John Marlow wrote:
> --- Adrian Tymes <wingcat@pacbell.net> wrote:
> > John Marlow wrote:
> > >
> > > > A: Become one of the super AIs.
> > >
> > > Q: How do you do this when the first AI decides it
> > > will be the ONLY AI?
> > >
> > > A: You don't.
> >
> > Q: How does the first AI in its earliest moments,
> > struggling to come to
> > terms with itself and learning how to manipulate the
> > world, keep other
> > AIs from coming into existence in the meanwhile
> > (which will probably be
> > at least several years)?
> >
> > A: It can't. Enforcement waits until it can -
> > providing a window for
> > people to upload, and if necessary, escape the
> > dominating AI's sphere
> > of influence.
>
> **I see people working to build an AI and making
> progress. I don't see the same progress toward
> uploading.
I do. In fact, I see more serious progress towards uploading than I do
towards AI - mainly because the upload research does not focus on
uploading per se, but on creating those technologies which will be
needed for uploading, while AI has not come to a consensus on why they
have failed thus far and started doing something about it.
> Besides--once the thing goes critical,
> you've lost control. That could take years, or
> seconds. You may not see it coming. The AI might
> conceal its progress, and so on.
My point was that the AI *would not be able to* conceal its progress at
first. You appear to have a concept of an all-knowing AI that knows
everything from the moment it is turned on. I am saying that, given
current research paths, such an eventuality is practically impossible,
at least without lesser AIs first being created and tested.
> **After all--would YOU, trust US?
Why, yes. As far as I had to.
> > > > Ok, so you'll have to still keep working on
> > saving
> > > > the world, providing
> > > > for you and your kin, and generally making
> > things go
> > > > 'round. If you
> > > > ever stop, you're superfluous to those who
> > haven't,
> > > > whatever form they
> > > > may take. This is true with or without AIs, no?
> > >
> > > No. A Rembrandt is not superfluous--whether he
> > > contributes to making things go round or doesn't.
> > >
> > > Same for a John Lennon.
> >
> > Ah, but they do. Art isn't superfluous: someone has
> > to dream the dream.
>
> **I suggest to you that art WILL be superfluous to an
> AI--and that anything which cannot fathom the value of
> art is unsuited to make our decisions for us. (Okay,
> maybe this is an eccentric POV; I think it's valid.
> And to appreciate art, you need emotions. And an AI
> with emotions is unsuitable to make our decisions for
> us. And...)
Some forms of art MAY be superfluous. Do you have any evidence or
logical proof that it will? Granted, I do not have any solid evidence
that it will not, except for the very limited AI I currently employ
that cares about nothing but music. (Specifically: a music generator,
more of an expert system than a sentient AI.) But, if that is the only
evidence on the table...
> > I'm talking about doing things for pure pleasure,
> > that produces nothing
> > for anyone outside of the doer. I.e., indulging in
> > luxury, and letting
> > pleasure, as opposed to providing value to others,
> > be the only thing
> > that drives you.
>
> **Ah--but Rembrandt died a pauper BECAUSE his works
> did not 'provide value to others' in his time. And
> that was in the judgement of his fellow humans.
> Imagine, if you can, what an AI would think of it.
Imagine, if you can, what a random person would think of it. Same dif:
the set of all AIs can potentially have as much or more variance as the
set of all intelligent humans currently living.
This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:56:19 MDT