“Naturalized induction” is actually a mathmatical way of saying “AGI”.
Well, firstly, it’s not a complete agent in the slightest: it lacks a decision theory and a utility function ;-).
And requesting that there be MIRI(x) workshops on naturalized induction is requesting that MIRI include actual AGI design in its scope.
MIRI already includes naturalized induction in its scope, quite explicitly. There’s just not a lot of discussion because nobody seems to have come up with a very good attack on the problem yet.
Well, firstly, it’s not a complete agent in the slightest: it lacks a decision theory and a utility function ;-).
MIRI already includes naturalized induction in its scope, quite explicitly. There’s just not a lot of discussion because nobody seems to have come up with a very good attack on the problem yet.