Thanks for the writeup Oscar—I’m generally excited for more exploration around prizes in AI safety too. Will try to get more notes published, but briefly:
I think I’m a bit more bearish on things like “essay prizes” since the future fund era, but prizes overall still worth experimenting with, and ambitious versions could be good—AMCs especially
I think people who like prizes generally underestimate how important deal flow is—you talk about getting publicity which is definitely part of it, then there’s actively chasing potential prize submitters, lowering opportunity cost of participation, fast feedback, eois, etc
also how non important a big pot of money is—eg ACX book reviews are really good at surfacing good writing and talent, while the very high paid OpenPhil cause exploration/criticism prizes weren’t that good imo
Prizes are also great at surfacing talent. Manifold partnered with ChinaTalk for an essay competition, and they ended up hiring the winning writer as an editor I think. Dwarkesh also had this goal with his recent essay prize.
I also like “Nobel prize for ai safety”, like we get a panel of trusted folks to pick out who did the best tais, policy, fieldbuilding etc work on any given year. Could staple some money to it but prestige is probably the main prize—also such a thing could make safety more legible to outsiders.
Kind of weird, but prizes and prize judging are helpful for visas
Thanks for the writeup Oscar—I’m generally excited for more exploration around prizes in AI safety too. Will try to get more notes published, but briefly:
I think I’m a bit more bearish on things like “essay prizes” since the future fund era, but prizes overall still worth experimenting with, and ambitious versions could be good—AMCs especially
I think people who like prizes generally underestimate how important deal flow is—you talk about getting publicity which is definitely part of it, then there’s actively chasing potential prize submitters, lowering opportunity cost of participation, fast feedback, eois, etc
also how non important a big pot of money is—eg ACX book reviews are really good at surfacing good writing and talent, while the very high paid OpenPhil cause exploration/criticism prizes weren’t that good imo
Prizes are also great at surfacing talent. Manifold partnered with ChinaTalk for an essay competition, and they ended up hiring the winning writer as an editor I think. Dwarkesh also had this goal with his recent essay prize.
I also like “Nobel prize for ai safety”, like we get a panel of trusted folks to pick out who did the best tais, policy, fieldbuilding etc work on any given year. Could staple some money to it but prestige is probably the main prize—also such a thing could make safety more legible to outsiders.
Kind of weird, but prizes and prize judging are helpful for visas
Money can help bootstrap prestige.