Re: Why don't we have a stong AI by now? (To Curt Welch and other)

In article
<7d9e85e5-643c-4b3d-a986-fd889ce27119@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
casey <jgkjcasey@xxxxxxxxxxxx> wrote:

On Aug 11, 3:42 am, Doc O'Leary <droleary.use...@xxxxxxxxxxxxxxxxxx>
wrote:
People mistakenly got the notion that a program that wins
at chess is thus intelligent, when nothing could be further
from the truth.

When something does something with an apparent purpose as
opposed to just nonsense we call that intelligent behaviour
and a chess program fits that description even if we know
it is limited to that domain.

*You* might label that intelligence, but *I* don't. I think most people
would agree that pretty much *all* machines serve a purpose, but
absolutely none of them would be considered intelligent.

Likewise, just because a computer is programmed by a human to play chess
instead of doing word processing, that doesn't make it intelligent. It
*might* be intelligent (even in the limited domain of chess), but that
is something that would have to be determined beyond the mere measure of
a win-lose tally.

The ant¹s behaviours are clearly not
stupid even if limited compared with ours and are thus to
some extent show intelligent behavior.

I don't know that to be true. Again, I'm still waiting to hear what the
definition of intelligence is such that we *can* say a behavior reflects
intelligence. An ant does a lot of things based on genetic programming
and simple reflex behavior. A bee might be a better example, but in any
case you *do* have to show that there is some mental model in place that
can reasonably be said to "think" in a manner that can reasonably be
said to be "intelligent".


It's not just the "AI effect", either, but the whole notion
of the end behavior being the definition that is wrong.
It is all well and fine to have a *test* of intelligence be
based on behavior, but it is a false start (to the point of
being a logical fallacy) to *define* intelligence that way.

But until you open up a brain that IS how we define intelligence.

No. Again, maybe *you* define it that way, but in the same way that
computation was decomposed into the Turing Machine, I think intelligence
can be decomposed into something more specific than "a brain". Indeed,
the whole field of AI is *supposed* be be based on the notion that "a
brain" is not necessary.

If someone consistently succeeds at solving problems wouldn't you
say they were more intelligent than someone who consistently
fails to solve those same tasks?

No. For a real world example, just look at education. It's quite
possible the most intelligent kid in the class gets poor grades because
they are bored and have mentally checked out. It's also quite possible
in our "teach to the test" system that kids will perform well without
really understanding the subject matter.

Isn't that based on their
observable behaviors. Intelligence is a word we use for a class
of behaviors.

Again, maybe *you* use that circular definition, but I don't.
Intelligence is about internal processing (aka, thinking), and to what
extent that is reflected in behavior depends a great deal on the
situation.

I think that is the fundamental misstep that has taken AI down
the wrong path. We make the assumption that *because* we see
what we've accomplished, we can work our way back to the underlying
intelligence. Yes, it is great that we have extensive knowledge
to work with in the modern world, but *society* is the main
benefactor in that, not the individual.

As components of a society we also benefit by it.

You missed my point. You want to make intelligence about standing on
the shoulders of giants, but I think we need to normalize that effect
out. Intelligence is about what each person *can* see, not about the
vantage point they happen to have. We might be able to *do* more thanks
to our advanced western educations, but that doesn't necessarily mean
any of us is inherently smarter than some kid in a primitive tribe in
the jungle.

We're not all that much different brain-wise than we were
50,000 years ago.

But we did develop the ability to cooperate and share knowledge
with future generations using language instead of just dna.
Knowledge in the form of language is the transmission agency
of new information playing the same role dna does for the body.

Information is not intelligence.

At the heart of it all, you still need that "limited" spark of
innovation from intelligence that *can* see just far enough to
take things the next step.

Sure but that is not complex any more than the innovations of
random changes in dna that take things to the next step through
a selective process. When you write "from intelligence" you make
it sound like it is a particular thing when intelligence is
not *something* is it a description of what something *does*.

That is the fundamental misstep I spoke about. I think AI will always
elude us so long as we follow the path of "what it does". Much like DNA
can be constantly changing behind the scenes without really changing a
species' immediate "what it does" until a moment of selection/punctuated
equilibrium is reached, I think intelligence works the same way.

Two humans can be doing the same thing over and over, and you might
naively think they're at the same level of intelligence, but one of them
might be *thinking* something different the whole time and, as a result,
eventually starts doing things in a different/better way. If you *only*
measure intelligence at the point the "what it does" changes, you are
really missing the BIG picture when it comes to understanding
intelligence.

Language and all the refinements in the brain that evolved to
make use of language is the "small quantum" difference between
human and animal intelligence that allows our social system
to evolve carrying us along with it.

Again, sharing information is great for advancing society, but it does
nothing to explain the intelligence that *necessarily* existed in the
first place to create the information from a world of semi-random data.
Language seems like it should more properly be seen as a *conclusion* of
intelligence, not a premise.

--
iPhone apps that matter: http://appstore.subsume.com/
My personal UDP list: 127.0.0.1, localhost, googlegroups.com, astraweb.com,
and probably your server, too.
.

Relevant Pages

  • Chess = Caesar ?
    ... In proposing the text below to Ray Keene, ... In discussing the origin of the word CHESS with Ray Keene, ... that English professors sometimes write as ... language, if you will. ...
    (rec.games.chess.misc)
  • Re: Chess = Caesar ?
    ... 1892 that the origin of the words check and thus chess is ... has the same root as jeu (i.e. game); if you want the theory you can ... cesare-pronounced chessare is italian for caesar ... language, if you will. ...
    (rec.games.chess.misc)
  • Re: Chess = Caesar ?
    ... **You should try upon the monastic link with chess. ... has the same root as jeu (i.e. game); if you want the theory you can ... cesare-pronounced chessare is italian for caesar ... language, if you will. ...
    (rec.games.chess.misc)
  • Re: Gates and Buffet start "Bridge in Schools effort.
    ... >>> tragically fashionable for schools to omit Latin from the curriculum. ... >> I agree-if they can teach bowling, they can teach bridge. ... Chess and Go for Mathematics. ... I'm pretty sure there are foreign language versions. ...
    (rec.games.bridge)