I’ve asked this several times before. As far as I can make out, no (published) text answers this question. (If I’m wrong I am very interested in learning about it.)
The CEV doc assumes without any proof, not just that we (or a superintelligent FAI) will find a reconciling strategy for CEV, but that such a strategy exists to be found. It assumes that there is a unique such strategy that can be defined in some way that everyone could agree about. This seems to either invite a recursion (everyone does not agree about metaethics, CEV is needed to resolve this, but we don’t agree about the CEV algorithm or inputs); or else to involve moral realism.
I’ve asked this several times before. As far as I can make out, no (published) text answers this question. (If I’m wrong I am very interested in learning about it.)
The CEV doc assumes without any proof, not just that we (or a superintelligent FAI) will find a reconciling strategy for CEV, but that such a strategy exists to be found. It assumes that there is a unique such strategy that can be defined in some way that everyone could agree about. This seems to either invite a recursion (everyone does not agree about metaethics, CEV is needed to resolve this, but we don’t agree about the CEV algorithm or inputs); or else to involve moral realism.