ACX Grants Results

Astral Codex Ten Podcast

https://astralcodexten.substack.com/p/acx-grants-results

Thanks to everyone who participated in ACX Grants, whether as an applicant, an evaluator, or a funder.

Before I announce awardees, a caveat: this was hard in lots of ways I didn't expect. I got 656 applications addressing different problems and requiring different skills to judge. I'll write a long post on it later, but the part I want to emphasize now is: if I didn't grant you money, it doesn't mean I didn't like your project. Sometimes it meant I couldn't find someone qualified to evaluate it. Other times a reviewer was concerned that if you were successful, your work might be used by terrorists / dictators / AI capabilities researchers / Republicans and cause damage in ways you couldn't foresee. Other times it meant it was a better match for some other grant organization and I handed it off to them.

Still other times, my grant reviewers tied themselves up in knots with 4D chess logic like "if they're smart enough to attempt this project, they're smart enough to know about XYZ Grants which is better suited for them, which means they're mostly banking on XYZ funding and using you as a backup, but if XYZ doesn't fund these people then that's strong evidence that they shouldn't be funded, so even though everything about them looks amazing, please reject them." I have no idea if things really work this way, but I needed some experienced grant reviewers on board and they were all like this. I took these considerations seriously and in some marginal cases they prevented funding.