Re: Definition of AI.

In sci.logic, Nam D. Nguyen
<namducnguyen@xxxxxxx>
wrote
on Sun, 28 Jan 2007 05:30:29 GMT
<VxWuh.814512$5R2.166670@pd7urf3no>:
The Ghost In The Machine wrote:
In sci.logic, David
<da5id65536@xxxxxxxxx>
wrote
on 27 Jan 2007 17:28:56 -0800
<1169947736.730480.37180@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>:
"The question of whether a computer can think is no more interesting
than the question of whether a submarine can swim."

- Edsger W. Dijkstra

Dijkstra pretty much sums up my view.

Nice way of putting it. Of course, there's the response
that if it looks like a duck, walks like a duck, and quacks
like a duck, then it's effectively a duck for all intents
and purposes, never mind whether its innards are carbon
or silicon-based and who or what created it.

(At least until one tries to shoot it with a shotgun.)

We also tend to delude ourselves. Ever "talked" to Eliza?
One wonders how much of that is Eliza's thought-patterns
(or our programming facsimile thereof) and how much of
it is our papering over the very real limitations of a
computer which has about the brainpower of a large spider,
and in most cases less mobility.

Hentai porn might also fall under that area, although it's
less AI, more cartoonification of various activities, usually
best left unexpressed in family-geared forums.

There are also the roboreceptionists running around.
They're a bit limited -- one can say "yes", "no", a number,
maybe one's name (which it records), and a few other
things, but for many tasks they're quite adequate, freeing
up the real humans to do other things, like figure out
precisely how the Donner file got filed under New Accounts
when they've been a regular customer for many years... :-)

David

On Jan 24, 10:10 pm, "Nam D. Nguyen" <namducngu...@xxxxxxx> wrote:
I think we could entertain a definition of AI (Artificial Intelligence),
by way of 1st Order formalization. A formal system T is said to be about
AI (or an AI system) iff:

1) T is consistent.
2) There exists a formula F1 which is a theorem in T.
3.a) There exists a formula F2 which is undecidable in T.
3.b) F2 is syntactically a part of F1.

This is so broad a definition that almost any
formal/logical system would be defined as being AI, which
is patently ridiculous.

Sure. As I mentioned, you could claim a rock of certain size
to be a chair. But that should prevent you from constructing
a "better" chair, say made of wood and has 4 legs, and still
conforming to the definition of a chair: that on which we could
sit!

An interesting point. Of course, at some point one's going to have to
identify the components of a chair, and then shape the rock to fit,
if one goes that route.



Classical AI, as *I* would define
it, would be the development of software algorithms and/or
data that would allow a computer to pass the Turing Test,
a good part of the time.

How would you *formally deal with* the "hard" issues of intelligence
such as "consciousness", "introspection", and the like?

The Turing test is the best I can do at the moment, and
it's an operational test, as opposed to something used
to analyze construction. One might call it a black-box
comparative test.

What is consciousness? What is introspection? How would
one tell an entity with true consciousness and introspection
from a very well-coached mimic? (Or, for that matter,
a proxy?)


(A modern formulation would
be to put such a test algorithm in back of an instant
messaging or Inter-Relay Chat server (RFC1459), and then
have people message it, and it respond back. I did see
some years back some advertisements -- somewhere, in the
sleazier backwaters of the Internet -- for "virtual women"
[the exact designation now escapes me]. This presumably
falls under the self-delusionment category I mentioned
earlier... :-) )

Thus far, I'm not sure AI has been all that successful,
though various offshoots thereof show some application in
the commercial realm.



--
#191, ewill3@xxxxxxxxxxxxx
Linux. Because vaporware only goes so far.

--
Posted via a free Usenet account from http://www.teranews.com

.

Relevant Pages

  • Re: Definition of AI.
    ... Dijkstra pretty much sums up my view. ... that if it looks like a duck, walks like a duck, and quacks ... A formal system T is said to be about ... messaging or Inter-Relay Chat server, ...
    (sci.logic)
  • Re: Definition of AI.
    ... Dijkstra pretty much sums up my view. ... that if it looks like a duck, walks like a duck, and quacks ... A formal system T is said to be about ... messaging or Inter-Relay Chat server, ...
    (sci.logic)