Re: Cool visual illusion


"feedbackdroids" <feedbackdroids@xxxxxxxxx> wrote in message
news:1134425035.352446.305930@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
>
> Allan C Cybulskie wrote:
> > So what you are saying is that our innate hardware limits the
"search-space"
> > for us so that we are not overwhelmed. Now, why do you think that such
> > innate systems are required for (artificial) intelligence? We have
> > computers now that can process millions of instructions a second. Could
the
> > need to reduce the search-space be a limitation of our physiological
> > hardware? If we off-load some processing to multiple components or
modules
> > and multiple threads, could we then search the entire visual range, say,
> > without any of the inherent limitations you talk about?
> >
> >
>
> Maybe with modern supercomputers, and gazillions of cycles to burn, the
> naiive learning approach will work out. However, so far it's hasn't in
> the past, and biology certainly didn't do it that way. And before you
> ask for the umpteenth time why that matters, the answer for the
> umptheeth time is that it's currently our only existence proof. No one
> has shown that having a gazillion cycles to burn will magically solve
> the problem.

If you assume that only the biological can be intelligent -- as you imply
above -- then you must accept that you cannot do AI since it WON'T be
biological. So give up now.

If you accept that things that aren't biological or human can be
intelligent, then your reply that the only thing that we know is intelligent
now are humans is an empty argument, since AI assumes that non-human things
can be intelligent and thus accepts implicitly that the infrastructure may
be different, and that different abilities may result.

>
> Regards the word "need", I think you have the concept in that sentence
> backwards.

Are you claiming that we don't need to limit the search-space? I thought
that was what you argued ...

>
> Regards the last, of course, theoretically infinite processing power
> can theoretically search an infinite search space, in finite [or
> infinite ??] amount of time. Before you answer, think what does "entire
> visual range" mean? Or maybe "entire behavioral repertoire". These
> terms don't make much sense.

Entire visual range -- which is the term I used -- means everything in view.
So, take a picture, and see what can be seen in it. Humans tend to focus
their attention, but a computer need not do so, at least to the same extent.
So the limitation we do of focusing our attention is at a minimum likely to
be less extreme with a computer than it is with us ... simply because of
processing power.

>
>
>
> >
> >
> > Since you want to talk about implementations, here's one that I've been
> > thinking about. Say that we have a bunch of different modules. When we
> > want to take an "intelligent" action, we have a central module that
kicks
> > off input to all the modules. It also determines a time limit on
getting
> > answers back. All the modules then respond either with more data or
with an
> > "answer" (which could be the same thing). When the time is up, the
central
> > module takes the answers it has and formulates an action from it. If it
> > can't formulate an answer, it waits for more data or reformulates the
> > question and sends it out again. For actions that have worked in the
past
> > consistently, it can store them in a "look-up table" to do them
immediately
> > (this would reflect conditioned automatic responses).
> >
> >
>
> Sounds like a rather common idea. I write s.w. like this now.

Do you think that AI can be implemented this way?

>
>
> >
> >
> > Would this be how the brain does intelligence? Well, it's possible, and
> > there is some evidence for such a system, but it doesn't look likely.
Would
> > it produce intelligence? Quite possibly. Could it process more than
our
> > brain does now? Depends on the hardware.
> >
> >
>
>
> Infinite processing power can theoretically process an infinite search
> space. But could it do it in a finite amount of time? I don't know.

This has no relation to what I said, which is that a system as described
above would do more than we could simply because of its design, and so the
limitation is not as severe.

But not that "the entire visual range" is not infinite, since for any
current sensing device there comes a point where the objects are so distant
that nothing can be determined about them. So it isn't infinite.

>
>
> >>
> > > > You will object that this is a long way from implementing Artifical
> > > > Intelligence. Quite so. But the reason for that is not that we don't
> > > > know how to build an intelligent structure, but that we don't know,
let
> > > > alone agree on, what's meant by intelligent behaviour.
> > > >
> > >
> > >
> > > This last is BS. This is not the limitation with implementing AI.
> >
> > Really? So tell me what "intelligent behaviour" is without resorting to
> > "it's what humans do". There are many rational fallacies that humans
make
> > that we wouldn't expect intelligent beings -- artificial or otherwise --
to
> > do.
> >
> >
>
> You and GS work it out.

We're trying. You're the one who's saying that you don't need that, and
that not knowing this definition is not the "limitation" (read: problem)
with implementing AI. So, put up or shut up. If you know so much about
what intelligent behaviour is, please educate us in what it is.

>
>
> > Oh, and if you know so much, tell me if emotions are required for an
> > intelligent being.
>
>
> Emotions are an integral component of biological organisms which
> exhibit intelligence.

But that doesn't mean that they are required for intelligence. It is a very
credible argument to claim that emotions impede intelligent behaviour, if
intelligent behaviour is considered have "rational" as a quality.

>
> But both of your comments here [as well as pretty much 100% of GS's]
> are off track. It's not yes or no, it's shades of grey. As long as you
> guys argue yes-or-no, N-v-N, etc, you'll just go round in the same
> little conceptual loops forever. No one is SUDDENLY going to produce an
> "intelligent being", yes or no. Some people have IQ=60, some IQ=80, and
> some IQ=100. Which person is "intelligent"? Stupid question.

Actually, it isn't a stupid question. I've raised it before. Computers now
can do "intelligent" actions that I and many others cannot do. For example,
some programs are much better at integrals than I am. So, are such things
intelligent? Is someone who can integrate not intelligent?

If we knew what intelligence meant, it wouldn't be an issue. But we don't,
so it is.

And, as usual, you miss the point. The first point claims that humans make
rational errors that we wouldn't expect intelligent agents to make. So
saying "intelligence is what humans do" fails because it would force our
intelligent agents to act irrationally, which is a ridiculous way to build
AI. In the second case, it IS a yes/no answer: does an intelligent agent
need to have emotions? The fact that you can't answer the question does not
mean it's stupid. It is critical, since if it isn't resolved if we build an
intelligent agent that doesn't have emotions we'll have to ask if it is, in
fact, really intelligent. Not what you want if you really want to build an
intelligent agent.



.

Relevant Pages

  • Re: Dresden
    ... Some think the hopes, dreams, emotions, of each of ... *Every* human person is of infinite value. ... Bob Harper ...
    (rec.music.classical.recordings)
  • Re: infinity
    ... >>> that if its true for any fixed limit L, it must be true for where there is ... It is NOT true that the set is finite without the finite size limitation. ... >>> If NO element is preceded by an infinite number of elements, ... elements are finite, and each element occupies a finite unit of that range, and ...
    (sci.math)
  • Re: Dresden
    ... Some think the hopes, dreams, emotions, of each of ... those human beings they were going to deprive of life, ... were innocent, do not mean anything. ... Some of less infinite value than others, ...
    (rec.music.classical.recordings)
  • Re: infinitely many nns = infinite nns?
    ... the LIMITATION that all of them are finite! ... The set FN itself is clearly infinite, by any of the usual definitions ...
    (sci.logic)