>However, please try to find a way to express what you want to put into her mind in some way we could
later work with in a flexible way.
Only just that ehh? Should be easy-peasy right? I'll have that for you by next Tuesday Anon, will that work for you? :^)
Yes, I agree with you about the need for flexibility & extensibility (as I mentioned ITT). In fact, I think such a tall order will be literally impossible if that development approach isn't established and maintained during the entire process. In large part we're attempting to tackle something here that AFAICT, has never
been done before in human history. And to do it on cheap, commodity compute hardware too
? Tall order indeed.
I will just add to that as a side-note response to the above that I see a real benefit to us the Christian theological view of human beings, and distinguishing between the heart
of a robowaifu (where these actual laws will be 'written'), and the mind
of a robowaifu (where the more immediate physical responses, etc., etc.) are controlled. At the very least this separation of concerns will ease to some degree our ability to reason about these things. It's also a much higher-level take (analogically) on biomimicry
, which I think almost all of us here already admit to being a sound engineering approach generally.
>I think a lot of the learning could come from extracting the normalcy of human behaviors.
Yes, exactly. We are each striving towards building robowaifus that--at the very least for ourselves--have to engage effectively & pleasingly with human beings. While cultural & social """norms""" will have some bearing on that, our goal ITT should be a narrowly-scoped set of laws that adhere to a small set of basic Christian principles. Not only will this dramatically-decrease the set of tasks we have ITT, I believe it will establish a basic foundation of ethics & morals that can be used & extended by any Anon as he sees fit, while still adhering to basic modes of behavior generally-acceptable in the (now-dead) former Western Tradition
>So the minimal understanding would be to
be careful about doing things to living things in general, and especially when doing things with them
which normally are only being done to not living things.
>...but about knowing if it is harmful and abnormal
in the first place.
Quite a large taxonomic problem-space, wouldn't you agree Anon?
>I think these things can probably best be expressed as graphs. Definitions and relationships between
things, but also some ambiguity. The information for it, could come from some language models, before
having it checked by a human. Someone would have to try that out.
I think you're generally-correct about these points Anon. Thus my pursuit of Halpin, et al's
work on Object Role Modeling
) . I think it's probably our single best shot at managing such complexity via a highly-flexible approach that doesn't inadvertently hamstring the effort from the get-go by dint of limiting syntax (such as UML).
And you mentioned more than once that the robowaifu should ask for help. I certainly agree, and that obviously the primary way that we ourselves learn as children. However, that in itself is a basic problem that's mostly separate from this one. Tying this all together functionally, in a way that doesn't kowtow to the Globohomo Big-Tech/Gov
, is going to be a major breakthrough in general once it's achieved. The ramifications of that will go far
beyond just our beloved robowaifus! :^)