Re: Baby AI's First Words?

On Tue, 25 Apr 2006 12:50:11 GMT, alan jones <ob2@xxxxxxxxxxxxxxxxxx>
in comp.ai.philosophy wrote:

Lester Zick wrote:
On Mon, 24 Apr 2006 22:25:44 GMT, alan jones <ob2@xxxxxxxxxxxxxxxxxx>
in comp.ai.philosophy wrote:


Lester Zick wrote:


On Mon, 24 Apr 2006 18:50:22 GMT, alan jones <ob2@xxxxxxxxxxxxxxxxxx>
in comp.ai.philosophy wrote:



Lester Zick wrote:


On Sun, 23 Apr 2006 16:52:07 GMT, alan jones <ob2@xxxxxxxxxxxxxxxxxx>
in comp.ai.philosophy wrote:




Lester Zick wrote:


[. . .]




How about this definition for lie: a contradiction to truth,


Which in turn assumes we understand what truth means to begin with.

Well we have an empirical idea of truth, which seems to stand up to
the best that logic can throw at it, so why not use that idea of
truth.




Because it isn't actually true? I've written so extensively on
empirical assumptions of truth I don't see much reason to revisit
the issue.

Ok I accept even without you stating, that as an idea, truth
isn't bound to anything other than our current perception of
truth. Even so why not build your machine with the ability to
reappraise its current standing? Build it to accept and adjust
to this limitation.


The difficulty here is that without an unambiguous universal
definition for truth you're really left with nothing to program
except various personal opinions as to what constitutes truth in
given situations. There are legal standards of truth and there are
emprical standards of truth but there are also a host of problematic
standards people use daily in a variety of contexts.

The problem should become apparent if anyone sat down to design
standards for truth in software terms. Not only will the problem
itself become intractable in size but what if criteria for truth
in particular contexts yielded incorrect information? The machine
would just have to go on calling it truth regardless. Of course
this is just what empirics themselves do. This is why we need
science in the first place. Otherwise we're left just mechanizing
personal opinions and prejudices.

Ah, but truth isn't fixed. Instead, truth might be regared as the
most complete idea of the moment. Any idea of truth which likens
it to a stone tablet, leaves this possibility of becoming unstuck.
Instead you allow for this limitation. Not even science is without
its limits. The challenge is to work witin those limits allowing
for change. This machine to be truly intelligent, would need to
adapt.


Well as I've pointed out repeatedly isolated truths may not be
immutable but the general idea of truth certainly is in universal
terms as that whose alternatives are necessarily false.



Or we could just start with simple mistakes. How is it that we
make mistakes? Then we could move on to lies.

Geez where to start, conscious or unconscious error? Or errors
based on false presumptions., where our mental constructions
leads us predict outcome, only to jump to the false conclusions.
Or error based on ....


Yeah I understand. The point of my suggestion is to eliminate the
mechanization of motive and collateral factors just to concentrate on
what mistakes mean and how they originate in basic mechanical terms.




Okay. So let's just back off and consider mistakes. Do rocks make
mistakes? Do computers make mistakes?

Surely the question ought to start with humans, or at least
those creatures with behavior which allows the possibility
of mistakes? Does it take intelligence to perceive mistakes?
We could say mistakes are bound to an idea of correction.
In which case we might find this feed back loop in other
areas besides animal behavior?


No I strenuously disagree. The whole thrust of the mechanics
I've been discussing recently is that any mechanical view of
anything requires definition of what the mechanics is not.
Of course we have to consider humans because we do know that
we make mistakes. But we nontheless have to extend whatever
mechanical definition of mistakes we come up with to show why
rocks etc. don't make mistakes in the same terms. In other
words we really have to come up with some kind of dualistic
or pluralistic mechanics of mistakes which shows how we on
the one hand make mistakes but rocks and other material
artifacts don't.

Ok, i'll conceed the usefulness of defining an object by its
'negative space'. Too many times we divise rules which are based
on apparent corallation, rules which fail on that basis. My
current bug-bare happens to be with the 'science' of behavioral
profiles, which could stand a similar analysis. It defines,
what is, before it looks to define what is not.


Of course that's exactly the whole problem with identity theories of
reality when true theories of reality can only get at what is true on
the basis of alternatives to what is false.



Well the actual mechanics of lies is undoubtedly premature.
Let's just try to figure how it is we make mistakes. Rocks
don't make mistakes. Computers don't make mistakes. We make
mistakes. So maybe we better just concentrate on the mechanics
of mistakes and go on from there.

Ok. In a round about fashion you could link mistakes to energy.
or a need to be efficient. Also linking mistakes to energy,
is the obvious idea that as grow tired we take short cuts.
Routine seems to lead to mistakes, suggesting that there is
some mechanism which takes over as we establish routines and
this leads to false predictions, or autonomous behavior which
then fails to notice. The alternative to making mistakes would
be to behave with the presumption that we were about to make
a mistake, in which case so much energy would be consumed
considering that possibility of mistake. Perhaps there is a
pay off between acceptance of the possibility and the mechanism
of correction. I'm thinking now of locomotion and the mechanisms
which seem better able to correct for mistakes.

Are any of these scenarios likely to affect a machine?


I think an energy budget explanation for mistakes would have to
apply equally to material artifacts such as rocks and computers.
At least any alternative would indicate that material artifacts
don't have energy budgets, which I think we can dismiss out of
hand. The difficulty is you seem to be considering just the
physics involved and not the mechanics of actually reaching for
some idea and getting a wrong answer.

I think any idea for the mechanics of mistakes would have to ask how
it is possible to reach for a match and come up with the wrong match.
Of course rocks don't reach for matches but computers do; however,
apart from the extremely rare occurence where error circuitry fails
computers can only come up with a correct match if there is a match.

... hmm, where as human beings might reach with some preprogrammed
instinct, to find what might not actually there. In other words,
human beings act as much by routine, as calculation. Routines which
might lead to an anticipation of a match, and lead to our mistakes.
This begs the question, what would create this anticipation, other
than a need to conserve energy?-)



But let's extend this to a situation where we work with differences
between differences. If I'm looking for say the antecedents for the
difference between 17 and2 I would be looking for 15. And I might
actually find that match just like a computer would and just as
reliably. But the problem would then be that I'd have no way to know
that what I found were the actual antecedents I'm looking for since
I'd only have their difference to work with.

So you are saying that this antecedents might be one of many,
sharing the same answer. And the answer would not rule out the
possibility of other 'differences' within that problem state.
Its seems to me this problem is bound by its definitions. What
one finds, is determined by how well the problem is defined.


But even if the problem is defined perfectly (which I assume it is for
mechanical purposes) we still face identical considerations. How we
define things doesn't really matter either unless we have some way to
know definitions themselves cannot be in error.



Still i like this idea of the relative difference. Differences
for which there are ambigious answers. I suppose a mathematical
language would go on to give the possible answers, an attribute
of some kind. 1st 2nd ... Nth.


Sure. The primary obstacle to mathematical definitions and language
turns out to be mathematikers whose only interest seems preservation
of the status quo in Cartesian geometry, Cantorian arithmetic, and TvN
mechanics.



This why why we make mistakes and material artifacts such as rocks do
not.Rocks are in the 1D category and only work in terms of differences
alone. Sentient beings on the other hand lie in the 2-4D category of
mechanical pluralisms which work in terms of differences between
differences. Hence they are invariably at a loss to know exactly when
they have found the correct antecedents even though their matching
algorithms are mechanically correct.

~v~~


You could say rocks don't make mistakes because mistakes are
an attribute of something, which isn't found in rocks. Call
it a property of dynamic systems. Dynamics and complexity,
allied to organization, then perhaps consciousness. The rock
being without all these properties is unable to make mistakes.


Yes. But the basic physical fact is that rocks as material artifacts
are only defined in terms of differences alone and not differences
between differences. We can call that characteristic all kinds of
things. But doing so we risk trying to analyze those characteristics
in non essential terms and obscuring underlying mechanical issues.



The rock is, without being, an actuator. [deliberate ambiguity]

You could ask instead, how do we make the correct choices?


No actually we can't because we don't have any way to define correct
choices in the absence of any way to define incorrect choices.



This would limit the question to those objects capable of
making choices. Choise: deciding between one course and another.
This might take us to 'why' we make choice, which in turn
might takes us back to energy. Energy attracting the behaviors
which we then say are choices.

[you migh like to add 'time' for all occurances of 'energy'.]


Allan, you're still stuck on the physics instead of the logic and the
mechanical considerations of the logic. It isn't that we make choices.
We do of course but the "choicing" mechanism are differences between
differences which rocks lack as material artifacts. "Why" in terms of
motivation has nothing to do with it either except as it stems from
definition of sentience in terms of differences between differences.
Given that circumstance we have no choice but to make errors and
occasionally make non errors as a byproduct in the course of making
errors.


Ha ha ha ;-)


Which of course clearly conforms to the normal course of events.

~v~~


Sorry, but you have to admit your conclusion was funny. We
make errors as the norm, with 'non errors' as a byproduct.

Of course it's hilarious. It was intended to be. I'm often accused of
being witty. But the humor is in the mechanics. Just think about it.
Why do you suppose mathematics journals are filled with line by
line minutiae. It isn't just because the subject matter is so arcane.
It's because we have to make exactly sure it conforms to the axioms
setting out the principles involved. Why do you suppose we have no
out-of-hand solutions to intelligence and AI? Because all we're doing
is guessing at this point and most of our guesses turn out wrong.

Tell me how do you differentiate between the physics and the
logic? Isn't at some level logic informed by the physics?

Well the distinction I draw is analogous to mathematics and
mathematical objects. We can and do analyze mathematical
principles and theorems without regard to objects. We derive
relationships like 2+2=4 without trying to comprehend quantum
relationships underlying apples or oranges to which they apply.
Identical considerations apply to logic. Someday we may get to
relationaships within the brain. But the same principles of logic
will apply regardless.

~v~~

.

Relevant Pages

  • Re: Baby AIs First Words?
    ... Which in turn assumes we understand what truth means to begin with. ... what mistakes mean and how they originate in basic mechanical terms. ... anything requires definition of what the mechanics is not. ... rocks etc. don't make mistakes in the same terms. ...
    (comp.ai.philosophy)
  • Re: Baby AIs First Words?
    ... Which in turn assumes we understand what truth means to begin with. ... what mistakes mean and how they originate in basic mechanical terms. ... anything requires definition of what the mechanics is not. ... rocks etc. don't make mistakes in the same terms. ...
    (comp.ai.philosophy)
  • Re: Baby AIs First Words?
    ... Which in turn assumes we understand what truth means to begin with. ... what mistakes mean and how they originate in basic mechanical terms. ... The whole thrust of the mechanics I've been discussing recently is that any mechanical view of anything requires definition of what the mechanics is not. ... But we nontheless have to extend whatever mechanical definition of mistakes we come up with to show why rocks etc. don't make mistakes in the same terms. ...
    (comp.ai.philosophy)
  • Re: Baby AIs First Words?
    ... Which in turn assumes we understand what truth means to begin with. ... what mistakes mean and how they originate in basic mechanical terms. ... The whole thrust of the mechanics I've been discussing recently is that any mechanical view of anything requires definition of what the mechanics is not. ... But we nontheless have to extend whatever mechanical definition of mistakes we come up with to show why rocks etc. don't make mistakes in the same terms. ...
    (comp.ai.philosophy)
  • Re: Baby AIs First Words?
    ... Which in turn assumes we understand what truth means to begin with. ... what mistakes mean and how they originate in basic mechanical terms. ... The whole thrust of the mechanics I've been discussing recently is that any mechanical view of anything requires definition of what the mechanics is not. ... But we nontheless have to extend whatever mechanical definition of mistakes we come up with to show why rocks etc. don't make mistakes in the same terms. ...
    (comp.ai.philosophy)