Re: Cool visual illusion

"Allan C Cybulskie" <allan.c.cybulskie@xxxxxxxx> wrote:
> "Glen M. Sizemore" <gmsizemore2@xxxxxxxxx> wrote in message
> news:43a56821$0$16727$ed362ca5@xxxxxxxxxxxxxxxxxxxxx
> >
> > "Allan C Cybulskie" <allan.c.cybulskie@xxxxxxxx> wrote in message
> > news:7Zbpf.353$1Y4.3148@xxxxxxxxxxxxxxxxxxxxxxxx
> > >
> > > "JGCASEY" <jgkjcasey@xxxxxxxxxxxx> wrote in message
> > > news:1134709306.936825.199950@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> > >>
> > >> Allan C Cybulskie wrote:
> > >>
> > >> > But I don't WANT to understand the hardware.
> > >> > I want to understand the software.
> > >>
> > >> What is the difference in terms of behavior?
> > >
> > > I never said there was a difference in terms of "behaviour". What I
> mean
> > > is
> > > that I want to know how it works independently of the hardware it's
> > > running
> > > on so that I can implement it on any hardware powerful enough to run
> > > it. And that was the point. I don't particularly care how the brain
> precisely
> > > runs intelligence or what components or areas of the brain are
> > > involved
> in
> > > what actions. I really only care about the functionality, not the
> > > particulars.
> >
> > How do you know that this analogy makes sense in circumstances divorced
> from
> > the digital computer?
>
> I fail to see why this matters. Call it the algorithm instead of the
> software and my point is even more clear. The reason I use the
> "hardware/software" divide is to make it abundantly clear that the
> processors or the biology is not what I really care about. The
> "hardware" that it runs on -- the brain, a computer CPU, a hypercube --
> is not (in my opinion) particularly relevant to AI. A good AI algorithm
> can be implemented on "hardware" that is not a brain.
>
> Consider an analog computer that finds a place that is
> > centrally located to a bunch of towns. Briefly, one takes a map, tapes
> > it
> to
> > a table, and drills holes through the towns' locations. One then passes
> > strings through the holes and ties weights to them beneath the table.
> > The other sides of the strings are then tied to a small ring. When the
> > ring is released, the weights pull it to an area that corresponds to
> > one that minimizes the squared deviations. What is the "software" and
> > what is the "hardware" here?
>
> The hardware is the "analog computer". The software is the design and
> the algorithm of finding a central location. To make it clear, the
> hardware is what stays with this implementation, and the software is what
> can be implemented elsewhere and in different ways. You don't assume
> that there is no way to implement finding the central location other than
> this set-up, right?

That's just not a very good way to describe what I believe you are really
thinking.

What you refer to as the "software" is just a limited subset of our
understanding of some physical device. The "software" or "algorthm" you
refer to, exists in our heads, not in the device we are studying or
describing. And only the subjet of our understanding which we consider
important to some current goal, is the part you talk about as the
"algorithm".

For example, I could describe in my head the important phsical
characteristis of a box as being a squre container with inside dimenstions
of 6" x 6" x 6" and with one side open.

The above description is the "algorithm" of the box. Those few words
completley describe all the functional characteristics of a box that I
happen to be interested in. I could build such a container out of wood, or
plastic, or cardboard, and each would have very different physical
characteristics. But yet all of them share in common my specification
because it's only a limited subset of all the physical characteristics I am
interested in.

We know that likewise, the intelligence in AI is a subset of all human
beahvior. We just don't know how to define that subset yet. If we were in
fact only interested in full human beahvior, then the only way to reproduce
it would be with a human. That's because full human behavior includes
things like bleeding human blood when you are cut. I doubt that anyone
working on AI feels that "bleeding human blood" is now, or every will be,
part of the functional requirements of AI.

In the example which Glen gave with the analog computer, the "algorithm" in
that case could be specified with the language of math as Glen has already
done. As in, "A device which locates the point on a map where the sum of
the square of the differences is minimized between three cities and the
point".

There is no way to find an "algorithm" for a device by itself. Because the
"algorithm" is an arbitrary subset of the full set of characteristics of a
device picked by the needs of the observer. Glen's map device might also
work sincely as cat toy and I might choose to define that as, "a table with
3 weighed strings where pulling any one string will make the others move".
That's the "algorithm" for my cat toy.

So what Allan is trying to describe as the "algorithm" of AI, is simply
some subset of human behavior which no one has yet figured out how to fully
specify. And that algorithm will never be independent of the observer
because we (as a culture/society) have to agree on what subset of human
behavior to assign to the defintion of the words "intellgent behavior".

And though we don't normally use the word algorithm outside the scope of
determinstic state machines, the idea which Allan I believe was really
getting at does exist outside that scope - it's just the "functional
specification" of a machine. It's the subset of it's full set of physical
characteristics we care about. It's the subset we would find in common
between all machines of the class defined by that functional specification.
It's the part of the behavior of the machine we care about.

--
Curt Welch http://CurtWelch.Com/
curt@xxxxxxxx http://NewsReader.Com/
.

Relevant Pages

  • Re: Havent done anything real with OOP yet.
    ... But it is an algorithm description and, ... simulation needs. ... hardware to/from the network port. ... driving the abstraction of a rather specialized chunk of hardware. ...
    (comp.object)
  • Re: qubits
    ... > the available hardware states), in which case the program might run ... Shor's factorization algorithm, tough, was explicitly tought for ...
    (comp.object)
  • Re: OT : Approximate / fast math libraries ?
    ... some practical work with the algorithm will tell you the real results. ... It costs no more to calculate doubles than singles except ... depends on the hardware implementations inside the processor and the ... cycle I put into it, along with some additional benefit from hardware ...
    (Fedora)
  • Re: Question about re-designing an algorithm when hardware architecture is fixed
    ... A company has designed an algorithm in MATLAB. ... hardware architecture on which the algorithm is going to be implemented. ...
    (comp.dsp)
  • Re: CPM3 Time and Date stamping, question
    ... Yes, indeed, monahanz, your messages are fascinating: ... care if it was really working, especially if they say they can program ... algorithm, ... If you prefer to code directly in assembly language, ...
    (comp.os.cpm)