The Future of Earning to Give

By PeterMcCluskey @ 2019-10-13T21:28 (+70)

This is a linkpost to http://www.bayesianinvestor.com/blog/index.php/2019/10/13/the-future-of-earning-to-give/

[I have medium confidence in the broad picture, and somewhat lower confidence in the specific pieces of evidence. I'm likely biased by my commitment to an ETG strategy.]

Earning to Give (ETG) should be the default strategy for most Effective Altruists (EAs).

Five years ago, EA goals were pretty clearly constrained a good deal by funding. Today, there's almost enough money going into far future causes, so that vetting and talent constraints have become at least as important as funding. That led to a multi-year trend of increasingly downplaying ETG that was initially appropriate, but which has gone too far.

Nothing in this post should be interpreted to discourage people from devoting a year or two of their life, at some early stage, to searching for ways that they can do better than ETG. A 10% chance of becoming a good AI safety researcher, or the founder of the next AMF, is worth a good deal of attention.

I'm assuming for the purposes of this post that nonhuman animals have negligible moral importance, since I'm mainly aiming at people who focus on human wellbeing. If I were to alter that assumption, I'd be a lot more uncertain about whether any specific cause should get more funding, but I'd also see many additional funding opportunities that look like low-hanging fruit.

I'll also assume that we should expect there will be less low-hanging fruit in the future, so that philanthropic money should be spent soon. That assumption ought to be somewhat controversial, and I'll discuss the alternative near the end of this post.

I will try to err in this post in the direction of being too pessimistic about our ability to distinguish good charities, in order to demonstrate that the value of ETG is not strongly dependent on the wisdom of donors.

Is there an efficient market in charity?

If there are well-endowed charities that are sufficiently wise and altruistic (Open Philanthropy and the Gates Foundation?), then maybe ETG is unimportant because we can count on them to do all the funding.

I find that less plausible than the idea that the two best VCs can fund all the startups that need funding. The VC world has better incentives than the EA world, and maybe better feedback. Yet I still see little reason for confidence that the right startups are being funded.

Also, philanthropy in general has a track record which suggests mediocre, but improving, efficiency. Those seem to be keys points of the original ideas that GiveWell, Peter Singer, and Will MacAskill made when helping to create the EA movement. It would be somewhat surprising if a pattern involving billions of dollars disappeared quickly after being criticized.

These reasons suggest we should have a clear prior that philanthropic institutions are not close to being as efficient as the stock market.

If we had wise billionaires who were funding all worthwhile charities, then most EAs should do direct work. But we should expect any really small group of funders to have quirks, blind spots, and selfish desires to avoid spending weirdness points. So we should expect them to leave funding gaps that can be filled by somewhat average EAs, and also plenty of need for further vetting by above-average EAs.

Near term opportunities

Here are some educated guesses about which EA charities can productively use more money this year:

• ALLFED
• GiveDirectly
• AMF
• CFAR
• Ought
• MAPS

Open Philanthropy's plans to fund at most half of the needs of most good charities creates a presumption that some of them won't be fully funded, unless there's some other large charity that seems willing and able to evaluate them. I see some large charities that partly qualify, but I don't see them as having a broad enough scope to fund all the opportunities of this type.

It's possible that some of these charities have been expanding as fast as their skill/manpower allows, and look underfunded because donors wait to donate until the charities need funds. But this seems to require a somewhat implausible degree of wisdom on the part of donors. I'm confident that the startup world hasn't worked that well, so why should it work better with young charities?

Maybe those opportunities will be fully funded soon. Many people should look a decade or so into the future before deciding whether to pursue an ETG strategy. Also, it would be nice to know whether ETG will become irrelevant if the Gates Foundation spends most of its money optimally and soon.

So I've decided that these near-term opportunities aren't all that important to my main point, and I'll focus instead on a longer term outlook.

Bigger, harder causes

I won't try to identify the best causes in this section. Instead, I'll describe causes that are beneficial enough and verifiable enough to justify an ETG strategy. I expect that by the time there are enough donors to fund the causes in this section, there will be more wisdom available to donors, who will find opportunities that are better than many of these. So please take this section as being somewhat closer to a worst case analysis than to a prediction of what EAs will fund.

One or two of these opportunities may surprise me by being cheap to solve, but I expect I've chosen hard enough problems that some of them will be expensive to solve.

Prizes

My first category is prizes for medical advances.

For example, an institution might offer $50 billion for a cure for aging, or for a general-purpose cure for cancer (maybe with partial payments for significant progress).

No, I don't mean offering rewards for a drug that would delay aging or cancer by a few months - there's plenty of money going into that already (mostly treating Western disease or a small subset of cancers). Practically all of that appears to be following a paradigm that shows little promise of curing aging. I mean something more audacious, in the sense that Aubrey de Grey's approach is audacious.

Producing a treatment that cures aging is likely more expensive and failure prone than is producing a drug that yields a small benefit in a large number of people, yet we've got a system that rewards the two about the same. That means there's little pressure to focus research on cures for aging. Prizes can be much more result-oriented than current funding of medical research, so I expect them to provide opportunities to redirect some of that research to the most valuable cures.

Aging is an area where it's unreasonably hard for most of us to evaluate whether any one research program is doing a good job, and I suspect that most investors in this area are doing poorly. But with prizes, the funders only need to be able to evaluate results well after they've been achieved, and whether they're giving the prizes to the people who are responsible for those results. There still needs to be somebody with skill at predicting which research will succeed, but it can be a VC-style expert who is reacting to good financial incentives.

Prizes aren't as efficient as direct grants to the best research programs, so it takes a fair amount of humility for a funder to choose prizes. I expect it requires an unusual person to adopt arrogant goals like curing aging, without also having arrogant beliefs about their understanding of which strategies are worth funding.

Sarah Constantin estimates that aging research is competitive with GiveWell's current top charities. I'll be pessimistic here, and estimate that it will be more like 1/20 as cost effective as GiveWell's charities, due to a combination of inherently low tractability and our poor ability to identify the best research strategies. That seems likely to look competitive after the next $10 billion of low-hanging philanthropic fruit has been picked.

Ok, some of you are probably saying I'm not being pessimistic enough. Maybe aging and cancer are intractable problems, and prizes for them won't attract any legitimate treatments. I don't have a simple way to convince you to trust my intuitions about their tractability.

It's likely that there are some other medical problems that are tractable, and where multi-billion dollar prizes would be productive.

If a cure for aging is intractable, then there's likely plenty of room for improvement in quality of life for the elderly. Note that hunter-gatherers seem to not get certain debilitating age-related diseases such as diabetes and dementia, and the elderly tend to remain active until a few days before death (see Lindeberg's Food and Western Disease).

Mental health seems like another area there's large room for improvement, but also large uncertainty about what strategies are tractable. Alas, I feel rather uncertain whether progress in mental health will be constrained by money or by something else.

I admit that the difficulty of choosing the right kinds of prizes weakens my argument by a modest amount. Yet even if half the prize money goes to poorly thought out goals, this approach will still shift medical research into directions that focus much more on maximizing benefits than is currently the case.

Affordable drugs

My next idea is Michael Kremer's proposal for drug patent buyouts (or see this version that's a bit more oriented toward laymen), under which a wealthy institution buys most drug patents and puts them in the public domain. This would dramatically reduce the problems associated with patent monopolies.

For example, drug companies sometimes try to sell drugs to poor countries at a relatively low price, and recoup their drug development costs by charging much higher prices in wealthy countries. Alas, that leads to drug smuggling. This makes it expensive to sell drugs in poor countries, likely leading to a situation where people in poor countries can't afford drugs that they would be able to afford if they could guarantee they wouldn't resell the drug. Patent buyouts can eliminate this perversity for a substantial fraction of drugs.

This strategy has some risks [1] if you try to implement it with a budget that's comparable to the market value of the drugs, so I'm reluctant to recommend attempting it with a budget as small as that of the Gates Foundation.

Global warming

Global warming is likely to cause widespread harm under standard forecasts, and we should worry more about small risks of a larger catastrophe from unexpected weather changes that might be triggered by warming.

There are a number of interventions that seem promising, such as preventing deforestation, reforestation, and albedo enhancement.

I don't know which interventions of this kind ought to be funded, but this seems like an obvious candidate for spending $10+ billion per year if we're running low on other philanthropic opportunities.

Seasteads / Charter Cities

For $100 billion or so, we could build a bunch of new territories that will provide people with more options to move away from regions with bad weather or bad governments. I.e. seasteads, or maybe charter cities if they're politically feasible.

In keeping with my pessimistic assumptions, I'll ignore the standard hopes for seasteads, and assume for this post that seasteads will mainly just provide real estate and fairly average governance, in order to enable the world's less fortunate people to lift themselves up to something close to the global average.

UBI

A universal basic income has some potential to protect against technological unemployment, and do a relatively efficient job of eliminating poverty, without the incentive problems of means-testing.

I'm not suggesting a political movement, since I'm trying to err in this post on the side of pessimism about our ability to identify good institutions, and it's too easy for political movements to end up being bent toward other goals.

This might be achieved by a large expansion of GiveDirectly.

There have been some concerns that GiveDirectly has problems with money going to positional goods. I'm unsure whether we should be concerned about that, but I expect those problems would diminish if GiveDirectly expands donations to give money to everyone in a given village, or larger region.

Manna is another example of a strategy that might lead to a UBI, although I don't want to endorse that particular organization.

Major Funders

Some of these causes may be taken on by major governments, making ETG irrelevant for those causes. But governments don't have a particularly great track record compared to the best charities. I'm betting that governments will either ignore some big causes of this nature, or will bungle the solutions in ways that leave needs for more EA money.

Why isn't the Gates Foundation funding these? Here are some guesses:

• the foundation expects to find enough better strategies to use up all their money before the low-hanging fruit is exhausted [requires moderate optimism about the foundation's abilities].
• the foundation expects that many of their recipients can't handle more money effectively today, but that some recipients will soon expand their ability to handle more.
• the strategies aren't prestigious enough, or look too weird.
• they are following standard philanthropic procedures for deciding where to spend money, and something like peer pressure has discouraged them from evaluating the alternatives.
• the foundation is unwilling to admit that there are important limits to how many projects their employees can supervise. Switching from direct grants to prizes would bypass some of those limits, at the cost of requiring more humility than I expect from someone who makes a career out of evaluating charities.
• See also some comments by Carl Shulman.

More low-hanging fruit next year?

How would the value of ETG be affected if, instead, we assume that we'll find better giving opportunities in the future?

It implies greater need for people to look for those opportunities, but still, the history of philanthropy suggests that only a tiny number of people succeed in creating a better opportunity than philanthropy previously had.

"Generate new charity" is much less amenable to decomposition into easy parts than, say, microprocessor design, so if many people attempt it, it will end up more like a competition to create the next Google or Facebook.

Maybe it's good for a hundred new people each year to enter some sort of competition to find/create the next EA charity, or try to do x-risk research, even if less than one per year will succeed. But most of these people should notice after a few years that they're not better at it than the people who created AMF, FHI, etc., and fall back on another strategy.

"The best way to save lives or reduce suffering" should be expected to produce a much narrower set of answers than "something consumers will pay for", so we should expect there to be a much smaller number of EA charities than good businesses.

Are EAs more productive at direct work?

I imagine that there a large fraction of EAs who expect to be more productive in direct work than in an ETG role. But I'm not too clear why we should believe that. The skills and manpower needed by EA organizations appear to be a small subset of the total careers that the world needs, and it would seem an odd coincidence if the comparative advantage of people who believe in EA happens to overlap heavily with the needs of EA organizations. Remember that EA principles suggest that you should donate to approximately one charity (i.e. the current best one). The same general idea applies to need for talent: there are a relatively small number of tasks that stand out as unusually in need of more talent. Talent is more complex than money, so it doesn't mean there's only one kind of talent that matters. But a heuristic which treats most talent as equally valuable seems as suspicious to me as a heuristic that treats all charities as equally valuable.

I can imagine that some of the hopes for doing direct work are due to selfish desires to signal one's EA credentials. That kind of selfishness seems fine if done honestly, but I want to keep that goal separate from altruistic goals.

What about vetting?

Don't we need more people doing vetting, and isn't that more like direct work than it is like ETG?

I'm fairly confident that a suboptimal amount of vetting is being done. But vetting more vettors doesn't seem any easier than vetting charities that do object-level work. If anything, it's harder.

As an analogy, I as an investor believe there's lots of money to be made by good VCs and VC-like funds, and I've seen a fair number of opportunities to invest in VC or similar funds. Yet I haven't invested in any such funds, because the ones that want more money are ones that I don't expect to be able to evaluate with a reasonable amount of effort.

When investors become eager to trust new VC funds, as often happens near stock market peaks, amateurs enter the field and run VC funds without acquiring much skill, and end up with poor returns on investment. If EAs were that eager to trust new people to disburse charitable donations, the same kind of problem would arise, and would be harder to detect, since charitable donors don't have feedback mechanisms that are as hard to fake as getting double their investment back.

Conclusion

There are some tough calls that need to be made, by anyone doing ETG, about comparing safe donations to donations that look more promising but run higher risks of biased evaluation. Still, there is a range of plausible-looking answers for which we ought to be moderately confident that ETG will remain valuable for quite some time.

Most likely there will be a trillion dollars or more in opportunities remaining for the foreseeable future. Trillions per year if we include a fully charity-driven UBI, but I'll guess that we'll end up instead with a patchwork of basic income programs that are funded by charities in some nations, and funded by governments in others.

My best guess is that the current marginal cost of saving a human life is around $5k to $10k, and that under mildly optimistic assumptions about the growth of EA style charity, it will rise to $100k in a decade or so. $100k is well within the range of costs that lead me to encourage ETG.

It's quite possible that an ETG strategy will produce poor results, but it sure looks like the most likely source of failure is poor choices of where to donate, not a shortage of low-hanging fruit.

Or maybe AI will render this all irrelevant soon. But that's not an ETG-specific risk.

Footnote

[1] - [Highly technical point, which most readers shouldn't worry about:] The joint randomization for substitutes works well if there's unlimited money to buy patents. I'm worried about what happens when a charity with a $10 billion budget tries to buy a patent that's worth about $5 billion. If patent holders with several $2 billion patents claim, with some exaggeration, that those other drugs are substitutes for the drug being bought, then the charity faces problems with either being unable to afford the purchase, or political fallout from unfairly(?) undercutting sales of some of the drugs. I'm unclear whether this causes problems in realistic cases.