Just Curious - Does anyone know, if...?
Robert
machine"?
I mean any research groups working on it?
I am just curious, if it is being actively pursued at the present.
Gordon D. Pusch
It is the ``Holy Grail'' of artificial intelligence, so in a sense
it is being pursued by essentially _everybody_ working in the field of AI;
however, current limits on computing power make a ``human equivalent''
(or even a ``cockroach equivalent''!) computer program probably unfeasible
for at least another several decades --- even if we understood what
it would take to produce ``sentience'' (which we don't!).
[NOTE: Crackpot Arthur T. Murray will no doubt chime in to claim that he
has _already_ written sentient software. Ignore him --- he's a nutball.]
-- Gordon D. Pusch
perl -e '$_ = "gdpusch\@NO.xnet.SPAM.com\n"; s/NO\.//; s/SPAM\.//; print;'
Arthur T. Murray
>
>"Robert" <star...@netzero.net> writes:
>
>> Does anyone know if any practical work is being done on creating a
>> "sentient machine"?
>>
>> I mean any research groups working on it?
>>
>> I am just curious, if it is being actively pursued at the present.
> It is the ``Holy Grail'' of artificial intelligence,
> so in a sense it is being pursued by essentially _everybody_
IMHO the greatest "competition" to the Mentifex AI comes from
http://www.realai.net/ -- the effort of alii et Ben Goertzel,
who has expressed his opinion of the Mentifex Theory of Mind at
http://sysopmind.com/archive-sl4/0205/0334.html -- an AI forum.
> working in the field of AI; however, current limits
> on computing power make a ``human equivalent'' (or even
> a ``cockroach equivalent''!) computer program probably
> unfeasible for at least another several decades ---
> even if we understood what it would take to produce
> ``sentience'' (which we don't!).
ATM:
Computer science students at the University of Richmond
in Virginia are being directed under "Resources on AI" for
http://www.mathcs.richmond.edu/~acharles/courses/cs333.html --
a class on "Parallel Programming" -- to a Theory of Mind at
http://www.scn.org/~mentifex/theory5.html "Brain-Mind: Know Thyself!"
>
> [NOTE: Crackpot Arthur T. Murray will no doubt chime in
> to claim that he has _already_ written sentient software.
> Ignore him --- he's a nutball.]
http://www.scn.org/~mentifex/jsaimind.html is indeed sentient and
artificially intelligent software, but you are free to ignore it.
http://www.cpan.org/authors/id/M/ME/MENTIFEX/mind.txt -- (Ignore)
http://www.nanomagazine.com/01_10_24 -- interview @ Nanomagazine;
http://www.virtualentity.com/mind/vb/ -- (ignore Mind.VB);
http://www.angelfire.com/nf/vision/ai/mjava.html -- (ignore Mind.JAVA).
Brian Short
I suppose there could be at least a couple philosophical issues:
What exactly does "sentient" entail? intelligence? consciousness?
Would it be possible to create an intelligent machine that lacks
consciousness? Would it be sentient?
In article <gisn2ve...@pusch.xnet.com>, gdp...@NO.xnet.SPAM.com
says...
--
--
People who think they know it all are annoying
to those of us who do.
--
See my web page: http://www.k7on.com/
--
Robert Twardowski
> make a ``human equivalent''
> (or even a ``cockroach equivalent''!) computer
> program probably unfeasible
> for at least another several decades ---
I disagree with the above statement. I believe it is impossible for anyone
to know how much computing power it would take to equal one "human level"
intelligence.
It is possible that maybe 20 or 30 desktop computers (1 ghz or faster)
networked together would be enough processing power to equal a human
sentient intelligence. Key word in this paragraph is "possible". Its also
possible that it may take 100, 1000, 10000 or 100000 computers networked
together to produce enough processing power. Its also possible that we do
not have enough processing power even if we linked every computer together
in the world. All those are "possibilities".
I believe its impossible to tell how much processing power is required to
create the equivalent of a "human in a box" on a computer, until such a
program is written. You need the software before you know how much
rocessing power it requires, and nobody as of yet has created this software.
I personally believe that we currently have at our disposable more then
sufficient processing power to achieve human level intelligence if a
reasonable number of computers are networked together to provide the
processing power. By reasonable number I mean under one hundred. This is
just my personal opinion.
> It is the ``Holy Grail'' of artificial intelligence, so
> in a sense it is being pursued by essentially _everybody_
> working in the field of AI
I don't believe it is being "actively" pursued, which is ashame.
Suppose there was no World War II, then we would probably still today NOT
have nuclear power plants. Its only because nuclear weapons were actively
pursued during World War II that we have the technology today. It may of
taken us a 100 years to develop nuclear energy if not for World War II.
We take for granted that the technology that we have today would of existed
no matter the circumstances. But I don't believe this is a "truth". The
technology that we have today exists because of what was actively pursued or
gained by chance (some inventions were gained thru chance, discovering
something when researching a totally different field.) But the key point
I want to make is that there is also a lot of technology that COULD of
existed today if it was actively pursued. How much technology are we
missing today because it was never actively pursued?
I believe that it could take 100 or 200 years at our current rate of
research into artificial intelligence to begin to create human levels of
intelligence in machines plus to begin to create sentience. This is ashame
because I think if we did actively pursue this field and if the industry had
an incentive to pursue it, we could accomplish this goal in 20 to 30 years
at most.
Robert Twardowski
intelligence, I mean "intelligence with sentience."
Arthur T. Murray
>> "sentient machine"?
> It is the ``Holy Grail'' of artificial intelligence, so in a sense it is
> being pursued by essentially _everybody_ working in the field of AI;
> however, current limits on computing power make a ``human equivalent''
>
> [NOTE: Crackpot Arthur T. Murray will no doubt chime in to claim that he
> has _already_ written sentient software. Ignore him --- he's a nutball.]
Mentifex is definitely NOT a nutcase! Mentifex only responds
in order to point out that since the Mind's Short Term Memory Array was
reimplemented as a doubly-linked list, the sentient nature of the Mind
has become clear for all to see.
For more information see:
http://mind.sourceforge.net/acm.html
A.T. Murray
Incidentally, Hugo de Garis is a guy who seems to be 'going for it' too.
Erik Max Francis
> Mentifex is definitely NOT a nutcase!
So says Mentifex.
--
Erik Max Francis / m...@alcyone.com / http://www.alcyone.com/max/
__ San Jose, CA, US / 37 20 N 121 53 W / ICQ16063900 / &tSftDotIotE
/ \ See the son in your bad day / Smell the flowers in the valley
\__/ Chante Moore
Bosskey.net: Aliens vs. Predator 2 / http://www.bosskey.net/avp2/
A personal guide to Aliens vs. Predator 2.
Gordon D. Pusch
> "Arthur T. Murray" wrote:
>
> > Mentifex is definitely NOT a nutcase!
>
> So says Mentifex.
...And the fact that Murray consistently refers to himself in the third
person using his nom de guerre (very much like Archimedes Plutonium ---
another infamous nutball) rather strongly suggests delusions of grandeur
on his part, and lends little credibility to his protestations of sanity...
DeanIson
>create the equivalent of a "human in a box" on a computer, until such a
>program is written. You need the software before you know how much
>rocessing power it requires, and nobody as of yet has created this software.
Considering that we have not straightened out the basic question of how much of
intelligence mimicry is tied up in hardware and how much in organization, this
seems like a very commonsense statement to me.
spr...@timecube.com
Sorry, this was a send-up. When I read you saying he'd put in 2 cents,
I couldn't resist. But, I don't want opinions of him based on mocking
versions of him. However, the doubly-linked list thing is real, from
a couple of years ago (though maybe not in the Short Term Memory)
and I tried to duplicate my impressions of his tone, sharpened a little
maybe.
Tim Tyler
: Incidentally, Hugo de Garis is a guy who seems to be 'going for it' too.
: see http://glendhu.com/ai/hardware/utahbrain/
Did Hugo ever manage to get the "keys" to his machine from Michael Korkin?
--
__________
|im |yler http://timtyler.org/ t...@tt1.org
Gordon D. Pusch
You mean you _forged_ a posting by ``Mentifex'' ?!? SHAME ON YOU !!!!!!!
I'm sure your ISP considers that to be a TOS violation. Better hope Murray
does not report you to them...
Robert Twardowski
>
> : see http://glendhu.com/ai/hardware/utahbrain/
Doing this just as a "hobby". I was curious when I followed the link below:
Quoting from web site listed above, where they discuss the "brain" for the
artificial cat: "These processors are of a type known as FPGA's. FPGA's
(Field Programmable Gate Arrays) are computer processors which don't have
fixed logic circuits."
Are FPGA's publically available for purchase outside of the academic
community and at what cost? I wouldn't mind experimenting with them, but
have limited resources.
Sincerely,
Robert Twardowski
"Tim Tyler" <t...@tt1.org> wrote in message news:Gz72H...@bath.ac.uk...
Tim Tyler
: Are FPGA's publically available for purchase outside of the academic
: community and at what cost?
They are available inexpensively. However programming them still remains
challenging.
http://www.optimagic.com/faq.html
http://www.io.com/~guccione/HW_list.html
http://www.mrc.uidaho.edu/fpga/fpga.html
http://www.cellmatrix.com/
http://www.xess.com/
Tim Tyler
: I believe it is impossible for anyone to know how much computing power
: it would take to equal one "human level" intelligence.
Well people have had stabs at estimating it. Moravec has a go in
"Robot", for example.
He puts the brain at somewhere around 30 copies of IBM's Big Blue - if
I read his results correctly.
http://www.frc.ri.cmu.edu/~hpm/book97/ch3/
His estimate is based on the density of receptors in the retina.
I think that is likey to make it something of an over-estimate.
Gordon D. Pusch
On the contrary, his estimate is based on taking another estimate that
a cat's retinal cell can be simulated using a mere 100 ops per second,
and naively multiplying that naive figure by an obsolete and far too
small estimate of the number of neurons in the human brain.
Since a cat's retinal cell is a very simple nerve cell, with a very
stereotyped behavior and connectivity, no ability to learn, and roughly
three orders of magnitude fewer synapses than a typical brain cell,
and since a typical brain cell performs a quite complex averaging over its
tens of thousands of synapses with respect to both space and time in order
to determine its average firing-rate, and to adjust its synaptic weights
as a function of both its own recent firing-history and the correlations
of its firings with those of the upstream neurons that synapse to it,
I think it far more likely that Moravec has made a gross UNDER-estimate ---
probably by at =LEAST= four to six orders of magnitude in speed, and likely
by a similar amount in required memory capacity.
Tim Tyler
:> : I believe it is impossible for anyone to know how much computing power
:> : it would take to equal one "human level" intelligence.
:>
:> Well people have had stabs at estimating it. Moravec has a go in
:> "Robot", for example.
:>
:> He puts the brain at somewhere around 30 copies of IBM's Big Blue - if
:> I read his results correctly.
:>
:> http://www.frc.ri.cmu.edu/~hpm/book97/ch3/
:>
:> His estimate is based on the density of receptors in the retina.
:> I think that is likey to make it something of an over-estimate.
: On the contrary, his estimate is based on taking another estimate that
: a cat's retinal cell can be simulated using a mere 100 ops per second,
: and naively multiplying that naive figure by an obsolete and far too
: small estimate of the number of neurons in the human brain.
That does not appear to be correct. Moravec does not base his estimate on
the retina of a cat - nor does he use an estimate of the number of neurons
in a human brain.
: Since a cat's retinal cell is a very simple nerve cell, with a very
: stereotyped behavior and connectivity, no ability to learn, and roughly
: three orders of magnitude fewer synapses than a typical brain cell,
: and since a typical brain cell performs a quite complex averaging over its
: tens of thousands of synapses with respect to both space and time in order
: to determine its average firing-rate, and to adjust its synaptic weights
: as a function of both its own recent firing-history and the correlations
: of its firings with those of the upstream neurons that synapse to it,
: I think it far more likely that Moravec has made a gross UNDER-estimate ---
: probably by at =LEAST= four to six orders of magnitude in speed, and likely
: by a similar amount in required memory capacity.
This criticism appears to be based on your apparent misunderstanding of
Moravec's method.
My reason for suggesting that retinal cells were likely to be doing
much more work than other cells comes from the /intense/ pressure for
minaturisation that occurs in the retina - together with the
demands the retina places on the mitochondrial energy supply -
demands that exceed that of any other group of cells. Those
cells are working *extremely* hard - for nerve cells.
DeanIson
does Drake's equation -- none to speak of and for the same reason: it's all
based on assumptions about which there exists zero information about the
relationship of what you're actually counting to what you're trying to
estimate.
Whenever I try to think about this problem, I keep coming back to a basic
unsettled question: how much of "a human level intelligence" can be attributed
to processing and computation power, and how much can be attributed to the
organization of its supporting memory and processing substructures?
We can treat the neural organization of the human brain as hard-wired
algorithms and roughly estimate the computing power the brain uses to work what
it's got -- but that only establishes an outer limit for the minimum necessary
conditions for this particular way of organizing the subsystems. The human
brain can operate a "human level intelligence" with only a fraction of the
computing power intact. At the very least, the main etranger phenomenon
suggests that a wholly "human level intelligence" can be operated with only 50%
(possibly less) of the circuitry available.
Where does that leave us?
Gordon D. Pusch
> Gordon D. Pusch <gdp...@no.xnet.spam.com> wrote:
>> Tim Tyler <t...@tt1.org> writes:
>>> Robert Twardowski <star...@netzero.net> wrote:
>
>>>> I believe it is impossible for anyone to know how much computing power
>>>> it would take to equal one "human level" intelligence.
>>>
>>> Well people have had stabs at estimating it. Moravec has a go in
>>> "Robot", for example.
>>>
>>> He puts the brain at somewhere around 30 copies of IBM's Big Blue - if
>>> I read his results correctly.
>>>
>>> http://www.frc.ri.cmu.edu/~hpm/book97/ch3/
>>>
>>> His estimate is based on the density of receptors in the retina.
>>> I think that is likey to make it something of an over-estimate.
>>
>> On the contrary, his estimate is based on taking another estimate that
>> a cat's retinal cell can be simulated using a mere 100 ops per second,
>> and naively multiplying that naive figure by an obsolete and far too
>> small estimate of the number of neurons in the human brain.
>
> That does not appear to be correct. Moravec does not base his estimate on
> the retina of a cat - nor does he use an estimate of the number of neurons
> in a human brain.
That was exactly his description of how he obtained his estimate, as given
during his seminar presented at Argonne National Laboratory ca 1995, which
I attended.
Note that there is hard experimental data on the capabilities of cat retinal
neurons, including results obtained by inserting microelectrodes directly
into cat retinal cells; in fact, the majority of what has been learned
about the mammalian visual cortex is based on research on cats. We do not
have comparable data on the human retina, for the quite obvious reason that
performing similar experiments on humans would land one in prison as a
vivisectionist.
>> Since a cat's retinal cell is a very simple nerve cell, with a very
>> stereotyped behavior and connectivity, no ability to learn, and roughly
>> three orders of magnitude fewer synapses than a typical brain cell,
>> and since a typical brain cell performs a quite complex averaging over its
>> tens of thousands of synapses with respect to both space and time in order
>> to determine its average firing-rate, and to adjust its synaptic weights
>> as a function of both its own recent firing-history and the correlations
>> of its firings with those of the upstream neurons that synapse to it,
>> I think it far more likely that Moravec has made a gross UNDER-estimate ---
>> probably by at =LEAST= four to six orders of magnitude in speed, and likely
>> by a similar amount in required memory capacity.
>
> This criticism appears to be based on your apparent misunderstanding of
> Moravec's method.
On the contrary; in my opinion I have a better understanding of it than you do,
based on a number of very pointed questions I and several other researchers
asked Moravec about his methodology, both during and after his talk at Argonne
National Laboratory.
> My reason for suggesting that retinal cells were likely to be doing
> much more work than other cells comes from the /intense/ pressure for
> minaturisation that occurs in the retina - together with the
> demands the retina places on the mitochondrial energy supply -
> demands that exceed that of any other group of cells. Those
> cells are working *extremely* hard - for nerve cells.
That observation is quite irrelevant to comparing the _computational_
capacity of retinal cells vis-a-vis, say, a typical pyramidal neuron.
Retinal neurons are not at all typical of neurons in the brain, and
computationally quite stereotyped and rather boring in their behavior.
Tim Tyler
:>>> Well people have had stabs at estimating it. Moravec has a go in
:>>> "Robot", for example.
:>>>
:>>> He puts the brain at somewhere around 30 copies of IBM's Big Blue - if
:>>> I read his results correctly.
:>>>
:>>> http://www.frc.ri.cmu.edu/~hpm/book97/ch3/
:>>>
:>>> His estimate is based on the density of receptors in the retina.
:>>> I think that is likey to make it something of an over-estimate.
:>>
:>> On the contrary, his estimate is based on taking another estimate that
:>> a cat's retinal cell can be simulated using a mere 100 ops per second,
:>> and naively multiplying that naive figure by an obsolete and far too
:>> small estimate of the number of neurons in the human brain.
:>
:> That does not appear to be correct. Moravec does not base his estimate on
:> the retina of a cat - nor does he use an estimate of the number of neurons
:> in a human brain.
: That was exactly his description of how he obtained his estimate, as given
: during his seminar presented at Argonne National Laboratory ca 1995, which
: I attended.
He describes how he obtained his estimate in the page I
cited. He states he is comparing with a /human/
retina, and he states how he extrapolates this to the
entire brain - it has nothing to do with an estimate of
the number of neurons in the brain.
I think the possibility of his methodology improving in the
years between the lecture you attended and him writing his book
needs to be considered.
: Note that there is hard experimental data on the capabilities of cat retinal
: neurons, including results obtained by inserting microelectrodes directly
: into cat retinal cells; in fact, the majority of what has been learned
: about the mammalian visual cortex is based on research on cats. We do not
: have comparable data on the human retina, for the quite obvious reason that
: performing similar experiments on humans would land one in prison as a
: vivisectionist.
Moravec states how the estimate for processing power is
derived as well. It is related to the volume of
instructions required in simulated retinas to produce
comparable outputs to comparable inputs.
You may be right that research on cat retinas is
involved in deciding what outputs a mammalian retinal
neuron typically produces. However Moravec doesn't
describe that - he derives his estimate from the number
of instructions required to produce edge- and motion-
detector impulses. There are other ways to get data on
human retinal proportions of such cells besides
vivisection - involving, for example tachistoscope
experiments and autopsies.
:>> Since a cat's retinal cell is a very simple nerve cell, with a very
:>> stereotyped behavior and connectivity, no ability to learn, and roughly
:>> three orders of magnitude fewer synapses than a typical brain cell,
:>> and since a typical brain cell performs a quite complex averaging over its
:>> tens of thousands of synapses with respect to both space and time in order
:>> to determine its average firing-rate, and to adjust its synaptic weights
:>> as a function of both its own recent firing-history and the correlations
:>> of its firings with those of the upstream neurons that synapse to it,
:>> I think it far more likely that Moravec has made a gross UNDER-estimate ---
:>> probably by at =LEAST= four to six orders of magnitude in speed, and likely
:>> by a similar amount in required memory capacity.
:>
:> This criticism appears to be based on your apparent misunderstanding of
:> Moravec's method.
: On the contrary; in my opinion I have a better understanding of it than
: you do, based on a number of very pointed questions I and several other
: researchers asked Moravec about his methodology, both during and after
: his talk at Argonne National Laboratory.
He may have changed his methodology as a result of such
comments.
At any rate the way he claims to have derived his
estimate in Robot (as visible via the page I cited)
bears little relation to the methodology you are
describing from 1995.
:> My reason for suggesting that retinal cells were likely to be doing
:> much more work than other cells comes from the /intense/ pressure for
:> minaturisation that occurs in the retina - together with the
:> demands the retina places on the mitochondrial energy supply -
:> demands that exceed that of any other group of cells. Those
:> cells are working *extremely* hard - for nerve cells.
: That observation is quite irrelevant to comparing the _computational_
: capacity of retinal cells vis-a-vis, say, a typical pyramidal neuron.
: Retinal neurons are not at all typical of neurons in the brain, and
: computationally quite stereotyped and rather boring in their behavior.
That's one of the reasons why Moravec's estimate
*doesn't* simply multiply retinal power by the relative
numbers of neurons in the two systems.
I'm sure the methodology isn't perfect - but it doesn't
pretend to be either.
The approach indicates that there are other methods of
determining the computational power of the brain which
do not involve reproducing its program in its entirity.
In particular, you can perform an estimate for a
sub-system (based on the work required under simulation
to produce functionally equivalent apparatus), and then
scale that result.
Lionel B
> "Robert" <star...@netzero.net> writes:
>
>
>>Does anyone know if any practical work is being done on creating a
>>"sentient machine"?
>>
>>I mean any research groups working on it?
>>
>>I am just curious, if it is being actively pursued at the present.
>
>
> It is the ``Holy Grail'' of artificial intelligence, so in a sense
> it is being pursued by essentially _everybody_ working in the field of AI;
> however, current limits on computing power make a ``human equivalent''
> (or even a ``cockroach equivalent''!) computer program probably unfeasible
> for at least another several decades
I think we might well have enough computing power to to "make a
cockroach equivalent". Whatever that might be. (Does it live in the real
world? In simulation in a virtual world? What kind of virtual world?
What does it do? What do real cockroaches think about it?)
> --- even if we understood what
> it would take to produce ``sentience'' (which we don't!).
Of course. Computing power is not the issue.
--
Lionel B
Ludwik Grodzki
just have a look at
http://www-ia.hiof.no/~rolando/adate_intro.html
I recon they have the right idea, after all, ability to learn is what makes
us sentient.
"Arthur T. Murray" <ment...@scn.org> wrote in message
news:yL6X8.13219$A43.1...@newsread2.prod.itd.earthlink.net...