I don’t think you can avoid solving this one in the first place, so I wouldn’t call it dumping. (The problem is already inherent in the friendliness problem, so we’re not adding anything new to the friendliness problem by keeping this out of the logic problem.) The fact is that goals which have to do with external objects need to make use of “fuzzy” descriptions of those objects. There should be some reasonable way of formalizing fuzziness. (Of course standard fuzzy logic theory is woefully inadequate, but it has roughly the right goal, so we can re-use the name.) Utility needs to be a function of some “empirical clusters” in reality, specified independently from the underlying physics (or complete list of border cases) of those entities. Encoding a different logic into our logic is not much different from encoding a physics, so I think there is not a separate problem to be solved.
Very true! But that is a problem that already needs to be solved, separately. :)
Yes, and friendliness is already a hard enough problem, without people dumping the hard parts of other problems into it.
I don’t think you can avoid solving this one in the first place, so I wouldn’t call it dumping. (The problem is already inherent in the friendliness problem, so we’re not adding anything new to the friendliness problem by keeping this out of the logic problem.) The fact is that goals which have to do with external objects need to make use of “fuzzy” descriptions of those objects. There should be some reasonable way of formalizing fuzziness. (Of course standard fuzzy logic theory is woefully inadequate, but it has roughly the right goal, so we can re-use the name.) Utility needs to be a function of some “empirical clusters” in reality, specified independently from the underlying physics (or complete list of border cases) of those entities. Encoding a different logic into our logic is not much different from encoding a physics, so I think there is not a separate problem to be solved.