Science and Technology news... :: View topic

Nam Nguyen · Guest

Kyle T. Jones wrote:
[quote]Nam Nguyen wrote:
Kyle T. Jones wrote:

The great AI accomplishment would be creating a machine that actually
had esthetic judgment.

Agree. If a machine exhibits *only* non-aesthetic behaviors then we would
call it machine, not an AI machine.

No one has the remotest clue how such a system could be built, even
in principle.

Not agree. How do you know that?

Both those lines come from the post prior to mine, Nam.
[/quote]
My apology. It>s an overlook mistake on my part.

[quote]
Occidental is the man you want!

Cheers.[/quote]

Nam Nguyen · Guest

Chris Menzel wrote:
[quote]On Wed, 30 Jul 2008 02:35:47 GMT, Nam Nguyen <namducnguyen@shaw.ca> said:
...
The great AI accomplishment would be creating a machine that actually
had esthetic judgment.
Agree. If a machine exhibits *only* non-aesthetic behaviors then we would
call it machine, not an AI machine.

Nonsense.
[/quote]
The sewing *machine* my wife has never exhibits any thing I>d call
as an aesthetic behavior. What nonsense about it that I don>t call
this machine an AI machine?

[quote]Building a machine capable of making aesthetic judgments has
*never* been a significant part (or any part?) of AI research.
[/quote]
Maybe that>s why we don>t seem to have huge successes so far in AI.
Maybe they should *first* just go ahead and create some automatons
that could make aesthetics judgments , and worry later whether or
not anything is a "significant part" of any "research"!

[quote]If we
had a machine that couldn>t tell a Chagall from a xkcd cartoon but which
could learn from its experience, draw relevant conclusions from given
information, and respond appropriate to novelty, there is no question it
would be considered a huge AI success.
[/quote]
Sure. "If we had [such] a machine"! Right now I>d say a few people
would be very happy if we could come up with machine that could make
some aesthetics judgments on their own.

Don>t you think if we could make machines that on their own could make
aesthetics judgments, we could also extend them so they would have
a few other Intelligent traits?

Just so you don>t misinterpret my intention here, "aesthetic judgment"
here to me is a generic name that implies some (high) degree of general
intelligences.

[quote]
No one has the remotest clue how such a system could be built, even
in principle.
Not agree. How do you know that?

For one thing, what would it even mean for a machine to "have aesthetic
judgment"?
[/quote]
Easy. It means at appropriate time the machine would output to an
appropriate media a string of ascii symbols like:

" The truth of GC is a beautiful truth"!

[quote]Even the problems facing the success story above, where the
goals are much clearer, are huge -- not least because most every
interesting AI problem is at least NP-hard. How do you fancy getting
around that impediment?
[/quote]
Why would you think the fundamental problems of AI (only) root in hard
issues such as NP-hard, *in the first place*?

>

|-|erc · Guest

"Lars Kecke" <larskecke@gmail.com> wrote ...
[quote]|-|erc schrieb:
"Lars Kecke" <larskecke@gmail.com> wrote ...
|-|erc schrieb:
In 5 to 20 years search engines will have a new feature,
contextual search.
you mean, like the Science Citation Index? Has been around since the
invention of <meta> tags.

I once found a book that contained how to write a web browser in visual basic.
Try searching for it, "web", "browser", and "visual basic" all produce a tonne
of fluff, its impossible to search for what you actually want.

Searching for "browser in visual basic" will lead to
http://www.acky.net/tutorials/vb/wbrowser/ with four clicks. What was
your problem?

1 year after contextual search is available, we will crack
artificial intelligence.
no way. Since we still don>t know what this "intelligence" thing is, we
won>t recognize AI until it kicks us.

if it understands sentences and makes up answers, its doing what
the language part of our brain is doing, that>s a big chunk of brain!
spatial and temporal reasoning should be easy once the computers
can talk.

but this point was "just around the corner" since the times of ELIZA
(1966), and there hasn>t been much sizeable progress so far. So your
statement is equivalent to "interstellar travel shoult be easy once we
get FTL drives".
[/quote]
Search engine technology is only 10 years old, and computers are
10,000 times faster and 10,000 times higher capacity than 30 years ago.

It>s hard to find an off the shelf transister with a frequency higher than 500MHZ,
we>ve hit the limit of computing speed only recently. Computers are much
more available now too, nearly everyone has one.

Everything is in place, the computers are just capable enough to handle thought
like computations, the internet is a formalised playground to learn on, AI can>t
be that hard to crack that it takes millions of researchers 100 years.

[quote]
By 2060 there will be natural feeling
bionic limbs, and bionic eyes for the blind.
Probably no need for it. Just use headless clones as organ donors.

how does a headless clone secrete the right hormones to grow properly?

you can just fiddle with those Hox genes to prevent the formation of
higher brain areas. The brainstem will remain intact.

how do you stop your brain from aging? you>d be in a fit body but you>d
be senile.

you wanted replacement limbs, I gave you replacement limbs. Ageing and
senility are good things, they prevent stratification of our society.
With a cure for senility, you would have to kill the old guy in power in
order to replace him.

Lars
[/quote]
Wouldn>t you rather live 1,000 years? Or more?

Herc

zzbunker@netscape.net · Guest

On Jul 31, 1:17 am, "|-|erc" <h...@r.c> wrote:
[quote]"Lars Kecke" <larske...@gmail.com> wrote ...

|-|erc schrieb:
"Lars Kecke" <larske...@gmail.com> wrote ...
|-|erc schrieb:
In 5 to 20 years search engines will have a new feature,
contextual search.
you mean, like the Science Citation Index? Has been around since the
invention of <meta> tags.

I once found a book that contained how to write a web browser in visual basic.
Try searching for it, "web", "browser", and "visual basic" all produce a tonne
of fluff, its impossible to search for what you actually want.

Searching for "browser in visual basic" will lead to
http://www.acky.net/tutorials/vb/wbrowser/with four clicks. What was
your problem?

1 year after contextual search is available, we will crack
artificial intelligence.
no way. Since we still don>t know what this "intelligence" thing is, we
won>t recognize AI until it kicks us.

if it understands sentences and makes up answers, its doing what
the language part of our brain is doing, that>s a big chunk of brain!
spatial and temporal reasoning should be easy once the computers
can talk.

but this point was "just around the corner" since the times of ELIZA
(1966), and there hasn>t been much sizeable progress so far. So your
statement is equivalent to "interstellar travel shoult be easy once we
get FTL drives".

Search engine technology is only 10 years old, and computers are
10,000 times faster and 10,000 times higher capacity than 30 years ago.

It>s hard to find an off the shelf transister with a frequency higher than 500MHZ,
we>ve hit the limit of computing speed only recently. Computers are much
more available now too, nearly everyone has one.

Everything is in place, the computers are just capable enough to handle thought
like computations, the internet is a formalised playground to learn on, AI can>t
be that hard to crack that it takes millions of researchers 100 years.
[/quote]
It will much larger than that. Since the math stooges
have also been told CONSISTENLY for the last 30 years
that the only thing the syncopated morons know less
about than transistors is compilers.

[quote]

By 2060 there will be natural feeling
bionic limbs, and bionic eyes for the blind.
Probably no need for it. Just use headless clones as organ donors.

how does a headless clone secrete the right hormones to grow properly?

you can just fiddle with those Hox genes to prevent the formation of
higher brain areas. The brainstem will remain intact.

how do you stop your brain from aging? you>d be in a fit body but you>d
be senile.

you wanted replacement limbs, I gave you replacement limbs. Ageing and
senility are good things, they prevent stratification of our society.
With a cure for senility, you would have to kill the old guy in power in
order to replace him.

Lars

Wouldn>t you rather live 1,000 years? Or more?

Herc- Hide quoted text -

- Show quoted text -- Hide quoted text -

- Show quoted text -[/quote]

zzbunker@netscape.net · Guest

On Jul 31, 1:17 am, "|-|erc" <h...@r.c> wrote:
[quote]"Lars Kecke" <larske...@gmail.com> wrote ...

|-|erc schrieb:
"Lars Kecke" <larske...@gmail.com> wrote ...
|-|erc schrieb:
In 5 to 20 years search engines will have a new feature,
contextual search.
you mean, like the Science Citation Index? Has been around since the
invention of <meta> tags.

I once found a book that contained how to write a web browser in visual basic.
Try searching for it, "web", "browser", and "visual basic" all produce a tonne
of fluff, its impossible to search for what you actually want.

Searching for "browser in visual basic" will lead to
http://www.acky.net/tutorials/vb/wbrowser/with four clicks. What was
your problem?

1 year after contextual search is available, we will crack
artificial intelligence.
no way. Since we still don>t know what this "intelligence" thing is, we
won>t recognize AI until it kicks us.

if it understands sentences and makes up answers, its doing what
the language part of our brain is doing, that>s a big chunk of brain!
spatial and temporal reasoning should be easy once the computers
can talk.

but this point was "just around the corner" since the times of ELIZA
(1966), and there hasn>t been much sizeable progress so far. So your
statement is equivalent to "interstellar travel shoult be easy once we
get FTL drives".

Search engine technology is only 10 years old, and computers are
10,000 times faster and 10,000 times higher capacity than 30 years ago.

It>s hard to find an off the shelf transister with a frequency higher than 500MHZ,
we>ve hit the limit of computing speed only recently. Computers are much
more available now too, nearly everyone has one.

Everything is in place, the computers are just capable enough to handle thought
like computations, the internet is a formalised playground to learn on, AI can>t
be that hard to crack that it takes millions of researchers 100 years.

By 2060 there will be natural feeling
bionic limbs, and bionic eyes for the blind.
Probably no need for it. Just use headless clones as organ donors.

how does a headless clone secrete the right hormones to grow properly?

you can just fiddle with those Hox genes to prevent the formation of
higher brain areas. The brainstem will remain intact.

how do you stop your brain from aging? you>d be in a fit body but you>d
be senile.

you wanted replacement limbs, I gave you replacement limbs. Ageing and
senility are good things, they prevent stratification of our society.
With a cure for senility, you would have to kill the old guy in power in
order to replace him.

Lars

Wouldn>t you rather live 1,000 years? Or more?
[/quote]
Or better stated as "Math was invneted for bean counters".
Physics was invented for Kant Wannabees".
Poetry was invented for wanks.
Engineering was invented for people of the 21st Century.

[quote]
Herc- Hide quoted text -

- Show quoted text -- Hide quoted text -

- Show quoted text -[/quote]

Lars Kecke · Guest

|-|erc schrieb:
[quote]"Lars Kecke" <larskecke@gmail.com> wrote ...

you wanted replacement limbs, I gave you replacement limbs. Ageing and
senility are good things, they prevent stratification of our society.
With a cure for senility, you would have to kill the old guy in power in
order to replace him.

Wouldn>t you rather live 1,000 years? Or more?
[/quote]
Actually, no. Only if I get a memory wipe every 100 years or so, but
then I could just as well die. Besides, what I want is irrelevant; a
society that keeps its elite in extended animation is incapable of
adapting to changes of environment, so even if my lifespan (and that of
the old guys in power) were 1ky+, our culture would be overrun by
quicker-evolving barbarians within a century. To use an old analogy,
cancer cells potentially live for ever (usually their mitosis counter is
disabled), but it does them no good.

Lars

Lars Kecke · Guest

|-|erc schrieb:

[quote]Yes, all we have to do is decipher the meaning of sentences algorithmically and AI
will be solved after that.
[/quote]
Unfortunately it seems to be the other way round; you need a lot of
general knowledge to decypher the meaning of everyday-language
sentences, so dinding meaning in sentences is equivalent to AI (that>s
what the Turing test is about).

[quote]It would be a strange universe if our brains could do it, a piece
of meat with predominently 1 type of cell - the neuron,
[/quote]
Glia cells tend to disagree with this statement.

[quote]but computers couldn>t do it.
[/quote]
but why would they want it? When AI comes it wil come creaping and
percolating into our lives (already does), so you won>t be able to draw
the line between "before AI" and "after AI". In a way, it>s already
there; today>s machines are much smarter than what was thought possible
in the 1950s, but to save our egos, all those feats of appearant
intelligence (like not losing a single game against the world chess
champion) are demoted to parlour tricks.

[quote]Hopefully we>ll solve AI in our lifetimes because it will revolutionise society.
[/quote]
also the end of oil or cheap clean water will revolutionize society;
change for change>s sake is usually not something to hope for.

Lars

Chris Menzel · Guest

On Wed, 30 Jul 2008 22:25:32 -0700 (PDT), galathaea
<galathaea@gmail.com> said:
[quote]Chris Menzel wrote:
On Wed, 30 Jul 2008 02:35:47 GMT, Nam Nguyen <namducnguyen@shaw.ca
said:
...
The great AI accomplishment would be creating a machine that
actually had esthetic judgment.

Agree. If a machine exhibits *only* non-aesthetic behaviors then we
would call it machine, not an AI machine.

Nonsense. Building a machine capable of making aesthetic judgments
has *never* been a significant part (or any part?) of AI research.
If we had a machine that couldn>t tell a Chagall from a xkcd cartoon
but which could learn from its experience, draw relevant conclusions
from given information, and respond appropriate to novelty, there is
no question it would be considered a huge AI success.

we already have those
[/quote]
Only to a limited extent. Our best AI systems are programmed to operate
in very constrained environments that involve little novelty beyond what
has been anticipated for that environment. The vehicles that completed
the DARPA Grand Challenge were terrific at what they did, but that>s all
they could do.

[quote]it _was_ a huge AI success

simple deductive systems that learn
have been foundational to all computer science
[/quote]
"simple" being the operative word here. There are some terrific theorem
provers out there. But a dedicated theorem prover is a far cry from a
system that can draw its own *relevant* conclusions for the purpose of
guiding action (as opposed to generating a theorem for a (typically)
antecedently given conclusion) and learn from its experience.

[quote]("learn" is usually equivalent to "respond appropriately to novelty")
[/quote]
Usually not, actually. It is quite conceivable, e.g., to have a machine
that is extremely good at dealing with novelty but which doesn>t do at
all well at *remembering* how it deals with novelty and hence has to run
thru the same reasoning process all over again when confronted at a
later time with a situation it encountered at an earlier time. Granted,
it would be *better* to build a machine that remembers and can build on
what it has done in the past, i.e., learn from it, but the point is that
the two skills are not the same.

[quote]even though the everyman home computer is an example of such a system
[/quote]
Well, if you think that, then your notion of AI is quite different from
anyone who actually works in the field. This is of course not to deny
that there are both hardware and software components of your Mac or
PeeCee that incorporate the results of AI research. But your
desktop/laptop is not an AI system, let alone an instance of the
"success story" above.

Chris Menzel · Guest

On Thu, 31 Jul 2008 04:26:32 GMT, Nam Nguyen <namducnguyen@shaw.ca> said:
[quote]Chris Menzel wrote:
On Wed, 30 Jul 2008 02:35:47 GMT, Nam Nguyen <namducnguyen@shaw.ca> said:
...
The great AI accomplishment would be creating a machine that
actually had esthetic judgment.
Agree. If a machine exhibits *only* non-aesthetic behaviors then we
would call it machine, not an AI machine.

Nonsense.

The sewing *machine* my wife has never exhibits any thing I>d call
as an aesthetic behavior. What nonsense about it that I don>t call
this machine an AI machine?
[/quote]
Good grief. I said that exhibiting aesthetic behavior is not necessary
for being an AI system. I did not say that *failing* to exhibit
aesthetic behavior is *sufficient* for being an AI system.

[quote]Building a machine capable of making aesthetic judgments has *never*
been a significant part (or any part?) of AI research.

Maybe that>s why we don>t seem to have huge successes so far in AI.
Maybe they should *first* just go ahead and create some automatons
that could make aesthetics judgments , and worry later whether or not
anything is a "significant part" of any "research"!
[/quote]
So your claim is that if AI researchers were to start focusing on
creating machines capable of making aesthetic judgments, then all the
other (well understood) goals of AI would suddenly become more
achievable? What possible reasons are there for thinking *that*? How
will building in a capacity for making aesthetic judgments help with the
fact that, e.g., planning problems are NP-hard? The theoretical hurdles
of traditional AI will not disappear simply in virtue of a concerted
push to build aesthetically sensitive machines.

[quote]If we had a machine that couldn>t tell a Chagall from a xkcd cartoon
but which could learn from its experience, draw relevant conclusions
from given information, and respond appropriate to novelty, there is
no question it would be considered a huge AI success.

Sure. "If we had [such] a machine"! Right now I>d say a few people
would be very happy if we could come up with machine that could make
some aesthetics judgments on their own.
[/quote]
Maybe so. So what? Once again, this has no obvious implications for
the general challenges of AI.

[quote]Don>t you think if we could make machines that on their own could make
aesthetics judgments, we could also extend them so they would have a
few other Intelligent traits?
[/quote]
I have no idea whether that would follow. It is certainly not obvious
that it would. For another thing, I>m not even sure what the goal is.
I can create a machine that can make aesthetic judgments right now --
program it to print "That sucked!" no matter what you exposed it to. A
silly example, for sure, but what are the success criteria here? What
makes an aesthetic judgment Intelligent with a big "I", or even a little
one? (I>m not really interested in hearing your answer, since I find
the idea largely pointless.)

[quote]No one has the remotest clue how such a system could be built, even
in principle.
Not agree. How do you know that?

For one thing, what would it even mean for a machine to "have aesthetic
judgment"?

Easy. It means at appropriate time the machine would output to an
appropriate media a string of ascii symbols like:

"The truth of GC is a beautiful truth"!
[/quote]
Ok, alter my critical machine above to print a whole array of such
strings at appropriate times. But, you ask, what>s appropriate? My
question exactly.

[quote]Even the problems facing the success story above, where the goals are
much clearer, are huge -- not least because most every interesting AI
problem is at least NP-hard. How do you fancy getting around that
impediment?

Why would you think the fundamental problems of AI (only) root in hard
issues such as NP-hard, *in the first place*?
[/quote]
Uh, well, because all (or very nearly all) AI problems *are* at least
NP-hard -- at any rate, those, like planning and scheduling (not to
mention simple theorem-proving), that involve explicit reasoning. You
don>t know this?

galathaea · Guest

On Jul 31, 1:31 pm, Chris Menzel <cmen...@remove-this.tamu.edu> wrote:
[quote]On Wed, 30 Jul 2008 22:25:32 -0700 (PDT), galathaea
galath...@gmail.com> said:
Chris Menzel wrote:
On Wed, 30 Jul 2008 02:35:47 GMT, Nam Nguyen <namducngu...@shaw.ca
said:

The great AI accomplishment would be creating a machine that
actually had esthetic judgment.

Agree. If a machine exhibits *only* non-aesthetic behaviors then we
would call it machine, not an AI machine.

Nonsense. Building a machine capable of making aesthetic judgments
has *never* been a significant part (or any part?) of AI research.
If we had a machine that couldn>t tell a Chagall from a xkcd cartoon
but which could learn from its experience, draw relevant conclusions
from given information, and respond appropriate to novelty, there is
no question it would be considered a huge AI success.

we already have those

Only to a limited extent. Our best AI systems are programmed to operate
in very constrained environments that involve little novelty beyond what
has been anticipated for that environment. The vehicles that completed
the DARPA Grand Challenge were terrific at what they did, but that>s all
they could do.
[/quote]
you missed my point

we already have what you described

we already have machines which can
- learn from their experiences
- draw relevant conclusions from their given information
- respond appropriately to novelty

of course you can always add x, y, and z
and eventually come up with some intersection of properties
we have not yet done

my point was that your criteria had nothing to do
with what is actually still needed in AI theory

[quote]it _was_ a huge AI success

simple deductive systems that learn
have been foundational to all computer science

"simple" being the operative word here. There are some terrific theorem
provers out there. But a dedicated theorem prover is a far cry from a
system that can draw its own *relevant* conclusions for the purpose of
guiding action (as opposed to generating a theorem for a (typically)
antecedently given conclusion) and learn from its experience.
[/quote]
relevance has never been a problem for any machine

in fact it is quite the opposite

it is difficult to get meaningful responses from algorithms
that are unrelated to its input

most simply massage the input in known ways to produce
automatically relevant conclusions

[quote]("learn" is usually equivalent to "respond appropriately to novelty")

Usually not, actually. It is quite conceivable, e.g., to have a machine
that is extremely good at dealing with novelty but which doesn>t do at
all well at *remembering* how it deals with novelty and hence has to run
thru the same reasoning process all over again when confronted at a
later time with a situation it encountered at an earlier time. Granted,
it would be *better* to build a machine that remembers and can build on
what it has done in the past, i.e., learn from it, but the point is that
the two skills are not the same.
[/quote]
actually
it is

the whole reason you don>t normally have this "conceivable" machine
response
is precisely because it is not normally seen as the _appropriate_
response
to the given novel input

the problem is in your notion of appropriate

this is not an a priori given term
and is determined by the researchers expectations for a project

usually in AI they simply want their machines to learn from novel
input
which is what i said

sometimes they also want certain novel outputs instead of expected
outputs
and sometimes they don>t need learning if deducing is enough
(particularly in stateless deducers)
but these are less commonly expected

and these hidden assumptions and intentions are a deeper problem with
your post
(and many of the others in this thread)
in that they obscure where the real effort of AI is

there seems to be a lot of handwringing
"oh if we only had <xyz> we>d finally solve AI"
where xyz is a vaguely worded description of hidden desires
that when viewed formally states nothing more than what already
exists

because the science of pattern deduction has actually solved many
problems
and the known still-existing problems
are much more technical and of very limited scopes
or are problems of scale

as you said
"'simple' being the operative word here"
but even here
the capabilities of modern systems are extremely complex
understanding and able to reason about huge ontologies of objects
and the transformations they undergo

[quote]even though the everyman home computer is an example of such a system

Well, if you think that, then your notion of AI is quite different from
anyone who actually works in the field. This is of course not to deny
that there are both hardware and software components of your Mac or
PeeCee that incorporate the results of AI research. But your
desktop/laptop is not an AI system, let alone an instance of the
"success story" above.
[/quote]
actually my point was that your notion of AI
was quite different than of many who work in the field

the PC most certainly has all the capabilities you mentioned

which is why i mentioned the point about drives
and other things that PCs don>t currently have
which you snipped

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
galathaea: prankster, fablist, magician, liar

galathaea · Guest

On Jul 31, 4:04 pm, Chris Menzel <cmen...@remove-this.tamu.edu> wrote:

[quote]Why would you think the fundamental problems of AI (only) root in hard
issues such as NP-hard, *in the first place*?

Uh, well, because all (or very nearly all) AI problems *are* at least
NP-hard -- at any rate, those, like planning and scheduling (not to
mention simple theorem-proving), that involve explicit reasoning. You
don>t know this?
[/quote]
i really do not understand this at all

an algorithm to solve a problem will not be considered intelligent
unless it is polynomial time?

or is it just time to solve?

will we consider it intelligent once it gets as fast as humans?
so all we have to do is wait on moore?

this focus on np-hard
which i>ve seen several times now
looks like someone found out about boolean satisfiability
and just metaphorically thought of AI

but it stuck

what i notice about most of these np-hard AI problems
is that they almost all also have well-known solutions...

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
galathaea: prankster, fablist, magician, liar

Nam Nguyen · Guest

Chris Menzel wrote:

[quote](I>m not really interested in hearing your answer, since I find
the idea largely pointless.)
[/quote]
You really have an attitude and a mind of an Inquisitor!

(Iirc, somewhere along time ago, I think you were advising someone
something to the effect that debate/discussion is a give and take
and a reflection of the exchanges of thoughts or ideas on the
problems. Along time ago indeed!)


www.GetXFactor.com Leading Technology, Science, Agriculture News and information	Part of the Identityscape.com network... getxfactor.com jmoodmusic.com smartbusinesschoices.com mintdepot.com lowfaresalways.com evangelicalview.com shoppingpodder.com soproudlywehail.com webnews.ws currenthumor.com

Science and Technology news... Forum Index -> Logic Forum	Goto page Previous 1, 2, 3, 4, 5, 6
Page 6 of 6