Re: AI
On 8 Nov 2002, at 8:54, jbrown105 at speedymail.org wrote:
>
> On 0, Kat <kat at kogeijin.com> wrote:
> > On 7 Nov 2002, at 13:17, C. K. Lester wrote:
> >
> > >
> > > > I agree with you,the a.i. should have a way to modify its own
> > > programming...
> > >
> > > The only intelligent creature I know of on this planet does not modify its
> > > own programming, so why do you think a simulated/emulated/man-created
> > > intelligence will need to?
> >
> > Really? You know a non-human intellect that doesn't learn?
> >
>
> Ok, we need to define what it means to learn.
> (And yes, I know lots of non-human intellects that dont learn,
> but they are too dumb for our purposes ;)
>
> > > Now, I'm not saying that what you propose is ludicrous! By programming a
> > > creature that can program itself, you will have an incredibly efficient
> > > entity! Imagine if human beings could get into the wiring of our brain...
> > > kinda like in "The Matrix," where one disk read and WHAMMO, "I know kung
> > > fu." ;)
> >
> > Actually, it's NOT efficient at all. Self modification requires a much
> > larger
> > computer/brain, and it has many more events to calculate. I think you are
> > confusing "efficient" with "quantity of possible outcomes". The human brain
> > is
> > nowhere fast enough to operate any part of the human body (other than the
> > ears` logarithmic feedback mechanism) with the speed of a hummingbird's
> > wings
> > (and even then, the human ear can't respond that fast).
> >
> > Were you born with the instinct to type? Were you plugged into the Matrix to
> > learn it?
> >
> > Kat
> >
>
> There are 3 levels in the brain:
>
> basic internal information flow of the infrastructure underneath the
> intellengence -- preprogrammed in Eu hardcode for us.
>
> basic non changing instincts -- preprogrammed as the inital state of the
> brain for us. (perhaps non-changing is inaccurate.)
>
> learned behaviors -- created by the flow of inputs to ouputs through the
> internal information flow to enhance what was previously stored in the
> brain.
>
> 1 & 2 is what doesnt need to change. 3 is the part that does change,
> but its also the part that is not hardcode Eu, but rather a set of
> data structures manipulated by Eu hardcode.
>
> 1 in the human brain: rules on how neurons fire according to what
> signals they recieve.
>
> 2: the basic behaviors of crying and such that a baby knows as soon as it
> is
> born (not much compared to an adult, but much compared to a conventinal
> computer program.)
>
> 3: what is learned as the baby (and brain) grows up.
But the human brain is not one massive neural net, it is divided ineto
descrete sections, and a few have been mapped reliably, like sight, hearing,
speech, and various limb move areas. And in the frontal cortex, some areas
are specific to reason, some to inhibitions, and some to interface to the non-
thinking parts of the brain, like the autonomic system and the lower
instincts. I believe the parts not found yet which regulate these smaller nets
is what is learning new functions, not the nets. The nets are reprogrammable
to different patterns, but what determines when and how to reprogram them?
Kat
|
Not Categorized, Please Help
|
|
