Contra MacAskill on saving money for the intelligence explosion

Will MacAskill recently argued that large donors should invest money now, and wait until the intelligence explosion to give away their money. In a draft Forethought previewed with us and a few others, Will estimates a large donor might want to give 5% of their current philanthropic budget before end of 2029; we think it should be more like 30-50%.

Here’s why:

Philanthropic funding will grow faster than returns from investing in AI.

Will says that investing in AI could have really high returns in the next few years. True! But:

• AI-pilled future donors are already invested in AI.

  • I don’t think there’s a big audience of donors out there who believe the singularity is coming in 2029 but have all their money in cash under a mattress.

• There are a lot of donors with wealth currently illiquid in Anthropic, OpenAI, etc. who have committed to donating but haven’t yet.

• So if you take the simple model that your impact is proportional to the percentage of philanthropic funding you are, your dollars go further now.

  • Say you could donate a billion dollars this year and be half of philanthropic funding this year, or you could wait for your billion dollars to 10x. By the time your investment 10xes, the 100B/year waiting to come online—all in AI stocks—will also have 10xed, and your 10 billion will be a drop in a trillion-a-year bucket.

The world he describes is one where humans have already lost influence.

Will argues that by waiting, you can spend your grantmaking dollars more effectively on AIs that fully substitute for human labor. You can even “offload” running your foundation to “some aligned AI.”

I don’t know, waiting until AI is even better and then handing it hundreds of billions of dollars to solve alignment sounds like … we’re just letting AI take over?

AI isn’t fully substitutable, but it can already do a lot of tasks. I think we should spend money while humans still have some power in the world, and try to scale up the use of AI for good while we still have some control over it.

Donating now creates more donation opportunities in 1-5 years.

They do talk about this—Will says “if you’ve got … 3 categories of philanthropy, one is just what you would have been donating to otherwise, second is investing in order to donate more in the coming years, and then third is spending now in ways that will enable you to give more and more effectively, faster, at that crucial point in time, then I think things in that third category will often actually look the best.”

I kinda agree with this, but I think category 1 and category 3 are actually quite similar.

• Donating now seeds orgs.

  • There are lots of reasons it takes time for orgs to scale up: figuring out what to do, getting product-market fit, and building up internal trust and external legitimacy.

• Donating now gives you information and practice.

  • The more we give now, the more we can do retrospective evaluations and iterate based on what works.

  • In practice, I don’t think scaling from giving away small amounts to giving away huge amounts quickly works that well … the classic person who has done this is like MacKenzie Scott.

I think both of these things take calendar time, and the more you donate now the more you have time for them to come to fruition.

We can do better than giving an extra few hundred million per year to x-risk charities.

Will doesn’t want us to actually donate zero now. He does say that we should take advantage of the opportunities we see now, which are “maybe on the order of like hundreds of millions per year.”

To bring up just one obvious and large category, there will be $20 billion spent on the 2026 election, and probably a lot more spent on the 2028 election. Having wise leaders in charge of the American government during the intelligence explosion seems extremely valuable.

Or idk, give $100m to Manifund Regranting and see what happens! Buy Hampshire College! There you go, I spent a few hundred million dollars in 5 seconds.