What does the growth of EA mean for our priorities and level of ambition?

Link post

Lots of people have claimed that effective altruism hasn’t been growing in recent years. In a talk at the most recent EA Global, I argued that it has.

I then explored how this growth has changed the priorities for the movement, and argued that we should be more ambitious.

You can see the video by clicking here, or read a transcript below. The talk was 30 minutes, followed by a Q&A with audience-submitted questions.

It summarises several recent pieces of research, including Is effective altruism growing?, How are resources allocated across causes?, Why to be more ambitious, and Why we need more megaprojects.

I added an explanation of why the large amount of additional funding available doesn’t mean that it’s easy to fundraise (and why me talking about a ‘funding overhang’ was probably a mistake). A better framing is that there is a lot of funding available for any projects that can clear the current funding bar, but this bar is still pretty high.

Finally, I suggest in the talk that the recent success of Sam Bankman-Fried is an additional reason to aim high.

First, he shows that it’s possible to succeed. Back in 2015, perhaps only about 1,000 people were seriously directing their careers on the basis of effective altruism. And now one of them has made billions of dollars to donate, and become the world’s richest person under 30 — that’s not bad odds!

Second, the increase in the funding available to the community means that new projects that couldn’t have happened before should now be possible.

Watch the talk with the Q&A here.

Or you can read the transcript (without the Q&A):

Hey everyone. Today, I’m going to talk about why I think the resources available to the effective altruism community have been growing, how that might change our priorities, and why I think it means we should try to be more ambitious.

The first part of the talk is about whether effective altruism has been growing. Lots of people have been saying that effective altruism isn’t growing, and they cite things like this chart, which is the number of people searching for the term ‘effective altruism’ on Google. And it does indeed look like this isn’t growing, which is not ideal.

But to get a better picture of the growth of the movement, it’s useful to make a couple of distinctions. The first distinction is between the number of people finding out about effective altruism each year, and the total number of people involved — so that’s the difference between the flow and the stock. And then the second is, it’s useful to distinguish between different levels of engagement. So there’s the number of people who’ve just heard about effective altruism, the number of people kind of reading a lot, and then there’s highly engaged people. And people often call this the top, middle, and bottom of the funnel.

The search volume chart basically shows that the flow into the top of the funnel is not growing, and that is obviously bad — it would be better if all aspects of the funnel were growing. But if I was going to pick just one thing to focus on in order to evaluate whether effective altruism is growing or not, it would be the growth rate of the stock at the bottom of the funnel. And then that could then further be divided into a couple of different things we could look at.
You can think of the impact of effective altruism as a function of how many people there are and how much funding there is, where the efficiency with which that’s turned into impact depends on how good our ideas are.

So this is an illustration of that. As you get more funding and people, impact hopefully goes up, though at a diminishing rate. And then if we have better ideas, it shifts how much impact we’re able to have with each level of people and funding. There’s a lot more complications you could add to this — like there’s probably some complementarities between funding and people and so on — but just as a very rough illustration.

I’m now going to look at the growth in funding and then the growth in people. Earlier this year, I made these estimates of how much funding is committed to effective altruism in total. These are all very rough figures — they could easily vary by a factor of five, either side — but in total, I added up to about $46 billion. And maybe more strikingly, I estimated that that seems to have grown almost fivefold since 2015, which is a 37% annual rate of growth.

What drove the growth? One thing was that Dustin Moskovitz’s net worth — he’s the main source of the funds behind Open Philanthropy — has grown a lot, as Facebook stock has gone up and his new company Asana also IPO’d.

And then perhaps more strikingly is Sam Bankman-Fried, the founder of FTX and Alameda. We met Sam in 2012 when Will was giving a talk at MIT at a student group event. And he made the arguments for earning to give, and then Sam turned around and actually went and did it. And the Financial Times now estimates that he’s the world’s richest person under 30, and Forbes recently estimated his net worth is over $20 billion — actually gone up another five or six billion since I made the estimates a few months ago. I definitely wouldn’t have predicted this would’ve happened back in 2012 when we started 80,000 Hours.

It’s not just Sam though; all these other sources I think have grown a lot as well since 2015. So for instance, Founders Pledge in 2015 only had under $100 million dollars pledged when I made the estimate. I was using a figure of $3 billion pledged, and I’ve heard that’s now actually more like over $5 billion. So there’s a lot of growth in other terms as well.

That’s the stock of available funds. Just to talk quickly about the amounts of funding deployed each year, I estimated that’s around $400 million. It’s just important to remember with this one, it’s very lumpy from year to year. Open Philanthropy accounts for a lot of those funds, and you can see that that grew very rapidly from 2015 to 2017. It’s very lumpy each year, but then they’ve held it roughly constant from 2017 — but that was a deliberate decision to consolidate their existing cause areas, build up staff capacity. The plan with Open Philanthropy is that Good Ventures — so Dustin Moskovitz and Cari Tuna — want to spend down all the money before they die. And in order to do that, they need to hit something more like a billion per year fairly soon, so that’s definitely expected to go up. Sorry, ignore the 2021 figure, that’s incomplete.

So that was funding. Estimating the number of people involved is a lot more difficult. This is my favourite way to do it. This is probably not obeying best practice for slides, but the basic idea is you look at the Effective Altruism Survey, you see how many people say they’re engaged, and then you adjust it by the response rate. And the way you estimate the response rate is you take a group, like everyone who works at CEA, and then find out exactly how many filled out the survey, and then that gives you a proxy for that.

Applying that method in 2019, there were 900 or so people who filled out the survey saying they were very engaged, and then we estimated a 40% response rate, so there were 2,300 highly engaged members of the community. If you used a wider definition, then that number would be a lot larger, but just focusing on that very core at the centre. Applying the same method in 2020, around the same number of people said that they were engaged, but we estimated the response rate was a bit lower, so this method ended up with a 13% growth rate.

I’ve tried a number of other methods, and I generally end up with estimates between zero and 30%. Many ways of estimating it don’t really work, because they don’t take account of the fact that it takes people several years between hearing about effective altruism and getting involved. And so if you look at like 2020 data, it’s missing almost everyone who actually came in in 2020.

Just one interesting thing I was figuring out today is, I also estimated last year that around maybe 3% of people leave each year in that group of engagement. That’s also very uncertain, but if that holds up, then the stock of the community would eventually tend to around 10,000 people at equilibrium size, if the flow stays constant. So even if we don’t grow the number of people getting involved each year, this will naturally drift up to around 10,000 — though hopefully we can.

That was number of people and funding. I just want to do a final bit of this section, just to talk about how that’s allocated across issues. I also made these rough estimates of the breakdown of funding by cause area. This is mainly Open Philanthropy’s funding — averaged over a couple of years to take account of the lumpiness — and then GiveWell added in. Open Philanthropy is about 60%, then GiveWell 20%, and then I made rough estimates of the remaining 20%. And you can see the big thing here is there’s a big allocation to global health, which is driven by GiveWell.

The situation for people is actually really different. This is people who said five out of five for engagement on the Effective Altruism Survey — so this is the most engaged people. But you can see, firstly, they’re much more spread out over issues, and then these more meta things and AI are actually at the top when it comes to the most engaged people. This is actually exactly what we should hope for in some ways, because in a cause where you can hire people outside the community to do really useful work — like in global health — then you can deploy lots of funding quickly. And then people who are engaged in effective altruism should focus on the areas where it’s hardest to hire people outside of the community. For instance, in movement building, you can’t really hire people who don’t believe in effective altruism to do movement building for it. And same with AI alignment, because there’s not such an existing field there. You need people who are really into the ideas themselves.

So in some ways, we should expect quite different allocation for both funding and people. Some final speculations around what might be the ideal allocation of people across issues. This was a survey in the 2019 Coordination Forum shown in blue there, where people were asked what would be the ideal fraction of resources put into each area. And then also, shown in red, is users of the Effective Altruism Forum in the EA Survey — which is about 1,000 people — saying which their top cause preference was. And you can see those two actually line up quite well, which is interesting.

Now, I made this very made-up actual figure, which is a combination of people and funding, where I used a $100,000 per year conversion rate to get the total in each one. And then I’ve compared that to one estimate of the ideal portfolio. And what jumps out here potentially — though it’s very sensitive to exactly how you aggregate them — is that it seems like global poverty is a bit overallocated. I wouldn’t suggest we cut back funding to global poverty, but rather my hope is that the other causes will be able to grow faster over the next 10 years.

The other thing that really leapt out to me is the biggest gap is actually the bottom one, which is broad longtermism. That’s things like reducing great power conflict and improving institutional decision-making — efforts to make society generally better at dealing with big risks or other important challenges. And people in the survey thought that that area should have 5% to 15% of resources, depending on the person. But as far as I can tell, it’s under 1% currently, so on a proportional basis, that’s the biggest gap in the portfolio currently.

That’s the first section. The second section is about what might be some of the implications of this, and in particular, what are the implications of how much funding there is now? $46 billion — what should we be doing in response to this? And you could see there being two main options.

Option 1 is just to find the most scalable thing we can find, and spend the money as quickly as possible on that. Within more neartermist issues, that might be something like GiveDirectly or other cash transfers, which could absorb billions of dollars of funding and achieve a lot. Within longtermism, it might be something more like green energy research and development, which currently receives $35 billion per year of funding. You could add billions of dollars to that with probably similar returns to the current amounts. Maybe even better might be scaling up some existing biosecurity pandemic prevention methods on a billion-dollar basis. So that’s Option 1.

Option 2 is to try to find something even more effective than these. This seems to be what funders are doing currently — basically, funders are holding their bar significantly higher than GiveDirectly. And that’s because they believe that by waiting, we’ll be able to find and also create new opportunities that are significantly more cost effective than GiveDirectly, and therefore over the long term have a much bigger impact. I’d say the current bar of funding within Global Development is more around the level of Against Malaria Foundation, which GiveWell estimates is 15 times more cost effective than GiveDirectly. Generally, charities that are around that level of cost effectiveness, that level of evidence base, and good in the same other ways, have a good shot of getting funding.

Within longtermism, it’s a bit harder to say where the bar is. Maybe projects that have reasonable shots of doing something useful in AI safety or global catastrophic biorisks have a reasonable shot of getting funded, which you could think of as some of the recent things that have been funded by the Effective Altruism Long-Term Fund. The hope is that these things might be 10 or even more times more effective than something like green energy R&D. Meta-charities would then be expected to provide a multiplier on either of the other two.

To get a more concrete thing, there was a good post on the Effective Altruism Forum recently, which is called “List of EA funding opportunities.” It lists specific places you can get funding from, which also helps to give an idea of what funders are currently focusing on. One point I want to make about this is that these are hard bars to clear — green energy R&D might well be one of the best things within climate change, and we’re saying people are trying to find things that are 10 or 100 times more cost effective than that. And the same within global health versus GiveDirectly. It’s even harder to clear those bars if you take account of the fact that funders are going to be worried about things like the estimates being too optimistic, discount rates, opportunity costs, lots of other complications.

So a startup charity might well need to be targeting a level that’s even 10 times higher than those bars again, with the expectation that it will come down over time. The challenge that the community faces is, can we prove these funders right and actually generate all these opportunities that they’re hoping to find, and perhaps even expecting to find? And therefore have much greater impacts with this money than if we just donated it all now?

And so this means in terms of careers, there are these opportunities to have a lot more impact than something like earning to give and then donating to GiveDirectly. Which is already an extremely impactful level — I wouldn’t be surprised if something like earning to give and giving to GiveDirectly was more impactful than 98% of jobs that people take, and it’s really making a big difference in the lives of hundreds of people over the course of your career.

But because of this huge amount of funding, there’s an opportunity to go even beyond that. Here are some of the things that seem like priorities based on it. The first would be active grantmakers: people able to help find and also cultivate opportunities and projects that are above those funding bars that I mentioned before. If you’re earning to give currently, this could be the direction you could take your giving — can you get into the position where you could be like an angel donor in effective altruism and spotting things that other donors aren’t yet spotting? That’s a really valuable way of kind of adding value as a donor currently.

Secondly, we need researchers to figure out what these ideas for spending a lot more funding above the bar would be, or other ways to make the portfolio more effective. We need people founding projects that are more at this level of cost effectiveness. We need managers who are able to take the projects that are already there and help scale them up and reach their full potential. Then we also need people doing movement building to find all the people, those types of people above. There’s also a big need for feeder roles and supporting roles for all of these kinds of things — things like being a research assistant and all types of organisation-building roles, like operations management, office management, recruiting, things like that.

I would roughly say that if I had to put an estimate on it — so you could imagine we could magically create someone who’s earning to give donating certain amounts of money each year, or we could magically create someone who’s a good fit for one of these roles (especially the more senior versions) — I would often prefer the person to around millions of dollars of funding per year, in some cases over $10 million. That’s just a rough way that I would try to quantify the value of some of these positions in terms of earning to give; you could scale those figures down for the more junior equivalents.

Then based on this, I would say some rough priorities early career… Oh yeah, actually just before that… These things here are just the roles that impact has been changed by the funding situation. But there’s loads of roles that don’t require funding from the effective altruism community that are still really impactful. Some quick examples of those would be, people could have a lot of impact by improving things in government and policy, but they don’t require funding to do that — they can just take those positions. And I just want to flag, I’m not talking about all positions here, just the ones that have been changed by the funding situation. Just as another example, lots of people could work in journalism or the media, and promote really neglected and pressing issues and have a big impact doing that. And again, they don’t particularly need funding from effective altruism to do that necessarily.

Here’s very, very rough early-career priorities based on those roles. I’d say a big priority early would be trying to test out your fit for those roles, and then if you find a good opportunity, aiming at them. Secondly, there’s building career capital for those roles or building career capital for 20 years in the future when effective altruism will hopefully be a lot bigger and it’ll need a much wider range of skills — skills need to become broader over time. And then a third option is to just find a job that’s a good fit, and then on the side, don’t just donate, but you could also do community building and being a sensible advocate for some of these effective altruism issues. I think that can be a route to having just as much impact, and perhaps more than donating.

If you are currently earning to give, then I would say it’s worth seriously doing a round of applying to these types of roles and seeing whether there’s any opportunities that are a good fit — so seriously considering switching. It will be the best role for many people in the end.

In the past, I think I’ve helped to be fairly confusing a couple of times. One was by talking about talent constraints, the community being talent constrained. It makes it sound like it’s easy to get a job. And then more recently, talking about there being a funding overhang. It makes it sound like it’s easy to get funding.

Unfortunately, none of those two things are true. It’s more accurate to not use those terms. If you’re able to find opportunities that are much more effective than these funding bars — which are very, very high — then there’s a lot of funding available. But it’s not easy to think of something that’s 10 times more effective than something that’s already among the most effective things, which is earning to give to GiveDirectly, or the EA Long-Term Fund.

I should probably also be clear that allocation in the community is definitely not perfect. There’s going to be some good projects that are indeed above the bar, but getting missed. That’s always a risk. The way I would suggest orienting towards this is try to set up your life so that you can afford to fail — so make sure you have a backup plan, things like that. And then given that, try and aim as high as you can. The slogan would be, “Find ways to limit your downsides and then aim high.”

We can apply that on an individual level, and then I think also from the community, we can stand to be more ambitious. There’s many reasons why to be more ambitious. Partly there’s the general case that people who want to have an impact can afford to take more risks than people who are just in it for themselves personally — those kind of old-school effective altruism arguments.

Two new points — which I want to emphasise today — is that, firstly, I think recent events have shown that the community can achieve really, really big things by anyone’s standards. Coming back just briefly to Sam’s story: back in 2015, there were probably only around 1,000 highly engaged people in the community. So if one of those has become the world’s richest person under 30, that’s a pretty good base rate out of 1,000 people — those are pretty good odds. And I’ve seen people succeeding ahead of what I would’ve expected in other career paths as well — many people in effective altruism are just really capable of achieving a lot. And another interesting thing about Sam is, Sam had a lot going for him, and has a really privileged background — his parents are both Stanford law professors, I think. But one of his other really notable traits is that he aimed really high. I mean, he probably took the idea of becoming a billionaire more seriously than almost anyone else in the community. And I think that was maybe also necessary for his success. That’s showing that maybe we could raise our sights. It is possible.

The second point is that now there’s just a lot more funding available, so that should mean that we’re able to do things that just weren’t possible five or 10 years ago. In contrast to that though, a lot of effective altruism projects look quite similar to how they did back in 2015. Many of them would struggle to deploy more than $10 million per year effectively. And 80,000 Hours is definitely in that category. Our programmes are the same as before there was all this funding. But it seems like, looking forwards, we should be thinking about projects that are potentially a lot more scalable than our current ones.

If the challenge is we’re currently deploying $400 million a year, but we need to get that to above $1 billion a year, either we’re going to have to massively grow the community, or we’re going to have to set up projects that can deploy a lot more funding effectively than the current projects — like those higher bars of funding that I mentioned earlier. People have been talking about this idea of ‘megaprojects’ — this is just defined as a project that could cost effectively deploy $100 million a year rather than $10 million a year, which is where many of the current projects are.

I don’t know what these projects should be. There are some very rough ideas that have been floating around. One is something like a global screening programme to detect new pathogens to help with biosecurity. Another is an effective altruism–inspired school or university. A third would be to step into and try to become the biggest funders of nuclear security, because the biggest funder, MacArthur Foundation, has recently quit that area. So that could be another opportunity, and there’s a couple of talks about that at the conference.

That’s another way of seeing the challenge: what might these megaprojects be, and how could we get them started? It would be really exciting if something like this could emerge from this conference. I hope that is something to think about for the rest of the weekend. Thank you for listening.