Hiring is also hard if you're progressively eliminating candidates across rounds, because you never can measure the candidates you rejected. The candidate pool is always biased by who you chose to advance already. This makes me feel like I'm never collecting particularly useful data on hiring in hiring rounds. I don't ever learn how good the people I rejected were!

Isn't this avoidable? I could imagine a system where you allow a small percentage of randomized "rejected" candidates to the next hiring round and, if they properly succeed in the next, allow them into the third round. I have essentially no experience with how hiring works, but it seems to me that this could i) increase the effort that goes into hiring only moderately, ii) still sounds kind of fair to the candidates, iii) and would give you some information on what your selection process actually selects for.

Holden Karnofsky has described his current thoughts on these topics "how to help with longermism/AI" on the 80000hours pocast – and there are some important changes:

And then it’s like, what do you do to help? When I was writing my blog post series, “The most important century,” I freely admitted the lamest part was, so what do I do? I had this blog post called “Call to vigilance” instead of “Call to action” — because I was like, I don’t have any actions for you. You can follow the news and wait for something to happen, wait for something to do.

I think people got used to that. People in the AI safety community got used to the idea that the thing you do in AI safety is you either work on AI alignment — which at that time means you theorise, you try to be very conceptual; you don’t actually have AIs that are capable enough to be interesting in any way, so you’re solving a lot of theoretical problems, you’re coming up with research agendas someone could pursue, you’re torturously creating experiments that might sort of tell you something, but it’s just almost all conceptual work — or you’re raising awareness, or you’re community building, or you’re message spreading.

These are kind of the things you can do. In order to do them, you have to have a high tolerance for just going around doing stuff, and you don’t know if it’s working. You have to be kind of self-driven.

He goes on to clarify that today, he sees many ways to contribute that are much more straightforward:

[...]. So that’s the state we’ve been in for a long time, and I think a lot of people are really used to that, and they’re still assuming it’s that way. But it’s not that way. I think now if you work in AI, you can do a lot of work that looks much more like: you have a thing you’re trying to do, you have a boss, you’re at an organisation, the organisation is supporting the thing you’re trying to do, you’re going to try and do it. If it works, you’ll know it worked. If it doesn’t work, you’ll know it didn’t work. And you’re not just measuring success in whether you convinced other people to agree with you; you’re measuring success in whether you got some technical measure to work or something like that.

Then, from ~1:20:00 the topic continues with "Great things to do in technical AI safety"

This is an interesting idea, thanks for sharing!

While thinking about your suggestion a little, I learned about RUTF (peanut-based ready-to-eat meals used to treat child malnutrition) which appear to be rather established. Using the UNICEF 2024 numbers they distributed enough RUTF to feed half a million people throughout the year if we go by calories^[1]

According to UNICEF:RUTF-price-data a box of 150 meals (92 g which should correspond to ~500 kcal) costs them around $50 (using 2023 numbers and rounding up). Boiling this down to 2 000 kcal per day, this corresponds to a $1.33/day nutrition.

A few important aspects of RUTF are different than your suggestion (aimed at temporary treatment of malnutrition in children and not general nutrition for all ages, the prices above are not consumer prices, main ingredients differ, RUTF avoids the requirement of adding safe water which might not be easily available), but it seems to me that this supports your cost-estimate at least for large-scale purchases.

I think that together with existing consumer brands for meal-replacement powders, RUTF could be a second great reference for a similar and established idea :)

1. ^{^}

   Of course, RUTF is used as a temporary treatment so that the amount corresponds to a potential of 5.1 million treated children according to the linked dashboard

yes, totally agree 100% with this article—I have nothing to add.

I think it is good that (former) CEA has recognized that society has moved beyond distractingly specific names.

After Alphabet, Meta and X took the lead, I am excited about the impact that the Centre will create in adjacency to the EA community.

Update on the giving game contributions towards PlayPumps International: With some delay, we were able to find a way to donate the 2 € which were selected to go to PlayPumps by our participants.

That makes a lot of sense!

Consumers might not know or think much about the health aspects of things

This describes me quite well in many of my health choices, and unfortunately this is apparently really common.

potassium salt is 10x as expensive as normal salt

In my case, I also did not find salt that is pre-mixed at a price that makes sense to me - I bought a pharma-grade bag of KCl and mixed it with usual table salt myself^[1], which resulted in a net-price that is 3x of the usual sodium salt.

So it goes back to policy, and whether governments should just regulate sodium content even in salt - we didn't really explore this, given the higher evidence base and cheapness of salt policies.

That sounds very reasonable - I'll be looking forward to hearing about updates in the future!

That makes a lot of sense - in practice, there are many relevant considerations and other interventions might well be preferable in many contexts.

The expert opinion

[...] though a Chinese RCT does show positive results, and the current evidence is convincing, still more studies are needed, with the magnitude of benefit not as large as you would think.

also sounds as if potassium-enriched salt surely helps to some degree, but probably isn't a solution by itself. And I get the impression that research in the coming years will probably improve the uncertainties here.

Apart from this, I am a bit surprised that the costs ("perhaps double the price") would be a problem for richer countries. If I am understanding this right, this should still be obviously worth it as a health expenditure? A very simple estimate might be:

• lost expected life due to high blood pressure: ~2 years (scaling the DALY burden to a single person)
• expected gains from switching to potassium-enriched salt: ~1/2 year (I am guessing)
• expected costs: 80 years * 2/3 kg/year * $10/kg = ~$550
• resulting cost-effectiveness (assuming 1 year = 1 DALY): $1100 / DALY averted

Of course this isn't comparable to GiveWell effectiveness, but it is really cheap compared to other health expenses.