See for instance the stuff I’m funding. There’s a bunch of agent foundations, and the empirical work is more “people playing around and seeing what they find” than academic-style research.
I also like the work ARC is doing, though Paul doesn’t describe that as “agent foundations”.
My own “AI safety work” now also intersects a bunch with sociology, formal ethics, neuroscience, and psychology (e.g here, here). Meanwhile my “AI governance work” is starting from rethinking political philosophy (e.g. here, here).
Can you give a sample of those weird versions of AI safety?
See for instance the stuff I’m funding. There’s a bunch of agent foundations, and the empirical work is more “people playing around and seeing what they find” than academic-style research.
I also like the work ARC is doing, though Paul doesn’t describe that as “agent foundations”.
My own “AI safety work” now also intersects a bunch with sociology, formal ethics, neuroscience, and psychology (e.g here, here). Meanwhile my “AI governance work” is starting from rethinking political philosophy (e.g. here, here).