No; the concept of Friendly AI depends on the concept of CEV, so a proper tabooing (in either sense) of the concept of CEV would affect that, too.
You know how your mind returns an ‘I don’t know what you’re talking about’ error when asked direct questions about your self, since it doesn’t use that concept? What I’m suggesting is that you remove concepts such that it returns that same error when asked direct questions about CEV, and re-answer DavidM’s question using only the remaining concepts. You can rebuild the CEV concept, if you want, but I suggest you do so in a way that allows you to rationalist-taboo it.
I am unable to remove the Friendly AI concept without destroying the concepts of “good”, “bad”, “value”, “person”, “worthwhile”, “preferable”, “person”, “concious”, “subjective experience”, “humanity”, “reality”, “meaning”, etc. the list just goes on and on and on. They’re all directly or indirectly defined in therms of it. Further, since without those there is no reason for truth to be preferable to falsehood with this removed any model of my mind won’t try to optimize for it and just turns to gibberish.
Ouch. Okay, the above advice is probably too late to be useful at all, then.
If those are all defined in terms of CEV (subjective experience? really? I’m not sure I want to know how you managed that one, nor humanity or reality), then what’s left for CEV to be defined in terms of?
Ok, granted, I used a kind of odd definition of “definition” in the above post, but the end result is the same; The model I use to reason about all LW type things (and most other things as well) consists of exactly two parts; Mathematical structures, and the utility function the math matters according to. The later one is synonymous to CEV as closely as I can determine. Every concept that can’t be directly reduced to 100% pure well defined math much be caused by the utility function and thus removing that removes all those concepts. (obviously this is a simplification but that’s the general structure of it.)
You have an interim definition that does not rely on CEV or FAI of personhood that you can use to distinguish the persons from the non-persons who appear before you, do you not?
So I’d be just saying “Friendly AI” instead? I don’t see how that’s going to change anything except being even vaguer.
No; the concept of Friendly AI depends on the concept of CEV, so a proper tabooing (in either sense) of the concept of CEV would affect that, too.
You know how your mind returns an ‘I don’t know what you’re talking about’ error when asked direct questions about your self, since it doesn’t use that concept? What I’m suggesting is that you remove concepts such that it returns that same error when asked direct questions about CEV, and re-answer DavidM’s question using only the remaining concepts. You can rebuild the CEV concept, if you want, but I suggest you do so in a way that allows you to rationalist-taboo it.
I am unable to remove the Friendly AI concept without destroying the concepts of “good”, “bad”, “value”, “person”, “worthwhile”, “preferable”, “person”, “concious”, “subjective experience”, “humanity”, “reality”, “meaning”, etc. the list just goes on and on and on. They’re all directly or indirectly defined in therms of it. Further, since without those there is no reason for truth to be preferable to falsehood with this removed any model of my mind won’t try to optimize for it and just turns to gibberish.
Ouch. Okay, the above advice is probably too late to be useful at all, then.
If those are all defined in terms of CEV (subjective experience? really? I’m not sure I want to know how you managed that one, nor humanity or reality), then what’s left for CEV to be defined in terms of?
Math?
Ok, granted, I used a kind of odd definition of “definition” in the above post, but the end result is the same; The model I use to reason about all LW type things (and most other things as well) consists of exactly two parts; Mathematical structures, and the utility function the math matters according to. The later one is synonymous to CEV as closely as I can determine. Every concept that can’t be directly reduced to 100% pure well defined math much be caused by the utility function and thus removing that removes all those concepts. (obviously this is a simplification but that’s the general structure of it.)
You have an interim definition that does not rely on CEV or FAI of personhood that you can use to distinguish the persons from the non-persons who appear before you, do you not?