Scenario B: ... the existing grant makers can just have a low-ish bar and this be fine because there are many strong applicants
the major problem w/ this is that grants are often disbursed within a bunch of random constraints, e.g. grants must be made with an overhead acceptable to the original funder, grants must be within the philanthropic remit of the organization giving them, grants should not be made in ways that unduly put the original grantmaker's reputation at risk, etc.
a lesser problem but a still important one is operational overhead. There's a reason that it's taken so long for exchanges to operate ~24/5, despite arguably having been able to automate a bunch of the backend related to trading for many years now.
legal risk / operational cost for a grantmaking operation increases ~ linearly with the number of grants, irrespective of size of the grants (and there are a bunch of hidden cost factors, e.g. if the grant is going to an individual, if the grant is going overseas, if the grant needs to be given with restrictions because you don't trust the individual very much). I think this is somewhat unintuitive to people.
Scenario D: There are not very many strong applicants but lots of money. It's not obvious what grantmakers should be doing in this situation. Sounds like there isn't much work to do—grantmakers should stop hiring and their people should switch to direct work. Or maybe not?
this is pretty reasonable
It seems like "whether active grant making works" is the key question for how good additional grantmakers are. My prior is that active grant making is very hard because most field building is very hard in my experience. In fact, maybe instead of active grant making, I would just go harder on funding normal field building efforts?
this is pretty reasonable
overall, maybe we're in scenario B (depends on a lot of things, including how effective you think 'defense in depth' interventions are that people who don't necessarily have high credence on advanced ai soon still have high conviction in).
If we're in scenario D, agree that things are pretty confused. total lack of clarity re: active grantmaking. you might suspect that e.g. evidence from the for-profit world cuts against the likelihood that active grantmaking works. IDK man, do think it's confused. There are definitely several live case studies here for successes of the active grantmaking model, e.g. arguably ProEquip, many convergent research FROs, some projects from the launch sequence, some cG teams have done this, Forethought has done this like once, etc.
Scenario B: ... the existing grant makers can just have a low-ish bar and this be fine because there are many strong applicants
Scenario D: There are not very many strong applicants but lots of money. It's not obvious what grantmakers should be doing in this situation. Sounds like there isn't much work to do—grantmakers should stop hiring and their people should switch to direct work. Or maybe not?
It seems like "whether active grant making works" is the key question for how good additional grantmakers are. My prior is that active grant making is very hard because most field building is very hard in my experience. In fact, maybe instead of active grant making, I would just go harder on funding normal field building efforts?
overall, maybe we're in scenario B (depends on a lot of things, including how effective you think 'defense in depth' interventions are that people who don't necessarily have high credence on advanced ai soon still have high conviction in).
If we're in scenario D, agree that things are pretty confused. total lack of clarity re: active grantmaking. you might suspect that e.g. evidence from the for-profit world cuts against the likelihood that active grantmaking works. IDK man, do think it's confused. There are definitely several live case studies here for successes of the active grantmaking model, e.g. arguably ProEquip, many convergent research FROs, some projects from the launch sequence, some cG teams have done this, Forethought has done this like once, etc.