above comment wasn’t it.
https://www.alvea.bio/alvea-set-the-record-for-the-fastest-startup-to-take-a-new-drug-into-a-phase-1-clinical-trial-from-founding-to-first-in-human-date/
https://forum.effectivealtruism.org/posts/oGdCtvuQv4BTuNFoC/good-things-that-happened-in-ea-this-year