That said, I’d be wary duplicating existing programs; ie. the AI Safety Fellowship becoming a knock-off MATS.
This theme comes up a lot in AI Safety, and I really don’t think the reasoning is sound behind it. (See my post on a related topic).
Imagine you could snap your finger and create another organisation like MATS. Wouldn’t you want that, conditional on the org doing things just as well (or eventually becoming one that does things just as well)?
MATS is well-funded (having received a grant of over $30M recently, I believe), so it’s not as if they can magically absorb the money that could go to startup fieldbuilding projects. (Not to mention that smaller projects tend to be more cost-effective as long as they are good).
Imagine we lived in a world where 80k is still the only organisation doing career support.
Now we have HIP, Successif, Probably Good etc. These orgs are a blessing for the field, and if we could have twice as many of them, that would be great.
HIP and Successif focuse more on mid-career professionals.
Probably Good focusing on a broader set of cause areas; and taking some of the old responsibilities of 80k when it started focusing on more on transformative AI.
Yes, but there is still overlap in their work! It makes sense for orgs to find their nieche, but my stronger claim is that even if they didn’t, it would still be good to have double the amount of fieldbuilding orgs, assuming they are doing good work.[1]
This is dependent on funding availability, though—the background assumption here is that funders (OP) can’t give away money fast enough for some kind of fieldbuilding work (such as MATS)
If MATS was struggling for money, I would rather have them get the marginal dollar than another org that is doing something very similar but at an earlier stage. (but you could argue against this. One might want to invest into something speculative if they think it has the potential to outperform MATS on the long run)
This theme comes up a lot in AI Safety, and I really don’t think the reasoning is sound behind it. (See my post on a related topic).
Imagine you could snap your finger and create another organisation like MATS. Wouldn’t you want that, conditional on the org doing things just as well (or eventually becoming one that does things just as well)?
MATS is well-funded (having received a grant of over $30M recently, I believe), so it’s not as if they can magically absorb the money that could go to startup fieldbuilding projects. (Not to mention that smaller projects tend to be more cost-effective as long as they are good).
Imagine we lived in a world where 80k is still the only organisation doing career support.
Now we have HIP, Successif, Probably Good etc. These orgs are a blessing for the field, and if we could have twice as many of them, that would be great.
Sure, but these orgs found their own niche.
HIP and Successif focuse more on mid-career professionals.
Probably Good focusing on a broader set of cause areas; and taking some of the old responsibilities of 80k when it started focusing on more on transformative AI.
Yes, but there is still overlap in their work! It makes sense for orgs to find their nieche, but my stronger claim is that even if they didn’t, it would still be good to have double the amount of fieldbuilding orgs, assuming they are doing good work.[1]
(I think people who think this is wrong have the intuition of fieldbuilding being a zero-sum game, while in reality, we have a large amount of untapped talent, and orgs just don’t know how to reach them.)
This is dependent on funding availability, though—the background assumption here is that funders (OP) can’t give away money fast enough for some kind of fieldbuilding work (such as MATS)
If MATS was struggling for money, I would rather have them get the marginal dollar than another org that is doing something very similar but at an earlier stage. (but you could argue against this. One might want to invest into something speculative if they think it has the potential to outperform MATS on the long run)