Re: What is missing from AI ?


> > And any "goals" it seems to have, is just a
> > description of how it acts when it is turned on.
> > In general, the default behavior is to produce
> > all behaviors equally.
>
> But it doesn't produce all behaviors equally. It has
> its own internal emergent behaviors which make some
> i/o actions impossible and some almost a guarantee.
>
> Mind you I think a lot of the brains behavior
> is emergent in nature and *relies* on this being
> the case.
>
>
> > It's just as likely to send a pulse to the
> > "raise arm" output as it is to send a pulse to
> > the "lower arm" output. The odds of it sending
> > 100 pulses in a row the "raise arm" output is
> > very very small. But if you wait long enough,
> > it will happen.
>
> If this is how the brain worked it would never
> have had time to become intelligent. There are
> mechanisms and algorithms that evolution must
> have "discovered" that allow brains to quickly
> home in on a solution.


Hi John. Just a note regarding general issues of responsivity and
temporality. I'm not sure that either Louis or Curt has addressed this
in their networks. This is not so much about quickly homing in on a
solution as how to have a huge complex net that can also respond
quickly to changing real-world events.

Basically, this involves what are known as "small-world" nets, studied
by Strogatz and others. Take a look at the link to Corbacho et al on
the other thread. In short, you can have huge #s of interacting cells,
arranged in local clusters [cf, modules], and you can quickly
synchronize activity in a global manner across the clusters by adding
as few as 5 connections per 10,000 cells or so, added at random between
the various clusters. IOW, it actually takes a very small amount of
interconnection to radically improve the responsivity of the network as
a whole, as well as synchronizing gross activity in the myriad
clusters, while still allowing the individual clusters to engage in
their particular forms of activity.
A very interesting theoretical result, and as you can see, this is very
much related to the modular organization of the brain, overlaid by many
fiber pathways [white matter] interconnecting the modules.

.

Relevant Pages

  • Re: The Road with no Branches argument
    ... >(through the action of their more complex brains) simply have better behaviors ... Dennett dismisses this idea as silly. ... Dennett tries to substitute control for choice. ...
    (sci.logic)
  • Re: The Road with no Branches argument
    ... When we say that a person chooses among several possible behaviors is there ... (through the action of their more complex brains) ... Dennett dismisses this idea as silly. ... Dennett tries to substitute control for choice. ...
    (sci.logic)
  • Re: The Road with no Branches argument
    ... >>(through the action of their more complex brains) simply have better behaviors ... Dennett dismisses this idea as silly. ... >>If our mechanical brains are in control of our behavior and our brains produce ...
    (sci.logic)
  • Re: What is missing from AI ?
    ... > But it doesn't produce all behaviors equally. ... > its own internal emergent behaviors which make some ... > have "discovered" that allow brains to quickly ... main evolutional engine spanning millions of years. ...
    (comp.ai.philosophy)
Loading