I was going to write a comment responding but Neel basically did it for me.
The only thing I would object to is Holly being a āprominent member of the EA communityā. The PauseAI/āStopAI people are often treated as fringe in the EA community and the she frequently violates norms of discourse. EAs due to their norms of discourse, usually just donāt respond to her in the way she responds to others..
Yes, I do recall meeting at EAG two years ago and yes, I wasnāt giving any money to AI safety back then. I donāt have any criticism because I donāt know better about what policies to support and didnāt want to add noise.
I donāt want to be too critical of CAIP here and my original comment might have come off as too harsh. Iām not a big fan of the current funding ecosystem being so private with info and reasoning being so secretive. Itās caused us harm before and I expect will cause us harm again and I expect even greater it is causing passive harm where organizations donāt improve.
I really wish donors would write up their reasons for/āagainst funding things like Manifund does. It really doesnāt take long, so I donāt buy the usual ātime constraintsā argument, but I think the people at CAIP would waste a lot LESS time if they were told āwe donāt think your work is good/āmeets the funding bar. These other orgs are better at it for the cost and would rather scale them up than fund your organization.ā Itād waste a lot LESS grantee time. They can just add a disclaimer that this isnāt any way an abdication of their character or something.
That said, I looked at CAIP, and while you say here what youāve accomplished, it seems hard to verify that. I searched the EA forum and the internet, and only 2 people you list in your staff have posted on the forum here, and one more has an account. I didnāt do the same for LW or AF, but I expect similar there, maybe a bit more. Itās going to be very hard to get 7 figures per year for your non-profit if you arenāt very public about what you are doing. I know that takes time, but if you arenāt advertising the work you do, people wonāt know about it.
Here, in this post, you sort of just make a new account and say, āIf we donāt get $1.6M soon, we are shutting downā. Thatāsā¦ a lot of money. Itās hard to overstate how hard it is to make and then donate that amount of money per year. But more importantly, itād be very weird if this worked. There needs to be a lot more constant communication and updates from organizations if they want to continue to receive funding. It begs the question, what does the āCommunications Teamā do? I know they are probably primarily responsible for communicating with congresspeople, but at least some of that needs to be communicating with the EA/āAIS community.
How soon would you expect to see extremely fast economic growth? Say >20%/āyear GDP growth.
Another great post. I highly recommend this newsletter
Shartsis Friese was my law firm for something similar. They were pretty competent.
This is a great post. Way too many people in EA want to be able to do remote work behind a computer and not get into the thick of things that actually change things. I agree with your marathons and smores analogy; I donāt think I agree specifically with what you called the Levels for AI work but I digress since I like the main message.
Too often, I think people make the mistake that if they are working up the stack, they are getting a lot of leverage but we need 10x the amount of people at the lower trophic levels and without those, you donāt have any leverage at all. There are few exceptions to this of course, like MATS who produce a bunch of AI safety researchers but thatās directly increasing the number of ālevel 0ā people.
I wonāt say who it was (though they can out themselves) but someone convinced me that they do a donation strategy that I approve of. They donate to both sides to be able to lobby their congressperson on AI issues. I think this makes a lot of sense.
#changedmymind
First, sorry for the late reply. I thought I had sent it but it was still in autosave.
I had chatgpt analyze this paper on US House of Representatives. It finds that doubling spending for incumbents actually has ~no effect and in general you get about 4.5% increased win probability from doubling spending. You get the biggest gains for challengers with little name recognition. It also turns out that incumbents spend about $3M on a race and challengers spend about the same as well. So weāre talking about $3M to gain 4.5% extra chance of winning a house seat.
The paper goes on to explain that increases in spending faces increasingly diminishing returns.
To answer the question bluntly. Iāll just define past a certain point as 50% more than average spending. About 5% of races are ācloseā based on my crude metric of a margin of victory of less than 3 points.
Also, my criticism basically donāt apply (and in fact, I think we should be spending more money on) things like ballot initiatives and specific campaigns. Iām also much happier about things like primaries than general elections. If you are donating to just a generic race, even if itās close, I donāt think there is actually enough evidence that one party is much better than the other.
A lot of money is spent on politics already. Unless there is very very specific issues of EA concern, I donāt think itās worth donating to. There are tremendously good donation opportunities out there and political ads or Beyonce concerts arenāt among them IMO.
One thing Iāve been saying since the FTX days is that we absolutely shouldnāt be donating to politics in general (with exceptions of course but in broad strokes and definitely donāt donate to the Presidential race or a national party). The evidence that donations to political parties past a certain point do much is unclear (people only can see so many ads) and there are just much much larger problems in the world that EAs should be spending their money on, at worst you can just donate to AMF and this would do way better than most money spent on politics.
The exceptions would be that a very high integrity person was running who was a committed EA. In these cases, I think the math checks out but rarely will it for just some generic candidate.
Small bets allow you to get cheap signaling. Larger amounts force people to be at least a little more serious. FWIW, I offered Remmelt to do more.
It was difficult to come up with operationalizations for investment metrics. Remmelt wanted things like AI & ML VC deal activity being <30% and Anthropic valuation <$30B. I wanted things like % of YC startups being AI-related and didnāt want some large investment round in SpaceX or Stripe dominate funds raised.
Thanks for writing this. Iāve told other people this in the past.
Should EA avoid using AI art for non-research purposes?
To the contrary, you probably cost the AI labs a bunch of compute and this is the overwhelming effect.
Also, the environmental costs are tiny. If you are doing it for the environment, nearly all your environmental footprint is your diet and transportation, not electricity usage.
Sometimes rationalist-y people and lawyers really like to use/āinsist on maximally precise/āliteral language or legalese. I have a word for lawyers who do this in my business interactions with them; shitty lawyers.
Iām somewhat sympathetic to these people since it allows for exact communication and for some, it makes truthful information sharing easier. However, this just isnāt how the world works. What you point to simply isnāt fraud by the commonplace usage of the word. What you mean is that by your interpretation, they should advertise 285 piglets/ādollar and not 354 and that you donāt think their reason for stating 354 isnāt well supported.
Again, this over-the-top and inflammatory language is hurting rather than helping you and the animal welfare movement.
Vote power should scale with karma
To a point, maybe a bit less than it does currently but in general it seems to work well.
A great number of EAs in America/āEurope could learn a lot from you.
Remmelt, if you wish, Iām happy to operationalize a bet. I think youāre wrong.
Right but most of this is her āpre-AIā stuff and I am saying that I donāt think āPause AIā is very mainstream by EA standards, particularly the very inflammatory nature of the activism and the policy prescriptions are definitely not in the majority. It is in that sense that I object to Matthew calling her prominent since by the standard you are suggesting, Matthew is also prominent. Heās been in the movement for a decade and written a lot of extremely influential posts and was a well known part of Epoch for a long time and also wrote one of the most prescient posts ever.
I donāt dispute that Holly has been an active and motivated member of the EA community for a while