seem naive, maybe even ignorant, to all you highly intelligent folk
This seems pre-emptively hostile. I think you are intelligent, but you do also seem ignorant.
I like the direction of your questioning, it relates to one of my personal pet peeves, that of the distinction between “human values”, “humanity’s values”, “human friendly values”, and “humanity friendly values”, each of which seem very importantly distinct from one another, and only the last seems like a worthy goal for ASI.
But to clarify why you are getting downvoted, from my perspective: You seem ignorant of the field of AI Alignment and the way terminology is used within it.
(That is, aside from seeming to hold yourself and your opinions in very high regard while insulting other people’s ability to hold you in the regard you deserve, which, to a person with hierarchical obliviousness like me, seems like a trivial and unimportant flaw. It is possibly you genuinely mean your false modesty in good faith. Or maybe you just like playing the role of curmudgeon. It certainly seems like a fun role to me!)
About your ignorance of the AI Alignment field, to pick a specific example:
Greek “agape”, [...] The focus is not on self, or selfish ambitions, but on the Other, and their best interest.
If I understand you correctly, this is discussed in the book Superintelligence as “corrigibility” and “indirect normativity”. It’s actually not a simple matter to explain it to a computer, and is probably even more difficult than explaining it to a human.
In particular, I would recommend taking a look at infra bayesian physicalism which seems to be closely related to the problem of specifying the target of indirect normativity. Not that I want you to try to understand what they are talking about, but that I want to convince you that smart people are trying to focus on this, and it is a difficult problem.
I doubt you would, but please do not take my, or anyone else’s negativity as a sign you should stop engaging with these ideas. You do seem intelligent and well read, and I feel we need more people with cross-domain knowledge in this space, but in order to be “cross-domain” you do need to actually cross domains.
This seems pre-emptively hostile. I think you are intelligent, but you do also seem ignorant.
I like the direction of your questioning, it relates to one of my personal pet peeves, that of the distinction between “human values”, “humanity’s values”, “human friendly values”, and “humanity friendly values”, each of which seem very importantly distinct from one another, and only the last seems like a worthy goal for ASI.
But to clarify why you are getting downvoted, from my perspective: You seem ignorant of the field of AI Alignment and the way terminology is used within it.
(That is, aside from seeming to hold yourself and your opinions in very high regard while insulting other people’s ability to hold you in the regard you deserve, which, to a person with hierarchical obliviousness like me, seems like a trivial and unimportant flaw. It is possibly you genuinely mean your false modesty in good faith. Or maybe you just like playing the role of curmudgeon. It certainly seems like a fun role to me!)
About your ignorance of the AI Alignment field, to pick a specific example:
If I understand you correctly, this is discussed in the book Superintelligence as “corrigibility” and “indirect normativity”. It’s actually not a simple matter to explain it to a computer, and is probably even more difficult than explaining it to a human.
In particular, I would recommend taking a look at infra bayesian physicalism which seems to be closely related to the problem of specifying the target of indirect normativity. Not that I want you to try to understand what they are talking about, but that I want to convince you that smart people are trying to focus on this, and it is a difficult problem.
I doubt you would, but please do not take my, or anyone else’s negativity as a sign you should stop engaging with these ideas. You do seem intelligent and well read, and I feel we need more people with cross-domain knowledge in this space, but in order to be “cross-domain” you do need to actually cross domains.