Agreed there are many bright young people who want to work for EA charities. CG doesn’t seem to have that much trouble getting these folks to apply (especially now that there are opportunities to test your fit via something like the Astra fellowship).
Relatively senior (6+ years of work experience?) AI-safety-focused people who can manage others, scope out programmatic foci and strategies, and build teams are much more of a bottleneck. Of course, this is true throughout the AI safety ecosystem, but it seems particularly tricky for grantmaking, and makes it harder to hire a bunch of eager young generalists.
Speculation: Maybe people who are more experienced are less open to trying new types of work, and grantmaking is fairly unusual?
Grantmaking means giving someone decent amounts of influence over large amounts of money, and orgs are (reasonably!) pretty cautious about extending that authority.
Compare this to something like research, where you can hire a researcher and have them produce work that gets reviewed before it goes anywhere.
With a grantmaker, the stakes of a bad judgment call are higher and can be harder to quickly review if you’re a time-constrained manager.
This makes orgs slower to hire and slower to ramp people up to full autonomy than they would be for other roles.
CG has probably just straight up made mistakes around hiring too slowly in the past (and possibly the present). Ideally a single organization making a mistake like this wouldn’t affect the ecosystem so much, but CG is such a large portion of the total amount of grantmaking done in the AI safety ecosystem ATM.
Some thoughts:
Agreed there are many bright young people who want to work for EA charities. CG doesn’t seem to have that much trouble getting these folks to apply (especially now that there are opportunities to test your fit via something like the Astra fellowship).
Relatively senior (6+ years of work experience?) AI-safety-focused people who can manage others, scope out programmatic foci and strategies, and build teams are much more of a bottleneck. Of course, this is true throughout the AI safety ecosystem, but it seems particularly tricky for grantmaking, and makes it harder to hire a bunch of eager young generalists.
Speculation: Maybe people who are more experienced are less open to trying new types of work, and grantmaking is fairly unusual?
Grantmaking means giving someone decent amounts of influence over large amounts of money, and orgs are (reasonably!) pretty cautious about extending that authority.
Compare this to something like research, where you can hire a researcher and have them produce work that gets reviewed before it goes anywhere.
With a grantmaker, the stakes of a bad judgment call are higher and can be harder to quickly review if you’re a time-constrained manager.
This makes orgs slower to hire and slower to ramp people up to full autonomy than they would be for other roles.
CG has probably just straight up made mistakes around hiring too slowly in the past (and possibly the present). Ideally a single organization making a mistake like this wouldn’t affect the ecosystem so much, but CG is such a large portion of the total amount of grantmaking done in the AI safety ecosystem ATM.