Re: Blue Brain: Illuminating the Mind
J.Random.User
It looks like we may arrive at true artificial intelligence a lot sooner than we thought we would...
Quote from article below:
"Markram's EPFL team, collaborating with IBM researchers
and an online network of brain and computer scientists,
will use Blue Brain to create a detailed computer model
of the neocortex, the largest and most complex part of
the human brain. "That's going to take two to three years,"
he says. Then, with a bigger Blue Brain, he hopes to build
a cellular-level model of the entire brain. This may take
a decade..."
Full Article:
;
; BUSINESS WEEK NEWS ANALYSIS
; By Otis Port
; June, 2005
; http://www.businessweek.com/technology/
; content/jun2005/tc2005066_6414_tc024.htm
;
; Blue Brain: Illuminating the Mind
;
; Scientists will use the blazingly fast supercomputer to do
; never-before-possible research into how we think and how mental
; disorders arise
;
; On July 1, the Blue Brain computer will wake up, marking "a
; monumental moment" in the history of brain research, says
; neuroscientist Henry Markram, founder of the Brain Mind
; Institute at Switzerland's Ecole Polytechnique Federale de
; Lausanne (EPFL). The event could usher in a new era of
; scientific discoveries about the workings of the human mind.
;
; The Blue Brain computer is the latest installation of IBM's
; (IBM ) BlueGene/L system, a radically new approach in
; supercomputer design. EPFL's machine has a peak speed of some
; 22.8 teraflops -- meaning it can theoretically spit out 22.8
; trillion calculations every second. That blazing speed should
; put Blue Brain among the world's top 15 supercomputers. (The
; world champ is the BlueGene system at Lawrence Livermore
; National Laboratory -- when finished later this year, it will
; have a peak speed of 367 teraflops.)
;
;
; A UNIQUE FACILITY
; Markram's EPFL team, collaborating with IBM researchers and an
; online network of brain and computer scientists, will use Blue
; Brain to create a detailed computer model of the neocortex, the
; largest and most complex part of the human brain. "That's going
; to take two to three years," he says. Then, with a bigger
; Blue Brain, he hopes to build a cellular-level model of the
; entire brain. This may take a decade -- even with IBM's
; next-generation system, BlueGene/P. Markram can't wait to get
; his hands on one of these number-crunching beasts.
;
; BlueGene/P will have faster processors and could ultimately
; reach petaflops speeds-- quadrillions of calculations per
; second. "We're planning on a very long-term effort," notes
; Markram. "We're creating a unique facility for researchers
; worldwide." Adds Charles Peck, the IBM researcher who leads
; the Blue Brain effort at IBM's research division in Yorktown
; Heights, N.Y.: "There's now a tremendous opportunity to do
; some science that up to this point just hasn't been possible."
;
; THINKING MYSTERY
; The Blue Brain Project will search for novel insights into how
; humans think and remember. Plus, by by running accurate
; simulations of brain processes, "we'll be able to investigate
; questions about psychiatric disorders and how they arise,"
; Markram says. Scientists believe that autism, schizophrenia,
; depression, and other psychological problems are caused by
; defective or malfunctioning circuitry in the brain.
;
; Parkinson's disease is another target, adds Markram. "There's a
; group of cells deep down in the mid-brain that produce
; dopamine, and when these cells begin to die and dopamine
; production decreases, you get Parkinson's," he explains. "We'll
; be able to mimic this," creating simulations that should make
; Blue Brain an invaluable tool for drug-company researchers on
; the track of treatments or cures for Parkinson's.
;
; Learning how the brain works has been one of science's great
; challenges. Researchers still don't have a holistic grasp of
; how we think. One reason: Most research so far has been
; conducted with "wet" experiments -- stimulating or dissecting
; the brains of mice, rats, and other animals. Markram notes that
; "some 'wet-lab' experiments are incredibly complicated," taking
; up to three years and costing $1 million.
;
; With simulations on Blue Brain, he predicts, "we'll be able to
; do that same work in days, maybe seconds. It's going to be
; absolutely phenomenal."
;
; CONSTANTLY CHANGING CIRCUITRY
; Markram first broached the idea of a BlueGene-based
; collaboration five years ago, right after IBM unveiled the
; supercomputer system. "Even before that, Henry had been wanting
; to go down this path of computer simulations," says IBM's Peck.
; "But only now is it actually feasible."
;
; That's because the brain is so extraordinarily complex that an
; enormously powerful computer is required. The brain's physical
; structure and electrochemical operations are very intricate.
; Complicating things still further is its constantly changing
; internal circuitry. "The brain is in a very different state in
; the morning, when you wake up, than it is at noontime," Markram
; points out.
;
; Fifty years ago, he notes, "we believed that memories were
; somehow hardwired into the brain. But our lab [EPFL's
; Laboratory of Neural Microcircuitry] has been one of the main
; propagators of a new theory, in which the brain is incredibly
; fluid. It's restructuring itself continuously --
; self-organizing and reorganizing all the time."
;
; HUGE SIMULATION
; If brain circuitry is in a constant state of flux, Markram
; insists that long-term memories can't be permanent, hardwired
; fixtures. To explain how memories are preserved, he and his
; colleagues cooked up the "liquid-computing" theory. Validating
; this concept with Blue Brain, he hints, might point to new
; types of silicon circuits that perform new and more-complex
; functions -- which IBM could use to build a revolutionary
; brain-like computer.
;
; "That's a possibility," says Tilak Agerwala, a vice-president
; at IBM Research. "But we're still very far from understanding
; how the brain works, so it's much too early to know if we
; should build computers that way." However, the notion already
; has a fancy moniker: biometaphorical computing.
;
; For now, Markram sees the BlueGene architecture as the best
; tool for modeling the brain. Blue Brain has some 8,000
; processors, and by mapping one or two simulated brain neurons
; to each processor, the computer will become a silicon replica
; of 10,000 neurons. "Then we'll interconnect them with the rules
; [in software] that we've worked out about how the brain
; functions," says Markram.
;
; The result will be a full-fledged model of 10,000 neurons
; jabbering back and forth -- a simulation 1,000 times larger
; than any similar model to date.
;
; FANTASTIC ACCELERATION
; This setup will form the foundation for studying neocortical
; columns -- the building blocks of the cortex and the part of
; the brain that differentiates mammals from other animals. Each
; column is a bundle of networked neurons and is roughly 1/2
; millimeter in diameter and 2 millimeters long. That's only
; about the size of a pinhead, Markram notes. "But packed inside
; are 50,000 neurons and more than 5 kilometers [3 miles] of
; wiring," he marvels.
;
; "The neocortical column is the beginning of intelligence and
; adaptability," Markram adds. "It marks the jump from reptiles
; to mammals." When it evolved, it was like Mother Nature had
; discovered the Pentium chip, he quips -- the circuitry "was so
; successful that it's just duplicated, with very little
; variation, from mouse to man. In the human cortex, there are
; just more cortical columns -- about 1 million."
;
; Since the neocortical column was first discovered 40 years ago,
; researchers have been painstakingly unraveling how it helps
; perform the miracles of thought that enable humans to be
; creative, inventive, philosophical creatures. "That's been my
; passion, my mission for 10 years," says Markram. "Now, we know
; how information is transferred form one neuron to another. We
; know how they behave -- what they do and whom they talk to.
; We've actually mapped that out."
;
; Next, that knowledge will be transferred into a torridly fast
; silicon simulator. Blue Brain promises a fantastic acceleration
; in brain research. It could be as dramatic as the leap from
; chiseling numbers in Sumerian clay tablets 2,500 years ago to
; crunching them in modern computers. And the Blue Brain Project
; just might culminate in a new breed of supersmart computers
; that will make even BlueGene/L seem like a piker.
;
; Otis Port is a senior writer for BusinessWeek in New York
;
Robert Myers
>
>
>It looks like we may arrive at true artificial intelligence a lot sooner than we thought we would...
>
Not likely from this particular approach.
>Quote from article below:
>
> "Markram's EPFL team, collaborating with IBM researchers
> and an online network of brain and computer scientists,
> will use Blue Brain to create a detailed computer model
> of the neocortex, the largest and most complex part of
> the human brain. "That's going to take two to three years,"
> he says. Then, with a bigger Blue Brain, he hopes to build
> a cellular-level model of the entire brain. This may take
> a decade..."
>
I'll be fascinated to see how the connectivity issue with Blue Gene
plays out, or whether the issue is even allowed to surface. The rigid
compartmentalization and problem layout with nearest-neighbor
communication imposed by the Blue Gene architecture sounds wrong for
simulating the human brain, but what do I know?
RM
Jan Vorbrüggen
> Not likely from this particular approach.
Agreed.
> I'll be fascinated to see how the connectivity issue with Blue Gene
> plays out, or whether the issue is even allowed to surface. The rigid
> compartmentalization and problem layout with nearest-neighbor
> communication imposed by the Blue Gene architecture sounds wrong for
> simulating the human brain, but what do I know?
Your cortex uses two-level wiring: There's short-distance wiring using
non-myelinated fibes. Due to the delays introduced - signal speed is less
than a meter per second - these are at most a few tens of millimeters long,
and most are significantly shorter. The long distance wiring uses myelinated
fibres and is what you will find in your brain atlas as the white matter.
(The grey matter are the neurons mixed with the short-distance wiring - a
layer only a few millimeters thick on the surface of the cortex.) While
their speed is up to 100 m/s, there is a much smaller number of them - I
remember something 3 orders of magnitude less; nonetheless, due to the bulk
of the myelination, they actually occupy most of the cortex's volume.
Of course, the brain has the advantage of using three-dimensional, self-
connecting and adaptive wiring.
Jan
John Fields
<feedbac...@yahoo.com> wrote:
>Markram's statement above is very discouraging ... using 1 processor
>for each neuron, and ending up with a net with *ONLY* 10,000 neurons.
>Given 100B neurons in the brain, he's only off by about 10-million
>"orders of magnitude" on solving the problem.
---
Seven orders of magnitude.
1E11
------ = 1E7
1E4
--
John Fields
Professional Circuit Designer
Jan Vorbrüggen
>>for each neuron, and ending up with a net with *ONLY* 10,000 neurons.
>>Given 100B neurons in the brain, he's only off by about 10-million
>>"orders of magnitude" on solving the problem.
>
>
> 1E11
> ------ = 1E7
> 1E4
It's one less - cortex is only about 10B neurons, the other 90% are mostly
in the cerebellum doing God knows what.
Jan
Sir Frederick
Delusions such as qualia?
Sensor qualia such as color, odor, sound, etc.?
Self qualia such as personhood, consciousness,
awareness, etc.?
Religious qualia such as wonder, meaning,
conscience, etc.?
--
Best,
Frederick Martin McNeill
Poway, California, United States of America
mmcn...@fuzzysys.com
http://www.fuzzysys.com
http://members.cox.net/fmmcneill/
*************************
Phrase of the week :
"Somewhere in this process, you will come face-to-face
with the sudden and shocking realization that you are
completely crazy. Your mind is a shrieking, gibbering
madhouse on wheels barreling pell-mell down the hill,
utterly out of control and hopeless. No problem.
You are not crazier than you were yesterday.
It has always been this way and you never noticed."
- Henepola Gunaratama
:-))))Snort!)
*************************
alan jones
> How do you give a machine delusions?
> Delusions such as qualia?
> Sensor qualia such as color, odor, sound, etc.?
> Self qualia such as personhood, consciousness,
> awareness, etc.?
> Religious qualia such as wonder, meaning,
> conscience, etc.?
A short answer would say, you bind the machine's state,
whatever that be, to the information it processes.
Sir Frederick
>Sir Frederick wrote:
>
>> How do you give a machine delusions?
>> Delusions such as qualia?
>> Sensor qualia such as color, odor, sound, etc.?
>> Self qualia such as personhood, consciousness,
>> awareness, etc.?
>> Religious qualia such as wonder, meaning,
>> conscience, etc.?
>
>
>A short answer would say, you bind the machine's state,
>whatever that be, to the information it processes.
>
Does that assure us that the subjective experiences associated
with those delusions will be experienced by the machine?
I know we primate folk take our delusions for granted,
but here we have a sibling machine, wouldn't want to
"short change" it.
We want a person type machine, not a zombie type.
(Though some known defective humanoids are ersatz persons.)
Guy Macon
Content-Transfer-Encoding: 8Bit
So that's about 12 years if Moore's "Law" holds...
SioL
> Markram's statement above is very discouraging ... using 1 processor
> for each neuron, and ending up with a net with *ONLY* 10,000 neurons.
> Given 100B neurons in the brain, he's only off by about 10-million
> good functional model of cortical columns would go a long way towards
> understanding brain operation.
One tiny column is 50.000 neurons, so this will only be 20%....
As usually technically uneducated journalists get overly excited about something
they don't have a clue about. I hope it helps raise funds, at least.
--
Siol
------------------------------------------------
Rather than a heartless beep
Or a rude error message,
See these simple words: "File not found."
alan jones
You say we take our delusions for granted, but I wonder to
what extent we enter into agreements on what we experience,
only accepting as delusions those things for which there is
no agreement.
As much as we share very similar hardware, our experiences
are communicated by agreements. A machine would need to lean
our agreements to communicate its experience.
Leaving aside that aspect of imagination which creates its
own agreements, I wonder if communication, internal and
external, is the same as experience? Do we think in terms
of language / vision? When i say i love, or i feel pain, do
we mean the same thing? Can i assume experience, from the
language used to describe it? .
Put another way, a machine might see the color red, as a
particular frequency of light, its qualia of that experience,
might be everything else it associates with that color. Bound
to its internal state, as various systems are made to act
upon that experience, it would form a qualia of a sorts. Could
we accept that? Would that be the same as our experience of
the color? I doubt we would be satisfied, since we know we
have evolved bound to the many advantages of seeing the color
red. Yet it would be experience.
BTw I share your goal for this machine, yet i wonder where
the person is found, in the person. Is it found in the
difference between people, the differences between man and the
rest of nature? Or is it found in what man can claim as common?
Could we ever view a machine as person-like, and accept where
it differs, or would be reject the attempt, since from the
start we know it will always be different. Even when the
machine exceeds our capacity for thought, it would always be
classed a non-person.
RyanT
nature. I've noticed that computer systems and the internet works very
similarly to my old workplace which was the library. They're called
information sciences for a reason, I guess.
But models will be models and I don't really think it would be able to
*replace* humans. If anything, we'll come up with something that might
be better than us in certain ways, but it'll just be different, not
better.
Curt Welch
I suspect they aren't worring about the archtecture difference at this
point. They probably just want access to a lot of computing power to run
larger simulations to see how they behave.
I suspsect there are still some very important "parts" they don't unerstand
yet and when they try to simulate what they do know, all they will find is
that it's not doing anything "intelligent", which only tells them they are
still missing something important.
I think it's important to try these simulations to help lead to better
understanding of the parts, but I don't really expect it to make the type
of progress they seem to be hoping for. They really seem to be talking as
if the _only_ thing missing is a big enough machine to run the simulation,
and when they get that, they will have a human brain simulation creating
human behavior.
That would only be true if they correctly understood what the brain was
doing. I don't believe they understand that yet. And when they build a
supercomputer sized simulation of the wrong low level behavior, all they
will get out of it is billions of numbers per second of garbage. Only when
they get the full and complete picture of what is happening at the low
level will the simulation do something interesting.
But maybe they understand more than I realize?
--
Curt Welch http://CurtWelch.Com/
cu...@kcwc.com http://NewsReader.Com/
inv...@example.com
On 09 Jun 2005 11:59:59 GMT, J.Random.User <NOS...@NOSPAM.COM> wrote:
>It looks like we may arrive at true artificial intelligence a lot sooner than we thought we would...
>
>Quote from article below:
>
> "Markram's EPFL team, collaborating with IBM researchers
> and an online network of brain and computer scientists,
> will use Blue Brain to create a detailed computer model
> of the neocortex, the largest and most complex part of
> the human brain. "That's going to take two to three years,"
> he says. Then, with a bigger Blue Brain, he hopes to build
> a cellular-level model of the entire brain. This may take
> a decade..."
There used to be talk about doing a full computer models of
an ant/cockroach/flatworm/something-else-with-a-small-brain
that would do what the actual organism does in all situations.
There were even claims (with fuzzy details) that it had already
been done. *Has* such a simulation ever been done, and if so
how well did it work?
Sir Frederick
>The way I see it is that technology always tries to model itself after
>nature.
That's true. Fortunately our models of what nature is up to are constantly
being refined. At this time our culture supports abysmally anachronistic
models of brain structure and function. We are stuck similarly as the medical
community was stuck with the miasma model of disease or the engineering
community was stuck with the phlogiston model of heat.
>I've noticed that computer systems and the internet works very
>similarly to my old workplace which was the library. They're called
>information sciences for a reason, I guess.
>
>But models will be models and I don't really think it would be able to
>*replace* humans. If anything, we'll come up with something that might
>be better than us in certain ways, but it'll just be different, not
>better.
The machines probably won't *replace* us, they will be good
siblings and friends. The machines will help us produce improved
genetic modifications that will compete with the *natural* model.
When (and if) we encounter intelligent extraterrestrials we probably will learn
truly different ways, perhaps better. A putative ET race that has been around for a
billion years should have something to show for it.
Bob Monsen
As a side note, I suspect that computers (as such) days are numbered.
The problem is I/O. They won't be able to compete with information
systems that attach directly to our own 'wetware'. However, once that
happens, who is to say where our own consciousness ends, and the
'computer' begins? This is already happening, in a crude form. Many
people use Google as a kind of memory... once we can simply 'remember'
things on Google, as needed, what *are* we?
My view is that at that point, *we* will be the AIs. There won't be a
need for clumsy separate entities, with their own desires. We will
provide the 'intentionality', 'it' will provide the communication,
memory and computation.
If the computational and biological facilities become sufficiently
advanced, there is no telling what will happen to us, where 'we' will
end, and 'it' will begin. A whole new chapter.
---
Regards,
Bob Monsen
Guy Macon
Colonel Forbin wrote:
>
>J.Random.User <NOS...@NOSPAM.COM> wrote:
>>
>> "Markram's EPFL team, collaborating with IBM researchers
>> and an online network of brain and computer scientists,
>> will use Blue Brain to create a detailed computer model
>> of the neocortex, the largest and most complex part of
>> the human brain. "That's going to take two to three years,"
>> he says. Then, with a bigger Blue Brain, he hopes to build
>> a cellular-level model of the entire brain. This may take
>> a decade..."
>
>The quality of the simulation is dubious at best. From the
>physiological perspective, to fully simulate a single neuron
>in terms of all the relevant processes involved in receptor
>up and down regulation, protein synthesis, gene expression,
>autocrine, paracrine and endocrine responses, etc., all of
>which are vital to excitability, it might take the entire
>computer to do even a remotely decent simulation of one
>neuron, and that would assume we knew a lot more than we
>do now about all these processes.
I disagree. As simulations go, a neuron lacks the attributes
that tend to eat up huge amounts of CPU cycles and memory.
Most of the aspects that you list change rather slowly, which
eases the computational load considereably. I do agree about
the need to have an accurate model, of course, but that requires
hard work by scientists, not hard work by computers.
>If one wanted to draw a crude analogy between such a
>computer and a living animal brain, one might be tempted
>to equate one CPU with one neuron. In that case, one
>would have to equate the biologically important components
>such as amino acids and nucleotides to the transistors that
>make up the CPU. While it would be possible to multitask
>the simulation since not all of these things change in
>outwardly physiologically significant ways at nanosecond
>intervals, the number of transistors required would far
>exceed any known fabrication process, because the brain
>is an analog device, so modeling it with a digital device
>requires vastly more logic gates.
Here is an example of a neuron simulator that handles the
sorts of things you list above and which runs on ordinary PCs:
[ http://www.neuron.yale.edu/neuron/ ]
[ http://www.neuron.yale.edu/neuron/about/what.html ]
[ http://big.sfn.org/NDG/site/eavData.asp?o=28956 ]
I can imagine a future version that contains a lot more knowledge
about how neurons work that isn't all that much bigger than the
version now available.
>OTOH, if the goal is to model consciousness, most of the
>efforts in that direction have been quite abstract and
>departed largely from the microanatomy and physiology
>of the human brain. Rather, they have been modeled on
>more abstract observations of how the brain appears to
>behave in response to environmental stimuli.
This is to be expected. What's the point of making efforts that
require a simulation of a human brain when nobody appears to have
been able to simulate a flatworm or ant? That would be like
making efforts towards chessplaying automation in the days of
sliderules.
Jan Vorbrüggen
> doing. I don't believe they understand that yet. And when they build a
> supercomputer sized simulation of the wrong low level behavior, all they
> will get out of it is billions of numbers per second of garbage.
Correct assessment, I believe. The proper abstractions have not yet been
found.
Jan
Martin Brown
>>> Markram's statement above is very discouraging ... using 1 processor
>>> for each neuron, and ending up with a net with *ONLY* 10,000 neurons.
>>> Given 100B neurons in the brain, he's only off by about 10-million
>>> "orders of magnitude" on solving the problem.
>>
>> Seven orders of magnitude.
> It's one less - cortex is only about 10B neurons, the other 90% are mostly
> in the cerebellum doing God knows what.
I guess they are aiming for a very accurate realtime simulation of a
chunk of neurons about the same size as they can acheive in the lab by
wet chemistry. It will be interesting to see how well the refined
computer model can emulate actual behaviour.
I would be more interested to see a rough and ready simulation of a much
larger chunk of neurons and interconnects not necessarily in realtime.
Astronomy simulations of gravitating particles have just recently
crossed the 10^10 barrier eg.
http://www.pparc.ac.uk/Nw/millennium_sim.asp
Regards,
Martin Brown
Jan Vorbrüggen
> chunk of neurons about the same size as they can acheive in the lab by
> wet chemistry. It will be interesting to see how well the refined
> computer model can emulate actual behaviour.
...the problem being that it is currently not possible to _measure_ actual
behaviour in a way that it could be usefully related to the model's behaviour.
Jan
Frithiof Andreas Jensen
"J.Random.User" <NOS...@NOSPAM.COM> skrev i en meddelelse
news:11agapb...@corp.supernews.com...
>
>
> It looks like we may arrive at true artificial intelligence a lot sooner
> than we thought we would...
Is this good?
Eventually, some jerk is going to succeed in brute-forcing AI on some
quantum computer - without really knowing what he is dealing with - A thing
that learns at an exponential rate, is vastly smarter than himself and very
much faster, maybe smart and fast enough to learn and predict every thought
that the creators could possibly have..
A Machine God.
!Great ;-p
Then ...
Maybe that is why the Universe is so damn quiet - every intelligent lifeform
eventually creates something smarter than themselves and loose their
intelligence because all answers and needs will be provided for them by the
machines they made.
rbmye...@gmail.com
Do you really believe this? I'm not a sci-fi type, but I suspect that
machines that become smarter than their masters are a completely
worn-out sci-fi genre, just like predictions of an imminent
breakthrough in AI are by now a completely worn-out genre.
That said, AI is one of the places in computation where something
really exciting could happen, and really at just about any time, and it
needn't have anything to do with exotic new computers; in fact, I
rather think it won't. That's because mostly what seems to have been
accomplished so far in AI is the realization that no one has a clue.
As soon as someone gets a clue, maybe something really interesting will
happen. Until that first important clue arrives, I wouldn't be too
worried about machine gods. It's so sixties.
RM
Frithiof Andreas Jensen
<rbmye...@gmail.com> skrev i en meddelelse
news:1118512983....@o13g2000cwo.googlegroups.com...
> Do you really believe this? I'm not a sci-fi type, but I suspect that
> machines that become smarter than their masters are a completely
> worn-out sci-fi genre, just like predictions of an imminent
> breakthrough in AI are by now a completely worn-out genre.
Even when something is not in vogue it still happens - like a religious war
f.ex. One civilisation must destroy another to survive. Such a 12'th century
concept, is it not?
> That said, AI is one of the places in computation where something
> really exciting could happen, and really at just about any time, and it
> needn't have anything to do with exotic new computers; in fact, I
> rather think it won't. That's because mostly what seems to have been
> accomplished so far in AI is the realization that no one has a clue.
Too much top down design, planning and flawed modelling is what kills "AI"
I.M.O.
Real brains are chaotic, adaptive and evolved - not carefully designed and
planned.
> As soon as someone gets a clue, maybe something really interesting will
> happen.
The Point is:
Someone might not actually *need* a clue for something really interesting to
happen. Maybe general principles and lots of processing power is enough.
Maybe Intelligence is really simple to build, simply a property of certain
kinds of connected networks that we have not stumbled upon yet or build
large enough.
Like handing some U-235 to a medieval cannon-ball maker, "oh, them new
cannon balls are heavy and hard, just what we need - so let's make a Really
Big One" ;-)
Robert Myers
<frithio...@diespammerdie.jensen.tdcadsl.dk> wrote:
>
><rbmye...@gmail.com> skrev i en meddelelse
>news:1118512983....@o13g2000cwo.googlegroups.com...
>
>>... mostly what seems to have been
>> accomplished so far in AI is the realization that no one has a clue.
>
>Too much top down design, planning and flawed modelling is what kills "AI"
>I.M.O.
>Real brains are chaotic, adaptive and evolved - not carefully designed and
>planned.
>
There has been plenty of work on self-organizing systems. I haven't
heard anyone who's actually worked in the field say, "If only we had a
bigger or faster computer, it might work." Maybe the someone to say
that is out there and I can elicit that comment from someone who knows
what he's talking about from this post.
>> As soon as someone gets a clue, maybe something really interesting will
>> happen.
>
>The Point is:
>
>Someone might not actually *need* a clue for something really interesting to
>happen. Maybe general principles and lots of processing power is enough.
>Maybe Intelligence is really simple to build, simply a property of certain
>kinds of connected networks that we have not stumbled upon yet or build
>large enough.
>
The right idea often turns out to be stunningly simple once you have
the right idea. So far, no one seems to have stumbled onto the right
idea, and building big computers to see if they will turn themselves
into a thinking machine by random fiddling doesn't seem like a
promising way to go.
>Like handing some U-235 to a medieval cannon-ball maker, "oh, them new
>cannon balls are heavy and hard, just what we need - so let's make a Really
>Big One" ;-)
>
Have you read, say, Richard Rhodes on the making of the atomic bomb?
A significant amount of basic science by some very bright people went
into designing it, and the those laying the money bets had a pretty
good idea that it would work and why. Do you have any similar
confidence for any scheme for artificial intelligence?
RM
Curt Welch
> On Sun, 12 Jun 2005 14:03:32 +0200, "Frithiof Andreas Jensen"
> <frithio...@diespammerdie.jensen.tdcadsl.dk> wrote:
>
> >
> ><rbmye...@gmail.com> skrev i en meddelelse
> >news:1118512983....@o13g2000cwo.googlegroups.com...
> >
> >>... mostly what seems to have been
> >> accomplished so far in AI is the realization that no one has a clue.
> >
> >Too much top down design, planning and flawed modelling is what kills
> >"AI" I.M.O.
> >Real brains are chaotic, adaptive and evolved - not carefully designed
> >and planned.
> >
> There has been plenty of work on self-organizing systems. I haven't
> heard anyone who's actually worked in the field say, "If only we had a
> bigger or faster computer, it might work." Maybe the someone to say
> that is out there and I can elicit that comment from someone who knows
> what he's talking about from this post.
There have been many people over the years who felt that they had the right
approch but just needed a biger or faster version for it to be
"intelligent".
Most people however are wiser than that. If it's not acting intelligent on
the small scale, making it biger is never going to get us anywhere.
> The right idea often turns out to be stunningly simple once you have
> the right idea. So far, no one seems to have stumbled onto the right
> idea,
I think I've got a stunningly simple approach that will get us there. So
far, everyone I show it to seems to believe it's too simple to be anything
new or different or useful.
Whether my network design is going to do anything useful is yet to be seen,
but I do strongly believe that intelligence is simply a strong
reinforcement learning machine, and that when we do discover how to build
this, it will be stunningly simple in comparison to the type of complexity
people have been playing with for so long in the search for AI.
SioL
> The right idea often turns out to be stunningly simple once you have
> the right idea. So far, no one seems to have stumbled onto the right
> idea, and building big computers to see if they will turn themselves
> into a thinking machine by random fiddling doesn't seem like a
> promising way to go.
Wouldn't it be cool to have AI in a AVR, you'd just connect a mic
and tell it:
Howdy there partner, would you flash that LED once a second for me?
Frank Bemelman
news:1f%qe.12667$F6.27...@news.siol.net...
> "Robert Myers" <rmyer...@comcast.net> wrote in message
news:6n9oa1ho94mo50nji...@4ax.com...
> > The right idea often turns out to be stunningly simple once you have
> > the right idea. So far, no one seems to have stumbled onto the right
> > idea, and building big computers to see if they will turn themselves
> > into a thinking machine by random fiddling doesn't seem like a
> > promising way to go.
>
> Wouldn't it be cool to have AI in a AVR, you'd just connect a mic
> and tell it:
>
> Howdy there partner, would you flash that LED once a second for me?
Ask Eliza what would happen if you did:
http://www-ai.ijs.si/cgi-bin/eliza/eliza_script
--
Thanks, Frank.
(remove 'q' and 'invalid' when replying by email)
Guy Macon
SioL wrote:
>
>Wouldn't it be cool to have AI in a AVR, you'd just connect a mic
>and tell it:
>
>Howdy there partner, would you flash that LED once a second for me?
Then you would have to listen to it complain about how all the diodes
on it's left have begun to ache...
Del Cecchi
"Colonel Forbin" <for...@dev.nul> wrote in message
news:IIYqe.35931$JX5....@tornado.ohiordc.rr.com...
> In article <6n9oa1ho94mo50nji...@4ax.com>,
> Robert Myers <rmy...@rustuck.com> wrote:
>>
>>The right idea often turns out to be stunningly simple once you have
>>the right idea. So far, no one seems to have stumbled onto the right
>>idea, and building big computers to see if they will turn themselves
>>into a thinking machine by random fiddling doesn't seem like a
>>promising way to go.
>
> this way in evolving intelligent life, but perhaps we haven't found
> the right idea yet. Or perhaps there isn't one.
>
Biological Science takes the above statement on blind faith. I think
they haven't got a clue about how it happened, and invoking "Nature" as
if it were an entity when it is nothing and doesn't exist, is only
covering up that fact. What is this "Nature"? Anything like Gaia?
One would think scientists would be a little more honest.
del cecchi
Robert Myers
wrote:
>In article <3h4237F...@individual.net>,
>"Nature" is simply what is, not some conscious anthropomorphized
>entity. The Darwinian theory of evolution is largely thought to
>have involved random mutations, the most adapted of which survived
>to reproduce and spread their traits.
>
The safe statement here is: there's lots of exciting science left to
be done.
>With respect to AI, there seem to be two main camps. One wants to
>emulate the behavior of intelligent beings, the other wants to
>emulate the biophysics.
The field could be divided into those who are trying to identify
intelligence as emergent behavior and those who are trying to invent
algorithms that mimic intelligent behavior.
The class of those who are looking for intelligence as emergent
behavior is really much larger than those trying to do biophysics (or
biochemistry, or neuroanatomy or whatever). Identifying intelligence
as an emergent behavior of any system would be an exciting
breakthrough whether the simulated system had anything to do with
natural systems or not.
As distinct from building supercomputers and hoping that they will
turn into intelligent beings, the study of emergent behaviors seems to
be a serious activity that is worthwhile in its own right. If anyone
showed that size matters in the sense that a bigger system could
display behavior that is qualitatively different from what a smaller
system could display, such a result, in itself, would be fascinating;
but the current level of understanding wouldn't seem to warrant the
use of gigantic computers.
>It's not clear that either approach will
>result in any huge breakthroughs in understanding consciousness.
>
It isn't enough that we've danced on the edge of the volcano in
discussing evolution and intelligence? Now you want to drag
consciousness into the discussion? Consciousness is an entirely
separate puzzle. I don't think anyone really knows what consciousness
is, never mind how to get something like a computer to be conscious.
>There was some promise in the careful study of the effects of
>certain substances on perception and consciousness, but this
>research was largely suppressed for political reasons around
>1970 and nearly completely so later on for reasons which were
>and remain largely specious.
>
Oh, piffle. Some professors at Harvard were having fun with drugs.
The politics, the mythology, the recreational aspects of the story are
all interesting, but they have nothing to do with computers,
artificial intelligence, or even, in my humble opinion, the search for
the origins of consciousness.
RM
Jan Vorbrüggen
> heard anyone who's actually worked in the field say, "If only we had a
> bigger or faster computer, it might work." Maybe the someone to say
> that is out there and I can elicit that comment from someone who knows
> what he's talking about from this post.
As somebody who (used to be) in the field of self-organizing systems,
that is mostly true - we just have no good idea what the general principles
are that are needed for an intelligent system.
There does seem to be some effect of quantity with respect to intelligence,
however, for instance if you compare a bonobo and a human being.
> The right idea often turns out to be stunningly simple once you have
> the right idea. So far, no one seems to have stumbled onto the right
> idea, and building big computers to see if they will turn themselves
> into a thinking machine by random fiddling doesn't seem like a
> promising way to go.
In particular, I believe that, to a large part, intelligence requires the
use of experimentation to develop. Thus, you need sensors _and_ actuators
- or a body to go with the mind.
Jan
Robert Myers
<jvorbrue...@mediasec.de> wrote:
>> There has been plenty of work on self-organizing systems. I haven't
>> heard anyone who's actually worked in the field say, "If only we had a
>> bigger or faster computer, it might work." Maybe the someone to say
>> that is out there and I can elicit that comment from someone who knows
>> what he's talking about from this post.
>
>As somebody who (used to be) in the field of self-organizing systems,
>that is mostly true - we just have no good idea what the general principles
>are that are needed for an intelligent system.
>
>There does seem to be some effect of quantity with respect to intelligence,
>however, for instance if you compare a bonobo and a human being.
>
I thought there was something qualitatively different about the
neocortex.
>> The right idea often turns out to be stunningly simple once you have
>> the right idea. So far, no one seems to have stumbled onto the right
>> idea, and building big computers to see if they will turn themselves
>> into a thinking machine by random fiddling doesn't seem like a
>> promising way to go.
>
>In particular, I believe that, to a large part, intelligence requires the
>use of experimentation to develop. Thus, you need sensors _and_ actuators
>- or a body to go with the mind.
>
The system needs an environment in which to learn to cope, goals
required for coping, and a way of interacting with the environment
that allows for experimentation and learning.
Physical objects introduce the possibility of introducing something
you didn't think to introduce intentionally.
I think Boston's public radio station is set to discuss the Blue Brain
silliness today at 11:00 am EDT on The Connection. I heard the
announcement, but www.wbur.org doesn't have the slot description;
sounds like a desparate last-minute fillin. Can machines be made
conscious? If so, is it ethical to unplug them? Groan.
RM
jsa...@ecn.ab.ca
> The rigid
> compartmentalization and problem layout with nearest-neighbor
> communication imposed by the Blue Gene architecture sounds wrong for
> simulating the human brain, but what do I know?
Well, that does mean that the nodes will spend some of their time
passing messages along, instead of just simulating neurons, and, of
course, each node will simulate more than one neuron anyways.
It is indeed a non-optimal architecture, but the benefits of a more
flexible interconnection layout would be small, since a computer with a
flexible interconnect scheme still wouldn't allow a single node as many
distant interconnections as thousands of neurons might have.
If this architecture allows a more economical layout with very large
numbers of processors, that likely will be more important.
John Savard
Jan Vorbrüggen
> neocortex.
I don't think you can easily distinguish primates's brains from different
species just looking under the microscope...I believe that you'd have
difficulty even with rat and human tissue. That's the whole point of
mammalian neocortex: it's the quitessential information processing engine,
as far as we can tell, and it's only the way it's wired to peripherals,
as well as its interactions with its peers (other parts of neocortex)
that seem to determine its function.
> The system needs an environment in which to learn to cope, goals
> required for coping, and a way of interacting with the environment
> that allows for experimentation and learning.
Just so.
> Physical objects introduce the possibility of introducing something
> you didn't think to introduce intentionally.
Well, you could do this in a simulated world, I suppose. Then you might
run into the problem one of the guys from Thinking Machines had. It apears
he set up a simulation for an artificial life model as a test for the
CM-5 (IIRC). He said he debugged his physical world simulation by looking
at the creatures that evolved. For instance, some of his first versions
didn't quite get the principles of thermodynamics correct, so that some
of the "animals" learnt to turn in a circle to gain energy for a jump 8-).
Jan
rbmye...@gmail.com
> > I thought there was something qualitatively different about the
> > neocortex.
>
> I don't think you can easily distinguish primates's brains from different
> species just looking under the microscope...I believe that you'd have
> difficulty even with rat and human tissue. That's the whole point of
> mammalian neocortex: it's the quitessential information processing engine,
> as far as we can tell, and it's only the way it's wired to peripherals,
> as well as its interactions with its peers (other parts of neocortex)
> that seem to determine its function.
>
By chance there was a piece on the radio this morning about how the
songs of bird species are more heavily influenced by location than by
genetics: that's how we get a western something-or-other and and
eastern something-or-other, same plumage, characteristically different
song.
Maybe not only brain size, but learned behavior, the pace of brain
development and the length of the child-rearing period (incredibly long
for humans) and other environmental factors are important. If that's
so, and machine intelligence is to come from mimicking natural
processes, we're going to be waiting a long time for such a program to
come to fruition. Not only do you need a *very* big computer, and, as
you point out, a way of interacting with its environment to learn,
you'd also have to provide it with a great deal of "nurturing."
RM
Sleepyhead
... and if learning to think just is the process of neural
interconnection established by training then we won't learn anything
much by mimicking a static body of neurons.
Guy Macon
jsa...@ecn.ab.ca wrote:
>It is indeed a non-optimal architecture, but the benefits of a more
>flexible interconnection layout would be small, since a computer with a
>flexible interconnect scheme still wouldn't allow a single node as many
>distant interconnections as thousands of neurons might have.
I am not sure how to apply the concept "distant interconnections" to
the concept of "thousands of neurons." Are you calling a chain of
local interconnections a distant interconnection? That's not the
usage that computer scientists are taling about when they talk about
distant interconnections.
If you apply the definition of "distant interconnections" used in
computers to the brain, ther aren't any. Most neurons have an axon
tree that is less than 1 or 2 nM across[1]. Although some partly
myelinated axons can extend several millimeters in the neocortex
and hippocampus, most neurons only have somewhat local connections.
References:
[1]
Buhl EH, Halasy K, Somogyi P Diverse sources of hippocampal
unitary inhibitory postsynaptic potentials and the number of synaptic
release sites. Nature 368:823-828 (1994)
Sik A, Penttonen M, Ylinen A, Buzski G Hippocampal CA1
interneurons: an in vivo intracellular labeling study.
Journal of Neuroscience 15:6651-6665 (1995)
[2]
Sik A, Ylinen A, Penttonen M, Buzski G, Inhibitory
CA1-CA3-hilar region feedback in the hippocampus.
Science 265:1722-1724 (1994)
[3]
Kisvarday ZF, Beaulieu C, Eysel UT, Network of GABAergic
large basket cells in cat visual cortex (area 18):
implication for lateral disinhibition.
Journal of Computational Neurology 327:398-415 (1993)
Gupta A, Wang Y, Markram H, Organizing principles for a
diversity of GABAergic interneurons and synapses in the neocortex.
Science 287:273-278 (2000)
Salin PA, Prince DA, Electrophysiological mapping of GABAA
receptor- mediated inhibition in adult rat somatosensory cortex.
Journal of Neurophysiology 75:1589-1600 (1996)
--
Guy Macon <http://www.guymacon.com/>
Glen M. Sizemore
emulate the behavior of intelligent beings, the other wants to
result in any huge breakthroughs in understanding consciousness.
GS: This does not strike me as accurate, or useful. Any AI must simulate
animal behavior, at least in the same sense the flight of planes achieves
some of the properties of flight in animals. A plane does not, as we have
heard ad naseum, flap its wings, but it does go long distances without
touching the ground, and it is not merely gliding. [Of course,
interestingly, when we hear how AI might not resemble "natural intelligence"
and are fed the plane/bird argument, we are not told that it is the case
that the flight of the glider and the "flight" of the first animals DO
closely resemble each other]. Is a connectionistic approach the former or
the latter? I'm guessing that you would say that it is the latter (i.e.,
emulate biophysics), but this shows the conceptual error I am pointing out.
If the NN does not emulate some functions of animals, then in what sense is
it "AI." Incidentally - or maybe not so incidentally - a big part of the
problem is the term "intelligence." One should consider that the term does
not identify anything useful about behavior OR the brain.
CF: There was some promise in the careful study of the effects of
certain substances on perception and consciousness, but this
research was largely suppressed for political reasons around
1970 and nearly completely so later on for reasons which were
and remain largely specious.
GS: I'm not sure what you are talking about here, but it can't be the field
that continues to investigate the effects of drugs on behavior (i.e.,
behavioral pharmacology).
"Colonel Forbin" <for...@dev.nul> wrote in message
news:qu6re.38241$JX5....@tornado.ohiordc.rr.com...
> In article <3h4237F...@individual.net>,
> Del Cecchi <dcecchi...@att.net> wrote:
> >
> >
> >
> "Nature" is simply what is, not some conscious anthropomorphized
> entity. The Darwinian theory of evolution is largely thought to
> have involved random mutations, the most adapted of which survived
> to reproduce and spread their traits.
>
> With respect to AI, there seem to be two main camps. One wants to
> emulate the behavior of intelligent beings, the other wants to
> emulate the biophysics. It's not clear that either approach will
> result in any huge breakthroughs in understanding consciousness.
>
pr...@prep.synonet.com
> The class of those who are looking for intelligence as emergent
> behavior is really much larger than those trying to do biophysics
> (or biochemistry, or neuroanatomy or whatever). Identifying
> intelligence as an emergent behavior of any system would be an
> exciting breakthrough whether the simulated system had anything to
> do with natural systems or not.
One of your MIT friends did a PhD on self directed learning of speach.
As a side effect, he had a data compressor that is better than gzip.
Paper is on xxx, sorry I can't remember the name. This was 96-97 or
so.
--
Paul Repacholi 1 Crescent Rd.,
+61 (08) 9257-1001 Kalamunda.
West Australia 6076
comp.os.vms,- The Older, Grumpier Slashdot
Raw, Cooked or Well-done, it's all half baked.
EPIC, The Architecture of the future, always has been, always will be.
Del Cecchi
"Colonel Forbin" <for...@dev.nul> wrote in message
> In article <3h4237F...@individual.net>,
> Del Cecchi <dcecchi...@att.net> wrote:
>>
>>
>>
> entity. The Darwinian theory of evolution is largely thought to
> have involved random mutations, the most adapted of which survived
> to reproduce and spread their traits.
Of course. They teach that as revealed dogma in kindergarten these days.
So don't speak of a random process as if it were some sort of being as
you did in the original passage. Nature does this, nature does that,
bullshit that could all be replaced by "shit happens". But that doesn't
sound nearly as plausible as "as the waters receded, nature made the
fishies grow lungs"
or "Biological Science tells us that nature worked this way in evolving
intelligence". Intelligence happened. We have not a clue how.
Concienceness happened. We have not a clue what it is, or how it came
into being. Random process? Divine intervention? ET? There is
basically no evidence much for any of them.
Is intelligence and self awareness inevitable? Common? Rare? Where is
everyone?
>
> With respect to AI, there seem to be two main camps. One wants to
> emulate the behavior of intelligent beings, the other wants to
> emulate the biophysics. It's not clear that either approach will
> result in any huge breakthroughs in understanding consciousness.
>
> There was some promise in the careful study of the effects of
> certain substances on perception and consciousness, but this
> research was largely suppressed for political reasons around
> 1970 and nearly completely so later on for reasons which were
> and remain largely specious.
Go lick a picture of Jerry Garcia. You are sounding like a Tim Leary
wanna be.
>
>
Stephen Fuld
I listened to the show. They didn't talk much about the blue brain stuff.
More about intelligence and consciousness in general. They had Marvin
Minsky and Paul Davies, among others as guests. It wasn't bad for a show
aimed at the general public. The moderator tried to get people to talk
about the "moral issues", but mostly they didn't.
So I still don't see what is so new about Blue Brain, especially compared to
the existing parallel cortical simulation software such as NCS
(NeuroCortical Simulator) that runs on Beowolf clusters and, if an offhand
comment on their web site is to be believed, on Blue Gene.
http://www.interstice.com/drewes/brain/brainlab.html
Note that Brainlab is a front end to simplify setting up models for NCS.
--
- Stephen Fuld
e-mail address disguised to prevent spam
Maynard Handley
realized that the name of the simulation is Blue BRAIN not Blue MIND.
The scientists doing this do not pretend they are studying the mind (of
course one cannot speak for the PR people and journalists); they are
interested in simulating, right now, a very tiny structure in the brain
consisting of about 10,000 neurons. (Not just 10,000 random neurons ala
neural net. I don't know much brain anatomy, but apparently these 10,000
neuron elements make up distinctive elements of the brain that are
replicated to form the larger brain.)
Obviously 10,000 neurons is not a full human brain. Obviously 10,000
neurons is also more brain than you get in certain animals --- clearly
more than in a C. Elegans, and perhaps more than in a fairy fly.
If everything goes well, and if the calculations appear to provide
useful and intriguing results, this will be scaled up to cover
interacting versions of these 10,000 neuron elements.
The point, however, is not to simulate consciousness; it's to see if
anything surprising and useful comes out of treating the brain as a very
large neural net. If you believe Penrose there are limits to what will
happen bcs an important aspect of the brain, some sort of quantum
related stuff, is not being simulated. If you believe Timothy Leary
types, there are limits to what will happen bcs brain chemistry is not
(as far as I know) being simulated. Beyond brain chemistry there are
plenty of other things that are not being simulated --- nothing about
protein synthesis, nothing about cytoskeleton rearrangement.
Simply stating these issues makes part of the point of the exercise a
little more clear --- what CAN come out of such a simulation (and thus
is dependent only on neural connections) and what, if anything, appears
unable to come out of such a simulation?
And, of course, way before consciousness, there are plenty of more
trivial brain functionalities we barely understand; things like vision
and hearing and manipulating muscles in real-time to fly.
Maynard
Glen M. Sizemore
to get something like a computer to be conscious."
GS: I think that there are a good many people who know what "consciousness"
is. First, such people know that the word has a variety of meanings (i.e.,
its "use" is caused by different things at different times) and, second,
they know that one use of "consciousness" involves our acquired ability to
observe and talk about our own behavior. That is, essentially, what there is
to say. When you understand that perception is behavior, and what we observe
when we report our own perceptual behavior is that behavior, the problem of
"qualia" disappears as well.
"Robert Myers" <rmyer...@comcast.net> wrote in message
news:dnlqa194dum6bg8rh...@4ax.com...
Jan Vorbrüggen
> anatomy, but apparently these 10,000 neuron elements make up distinctive
> elements of the brain that are replicated to form the larger brain.)
Yes, that a column, but it has more on the order of 50k-100k neurons.
And the same thing holds: we can't even put it to a black box test, as
we don't have any really useful input-output data for such a column.
Be aware that such a model has on the order of 20-1000 degress of freedom
_per_neuron_, so you are poking in a very highly dimensional space for
a set of states that does "something useful".
> Obviously 10,000 neurons is not a full human brain. Obviously 10,000
> neurons is also more brain than you get in certain animals --- clearly
> more than in a C. Elegans, and perhaps more than in a fairy fly.
An insect brain is on the order of a million neurons. That would at least
make it possible to do useful tests. And C. elegans neural system certainly
is also a very valid target, as not only are neurons but also all synapses
are known.
Jan
Robert Myers
<s.f...@PleaseRemove.att.net> wrote:
>
>"Robert Myers" <rmyer...@comcast.net> wrote in message
>news:cnsqa1lkfrlhb2ei2...@4ax.com...
>
>>
>> silliness today at 11:00 am EDT on The Connection. I heard the
>> announcement, but www.wbur.org doesn't have the slot description;
>> sounds like a desparate last-minute fillin. Can machines be made
>> conscious? If so, is it ethical to unplug them? Groan.
>
>I listened to the show. They didn't talk much about the blue brain stuff.
>More about intelligence and consciousness in general. They had Marvin
>Minsky and Paul Davies, among others as guests. It wasn't bad for a show
>aimed at the general public. The moderator tried to get people to talk
>about the "moral issues", but mostly they didn't.
>
http://www.theconnection.org/shows/2005/06/20050613_b_main.asp
Worth the time, IMHO, just to hear what people are willing to say with
a straight face. It seems like a fairly realistic presentation,
except that there is really very minimal emphasis on the enormous gap
between current capability (incredibly primitive) and the kinds of
ethical questions they want to explore. Some reasonably incisive
comments about self-organization and complexity.
>So I still don't see what is so new about Blue Brain, especially compared to
>the existing parallel cortical simulation software such as NCS
>(NeuroCortical Simulator) that runs on Beowolf clusters and, if an offhand
>comment on their web site is to be believed, on Blue Gene.
>
>http://www.interstice.com/drewes/brain/brainlab.html
>
>Note that Brainlab is a front end to simplify setting up models for NCS.
The important ingredient is the corporate commitment and the marketing
muscle of IBM. Whenever something interesting happens in AI, IBM
wants to be there and wants to be known as a player.
RM
jsa...@ecn.ab.ca
> If you apply the definition of "distant interconnections" used in
> computers to the brain, ther aren't any. Most neurons have an axon
> tree that is less than 1 or 2 nM across[1]. Although some partly
> myelinated axons can extend several millimeters in the neocortex
> and hippocampus, most neurons only have somewhat local connections.
That makes sense, and simplifies matters. As long as each processor
simulates more than one neuron, then, there probably will not be any
connections from a processor to other than its nearest neighbors.
However, diagonally adjacent neighbors would have to be included as
well, and even that is probably "distant" in the computer-science
sense.
So now we know the ideal topology for a computer that simulates the
brain: each processor has to have links to its *twenty-six* nearest
neighbors in 3-D.
John Savard
min...@media.mit.edu
> On Tue, 14 Jun 2005 04:39:41 GMT, "Stephen Fuld"
> <s.f...@PleaseRemove.att.net> wrote:
> >
> >"Robert Myers" <rmyer...@comcast.net> wrote in message
> >news:cnsqa1lkfrlhb2ei2...@4ax.com...
> >
> >>
> >> I think Boston's public radio station is set to discuss the Blue Brain
> >> silliness today at 11:00 am EDT on The Connection. I heard the
> >> announcement, but www.wbur.org doesn't have the slot description;
> >> sounds like a desparate last-minute fillin. Can machines be made
> >> conscious? If so, is it ethical to unplug them? Groan.
> >
> >I listened to the show. They didn't talk much about the blue brain stuff.
> >More about intelligence and consciousness in general. They had Marvin
> >Minsky and Paul Davies, among others as guests. It wasn't bad for a show
> >aimed at the general public. The moderator tried to get people to talk
> >about the "moral issues", but mostly they didn't.
> >
> I finally got a chance to listen from the streaming archive
>
> http://www.theconnection.org/shows/2005/06/20050613_b_main.asp
>
> Worth the time, IMHO, just to hear what people are willing to say with
> a straight face. It seems like a fairly realistic presentation,
> except that there is really very minimal emphasis on the enormous gap
> between current capability (incredibly primitive) and the kinds of
> ethical questions they want to explore. Some reasonably incisive
> comments about self-organization and complexity.
Just as there is an enormous gap between our incredibly primitive
current ideas about ethics and the problems they are purported to
solve. For example, Brian Cantwell Smith (who was one of the
participants) and I raised the 'thou shalt not kill' issue. If
everyone made a daily backup copy, then murder would become a mere
misdemeanor. What should be the punishment for causing someone to lose
a half-day's work-considering that, with maintainable parts, that
person already is potentially immortal.
Um, in fact, perhaps the punishment for 'killing' say, two days of
backups should be more than that for killing the person itself.
> >So I still don't see what is so new about Blue Brain, especially compared to
> >the existing parallel cortical simulation software such as NCS
> >(NeuroCortical Simulator) that runs on Beowolf clusters and, if an offhand
> >comment on their web site is to be believed, on Blue Gene.
> >
> >http://www.interstice.com/drewes/brain/brainlab.html
> >
> >Note that Brainlab is a front end to simplify setting up models for NCS.
>
> The important ingredient is the corporate commitment and the marketing
> muscle of IBM. Whenever something interesting happens in AI, IBM
> wants to be there and wants to be known as a player.
Yes, as with Chess. Their publicity about Deep Blue said almost
nothing about how they began by taking (or buying?) Deep Thought from
Carnegie-Mellon.
>
> RM
Jan Vorbrüggen
> model each neuron. Interesting that they think this level of detail is
> what it takes to do the job. Someone mentioned that, if Moore's law
> holds, then it will only be a few years before they can have a
> supercomputer node for each of the 10B-100B neurons, up from 10K with
> present system. Unless, of course, they decide by then that they need
> to push the models even further down ... say to the level of quantum
> effects on the atoms in the molecules. Oy.
Personally, I believe that a much less detailed simulation must be work-
able. Our brain has evolved to work in the face of so many varying boundary
conditions, not to mention the physical and chemical harm we throw at it
regularly, that its continued well-being (mental illnesses notwithstanding)
cannot depend on such details of its physical structure. In my view, that's
a large part of the interest in such work: What are the principles behind
building such a complex system in such a way that it remains "stable" (in
some fuzzy sense) in the face of fluctuations in its components? If we
could better understand how that is done, we might be able to solve similar
problems, e.g., how to build more stable societies or economies.
Jan
Dr. Adrian Wrigley
...
> computers to the brain, ther aren't any. Most neurons have an axon
> tree that is less than 1 or 2 nM across[1]. Although some partly
> myelinated axons can extend several millimeters in the neocortex
> and hippocampus, most neurons only have somewhat local connections.
I'm not sure what unit (nM) is meant here!
1-2 nanometres is only a few atomic diameters (3-6 atoms across)
1-2 nautical miles is silly!
1-2 nano moles is non-sensical
My guess is than mm is meant(?)
Thanks for clarifying!
--
Adrian
Del Cecchi
<min...@media.mit.edu> wrote in message
news:1118788342....@g44g2000cwa.googlegroups.com...
That makes the presuposition that a person is no more than the state of
the organic materials that comprise their body. Many are unwilling to
agree to that, in the total absence of evidence. Of course we all recall
the "lime pit" thought experiment, in which travel is accomplished by
transmitting the above hypothesized "backup data" and destroying the
original. If one has enough bandwidth to record it routinely then one
can transmit it and not worry about the bother of dealing with the actual
carbon based matter. So, how many readers of this group would use such a
means of transportation?
>
> Um, in fact, perhaps the punishment for 'killing' say, two days of
> backups should be more than that for killing the person itself.
>
>> >So I still don't see what is so new about Blue Brain, especially
>> >compared to
>> >the existing parallel cortical simulation software such as NCS
>> >(NeuroCortical Simulator) that runs on Beowolf clusters and, if an
>> >offhand
>> >comment on their web site is to be believed, on Blue Gene.
>> >
>> >http://www.interstice.com/drewes/brain/brainlab.html
>> >
>> >Note that Brainlab is a front end to simplify setting up models for
>> >NCS.
>>
>> The important ingredient is the corporate commitment and the marketing
>> muscle of IBM. Whenever something interesting happens in AI, IBM
>> wants to be there and wants to be known as a player.
>
> Yes, as with Chess. Their publicity about Deep Blue said almost
> nothing about how they began by taking (or buying?) Deep Thought from
> Carnegie-Mellon.
As I recall, they hired at least part of the CM team that did deep
thought. Those folks then did deep blue as a follow on.
del cecchi
HMS Beagle
>
>
>It looks like we may arrive at true artificial intelligence a lot sooner than we thought we would...
>
Who is "we"? What does this writer mean by "true artificial
intelligence"? How early did "we" expect to have this?
In case you the reader was not aware of what cortical columns are:
* The cerebral cortex is approximately 3-6 mm thick and
neuroanatomists have observed that the cortical neurons appear to be
organized in columns (cf. especially Mountcastle 1998).
* Cortical columns (sometimes called minicolumns), oriented
perpendicular to the cortical surface, can be likened to cylinders (of
30-50 um diameter) with length spanning the 3-6 mm depth.
* Each column contains between 80 and 110 neurons.
* Their shape is quasi-hexagonal, because each cortical column is
typically surrounded by six other columns.
* They are believed to function as modules. "A cortical column is
a complex processing and distributing unit that links a number of
inputs to a number of outputs via overlapping internal processing
chains" (Mountcastle, 1998).
>; HUGE SIMULATION
>; If brain circuitry is in a constant state of flux, Markram
>; insists that long-term memories can't be permanent, hardwired
>; fixtures. To explain how memories are preserved, he and his
>; colleagues cooked up the "liquid-computing" theory. Validating
>; this concept with Blue Brain, he hints, might point to new
>; types of silicon circuits that perform new and more-complex
>; functions -- which IBM could use to build a revolutionary
>; brain-like computer.
>;
>; "That's a possibility," says Tilak Agerwala, a vice-president
>; at IBM Research. "But we're still very far from understanding
>; how the brain works, so it's much too early to know if we
>; should build computers that way." However, the notion already
>; has a fancy moniker: biometaphorical computing.
>;
"We are still very far from understanding how the brain works."
Repeat that 10 times. Now tell me how we are going to simulate a
brain in a supercomputer. Not making sense? Its not making sense to
me either. Strange that the paragraph right above this one says IBM
could use to build a revolutionary brain-like computer. I guess the
key word is "could".
>; creative, inventive, philosophical creatures. "That's been my
>; passion, my mission for 10 years," says Markram. "Now, we know
>; how information is transferred form one neuron to another. We
>; know how they behave -- what they do and whom they talk to.
>; We've actually mapped that out."
>;
This is not accurate. First of all, we don't know what the spike
train is encoding, or even if it is an encoding scheme to begin with.
Our current models of neurons say that the neuron simply sums the
incoming signals into one number and fires if its above a threshold.
Do real biological neurons do this? We simply don't know.
>; Next, that knowledge will be transferred into a torridly fast
>; silicon simulator. Blue Brain promises a fantastic acceleration
>; in brain research. It could be as dramatic as the leap from
>; chiseling numbers in Sumerian clay tablets 2,500 years ago to
>; crunching them in modern computers. And the Blue Brain Project
>; just might culminate in a new breed of supersmart computers
>; that will make even BlueGene/L seem like a piker.
This whole article does not explain exactly what is being simulated.
And we are about to find out where below.
>;
>; Otis Port is a senior writer for BusinessWeek in New York
>;
Well, now I know not to read BusinessWeek for the latest on AI
research. A lot of times business magazines are really only read by
people who are wondering where to buy stock next. So IBM "places"
articles into these magazines to make it seem like they are on the
verge of a "breakthrough".
HMS Beagle
>I think it's important to try these simulations to help lead to better
>understanding of the parts, but I don't really expect it to make the type
>of progress they seem to be hoping for. They really seem to be talking as
>if the _only_ thing missing is a big enough machine to run the simulation,
>and when they get that, they will have a human brain simulation creating
>human behavior.
>
Considering they dont have mouse behavior yet I doubt hooking up a
bunch of CPUs will create human behavior.
>That would only be true if they correctly understood what the brain was
>doing. I don't believe they understand that yet. And when they build a
>supercomputer sized simulation of the wrong low level behavior, all they
>will get out of it is billions of numbers per second of garbage. Only when
>they get the full and complete picture of what is happening at the low
>level will the simulation do something interesting.
>
>But maybe they understand more than I realize?
They do not. This article was taken out of the magazine
BusinessWeek, and there are subtle uses of the words "maybe", "might",
and "could." Never anywhere in the article does it say he will have
an ACTUAL SIMULATION of a brain. They subtley place the word "might"
where needed. As an AI researcher yourself, you should have
realized immediately that raw computing power is not the key to AI.
The article does not completely explain EXACTLY what he is simulating.
It sounds like he is just simulating groups of neurons to check their
"self-organizing" behavior, and nothing more. Notice he says nothing
about using this "simulated brain" to control a robot that can pick up
trash in a subway (for instance). This is important.
Rich The Philosophizer
> <min...@media.mit.edu> wrote in message
...
>> Just as there is an enormous gap between our incredibly primitive
>> current ideas about ethics and the problems they are purported to
>> solve. For example, Brian Cantwell Smith (who was one of the
>> participants) and I raised the 'thou shalt not kill' issue. If
>> everyone made a daily backup copy, then murder would become a mere
>> misdemeanor. What should be the punishment for causing someone to lose
>> a half-day's work-considering that, with maintainable parts, that
>> person already is potentially immortal.
>
> That makes the presuposition that a person is no more than the state of
> the organic materials that comprise their body. Many are unwilling to
> agree to that, in the total absence of evidence. Of course we all recall
> the "lime pit" thought experiment, in which travel is accomplished by
> transmitting the above hypothesized "backup data" and destroying the
> original. If one has enough bandwidth to record it routinely then one
> can transmit it and not worry about the bother of dealing with the actual
> carbon based matter. So, how many readers of this group would use such a
> means of transportation?
make any sense to anyone if they hadn't read some of the various sci-fi
stories about stuff like that, and there's transporter stuff from The
Fly to Star Trek. But the one that fascinates me is the one with the
"source passenger" that, if the "destination passenger" wakes up alive,
and is superficially indistinguishable from the "source passenger", like
the storybook clone, but all grown-up, scars and all, then what happens
if they wake up the "source passenger" instead of kill him/her and recycle
the carbon etc.?
I submit, what happens is fragmentation. A spirit, or consciousness, or
mind, or whatever you wish to call it, literally gets torn into two
"pieces", one of which is animating each body, albeit with only half the
energy of the whole mind. And, yes, chances are, they'd be polar opposites
like the two Kirks. And, of course, they'd push each other's buttons,
because each would be either terrified of, or enraged at, the missing part
of himself that shows up in the other, his doppelganger, or fragment,
depending on your POV.
And the same thing happenes, only weirder, when people breed and make
babies thinking they're creating new people. Sigh.
But as to why machines will never _think_, that's a whole different
sermon. ;-)
--
Cheers!
Rich
for further information, please visit http://www.godchannel.com
Frithiof Andreas Jensen
"dan michaels" <feedbac...@yahoo.com> skrev i en meddelelse
news:1118847850....@g43g2000cwa.googlegroups.com...
>brain, and the feedback is both positive and negative in variety. How
>can you EVER make such a system stable?
Stuart Kaufmann argues very well for his theory that stability and order are
emergent properties of most complex, interconnected systems; Like genes,
brains and ecosystems. It just happens - there might be cycles but overall
the system will be stable and adaptable IF conditions are right (mostly
loosely connected)
The "Stuart Kaufmann Lite" is: "at home in the univers - the search for the
laws of self organisation and complexity", ISBN 0-19-511130-3.
Frithiof Andreas Jensen
"Robert Myers" <rmyer...@comcast.net> skrev i en meddelelse
news:6n9oa1ho94mo50nji...@4ax.com...
> On Sun, 12 Jun 2005 14:03:32 +0200, "Frithiof Andreas Jensen"
> <frithio...@diespammerdie.jensen.tdcadsl.dk> wrote:
> There has been plenty of work on self-organizing systems. I haven't
> heard anyone who's actually worked in the field say, "If only we had a
> bigger or faster computer, it might work." Maybe the someone to say
> that is out there and I can elicit that comment from someone who knows
> what he's talking about from this post.
That's because nobody really knows - like running the 4 minute mile,
everybody in authority at the time had lots of sound reason why it was
impossible until somebody went and did it. Then it was easy, lots of people
were able to do it. Maybe some crazy genious will come along with "why would
this processing structure not work", do it, and find out it works quite
nicely on the "100 Ghz quantum" processor in his mobile phone.
> The right idea often turns out to be stunningly simple once you have
> the right idea. So far, no one seems to have stumbled onto the right
> idea, and building big computers to see if they will turn themselves
> into a thinking machine by random fiddling doesn't seem like a
> promising way to go.
Weeelll - lumping lots of cells together with "wiring" connected by
evolution should not bring forth intelligence either ;-)
> Have you read, say, Richard Rhodes on the making of the atomic bomb?
> A significant amount of basic science by some very bright people went
> into designing it, and the those laying the money bets had a pretty
> good idea that it would work and why. Do you have any similar
> confidence for any scheme for artificial intelligence?
No - the point is merely that since we have not really got an understanding
of what is involved, there might be a surprise waiting. Like the mediaeval
gunsmith might not succeed in building a bomb, but he would surely succeed
in radiating his gonads off.
In the next 40 years - if moores law holds - computing capacity will be
close to infinite and cheap. It would be temptingly easy to build an
simulated world, evolve intelligence and than set is loose in cheap systems
excceding our own brains processing power with little understanding on what
the limits of the evolution on such a "creature" might be.
Or maybe not - maybe brains are alredy the most powerful "computers"
around - and that is why brains are intelligent and not computers.
Rich The Philosophizer
> Interesting. Thanks for the comment. I can understand why stability "just
> happens" as an emergent property of natural selection, since any unstable
> mutations will no doubt die off quite rapidly, and those remaining will be
> characterized by being stable.
>
> However, what would be the nature of "loose-coupling" in a system with
> 100B neurons and 100T synapses? How do we actually connect all of those
> cells together in a practical manner to attain this goal, rather than just
> saying that loose-coupling theoretically implies stability?
Two neurons happen to like each other's feel, so they decide to connect.
This is catalyzed by hormones released during emotional movement. :-)
Robert Myers
<frithio...@diespammerdie.jensen.tdcadsl.dk> wrote:
>
>"Robert Myers" <rmyer...@comcast.net> skrev i en meddelelse
>news:6n9oa1ho94mo50nji...@4ax.com...
>
>> There has been plenty of work on self-organizing systems. I haven't
>> heard anyone who's actually worked in the field say, "If only we had a
>> bigger or faster computer, it might work." Maybe the someone to say
>> that is out there and I can elicit that comment from someone who knows
>> what he's talking about from this post.
>
>That's because nobody really knows - like running the 4 minute mile,
>everybody in authority at the time had lots of sound reason why it was
>impossible until somebody went and did it. Then it was easy, lots of people
>were able to do it.
Take it from me. Running a four minute mile is not easy. And no one
just steps onto a track and runs a four minute mile without
significant preparation and training--even today.
>Maybe some crazy genious will come along with "why would
>this processing structure not work", do it, and find out it works quite
>nicely on the "100 Ghz quantum" processor in his mobile phone.
>
>> The right idea often turns out to be stunningly simple once you have
>> the right idea. So far, no one seems to have stumbled onto the right
>> idea, and building big computers to see if they will turn themselves
>> into a thinking machine by random fiddling doesn't seem like a
>> promising way to go.
>
>Weeelll - lumping lots of cells together with "wiring" connected by
>evolution should not bring forth intelligence either ;-)
>
And it doesn't. Interaction with an outside world brings forth
intelligence.
>> Have you read, say, Richard Rhodes on the making of the atomic bomb?
>> A significant amount of basic science by some very bright people went
>> into designing it, and the those laying the money bets had a pretty
>> good idea that it would work and why. Do you have any similar
>> confidence for any scheme for artificial intelligence?
>
>No - the point is merely that since we have not really got an understanding
>of what is involved, there might be a surprise waiting. Like the mediaeval
>gunsmith might not succeed in building a bomb, but he would surely succeed
>in radiating his gonads off.
>
Your comment trivializes the process of scientific and technlogical
discovery in an embarrassing way. People might just get lucky at the
horse track or in a singles bar, but not at the cutting edge of
science and technology, where insight almost always precedes
discovery. I say "almost" even though I can't offhand think of
exceptions.
>In the next 40 years - if moores law holds - computing capacity will be
>close to infinite and cheap. It would be temptingly easy to build an
>simulated world, evolve intelligence and than set is loose in cheap systems
>excceding our own brains processing power with little understanding on what
>the limits of the evolution on such a "creature" might be.
>
We're already seeing some important limitations in computers based on
silicon. Those limitations are concealed by the numbers that are most
often used to announce supercomputers in the press.
By one measure, only about 512 processors of genus Blue Brain work
together effectively in problems requiring global communication, no
matter how many processors you wire up to make a "supercomputer.".
>Or maybe not - maybe brains are alredy the most powerful "computers"
>around - and that is why brains are intelligent and not computers.
>
The whole idea of using a one-dimensional measure to characterize
computers as more or less powerful is profoundly misleading.
Blue Gene computers can do calculations that no single human or group
of humans could ever do. On the other hand, Blue Gene couldn't pass
the Turing test--not necessarily because it isn't "powerful" enough,
but because no one knows how to program it to pass the Turing test.
RM
Maynard Handley
Robert Myers <rmyer...@comcast.net> wrote:
(1)
> >
> >Weeelll - lumping lots of cells together with "wiring" connected by
> >evolution should not bring forth intelligence either ;-)
> >
> And it doesn't. Interaction with an outside world brings forth
> intelligence.
>
(2)
> >No - the point is merely that since we have not really got an understanding
> >of what is involved, there might be a surprise waiting. Like the mediaeval
> >gunsmith might not succeed in building a bomb, but he would surely succeed
> >in radiating his gonads off.
> >
> Your comment trivializes the process of scientific and technlogical
> discovery in an embarrassing way. People might just get lucky at the
> horse track or in a singles bar, but not at the cutting edge of
> science and technology, where insight almost always precedes
> discovery. I say "almost" even though I can't offhand think of
> exceptions.
Hmm, Robert, someone making comment (1) hardly seems to be in a
position, in comment (2), to talk about trivializing issues.
Why is it so hard for people to admit that
(a) we don't have a freaking clue about what generates intelligence. Not
even a hypothesis. Not even a vague guess. Therefore
(b) who knows what will come out of an experiment like the Blue Brain?
Maybe nothing, maybe something of mild interest to CS people, maybe
something of real interest to philosophers. Would people have guessed,
without doing the experiments, that stirring the appropriate chemicals
together and hitting them with lightning and UV for a few days would
give amino acids?
Maynard
Robert Myers
> In article <hjh6b15nd4664gm49...@4ax.com>,
> Robert Myers <rmyer...@comcast.net> wrote:
>
> (1)
> > >
> > >Weeelll - lumping lots of cells together with "wiring" connected by
> > >evolution should not bring forth intelligence either ;-)
> > >
> > And it doesn't. Interaction with an outside world brings forth
> > intelligence.
> >
>
> (2)
> > >No - the point is merely that since we have not really got an understanding
> > >of what is involved, there might be a surprise waiting. Like the mediaeval
> > >gunsmith might not succeed in building a bomb, but he would surely succeed
> > >in radiating his gonads off.
> > >
> > Your comment trivializes the process of scientific and technlogical
> > discovery in an embarrassing way. People might just get lucky at the
> > horse track or in a singles bar, but not at the cutting edge of
> > science and technology, where insight almost always precedes
> > discovery. I say "almost" even though I can't offhand think of
> > exceptions.
>
> Hmm, Robert, someone making comment (1) hardly seems to be in a
> position, in comment (2), to talk about trivializing issues.
>
> Why is it so hard for people to admit that
> (a) we don't have a freaking clue about what generates intelligence. Not
> even a hypothesis. Not even a vague guess.
We do have some important clues. The evidence that sensory deprivation
produces congitive impairment related to the sensory deprivation is
substantial.
That means that _all_ intelligence is developed in response to an
external environment? Maybe not, but if you're looking for something
to pass the Turing test, I'll put my money safely on learning from
actors in an external environment who themselves know how to pass the
Turing test.
As it is, I wouldn't know how to test "intelligence" other than by
challenging it with external stimuli to see how it responds. To get an
appropriate response, you either have to
1. Program it, or
2. Let it learn adaptively.
Do genetically-hardwired (programmed) behaviors qualify as
intelligence?
> (b) who knows what will come out of an experiment like the Blue Brain?
Maybe some clues about how neurons interact. The hype about artificial
intelligence is just that: hype.
> Maybe nothing, maybe something of mild interest to CS people, maybe
> something of real interest to philosophers. Would people have guessed,
> without doing the experiments, that stirring the appropriate chemicals
> together and hitting them with lightning and UV for a few days would
> give amino acids?
>
It would be interesting to know what the grant proposal looked like. I
suspect it was a little more than: we're going to put some chemicals
into a reactor and hit them with UV and electrical discharges to see
what happens.
RM
Guy Macon
Maynard Handley wrote:
>Why is it so hard for people to admit that
>(a) we don't have a freaking clue about what generates intelligence. Not
>even a hypothesis. Not even a vague guess. Therefore
>(b) who knows what will come out of an experiment like the Blue Brain?
>Maybe nothing, maybe something of mild interest to CS people, maybe
>something of real interest to philosophers. Would people have guessed,
>without doing the experiments, that stirring the appropriate chemicals
>together and hitting them with lightning and UV for a few days would
>give amino acids?
The assertion "machines cannot be intelligent" is a statement of
religion, not science. The reason that they "know" that the Blue
Brain project won't give us an intelligent machine is because it
is an article of faith to them.
Robert Myers
>
> The assertion "machines cannot be intelligent" is a statement of
> religion, not science.
Who in this thread ever made such a statement? In any case, whatever I
may think of the prospects for machine intelligence, it has nothing to
do with religion.
> The reason that they "know" that the Blue
> Brain project won't give us an intelligent machine is because it
> is an article of faith to them.
Whether Blue Brain comes up with anything resembling intelligence will
have to depend on how intelligence is defined. As it is, I think it
highly unlikely that Blue Brain will exhibit any kind of behavior that
justfies the hype about intelligence.
Blue Brain may well help neuroanatomists to understand what's really
happening in the brain. If they *really* simulated the development and
interaction of neurons at the molecular level, that would be exciting.
No need to pump up the hype about intelligence and consciousness. The
basic science is plenty exciting enough. Apparently, good basic
science doesn't make press releases with enough zing.
RM
Del Cecchi
"Robert Myers" <rbmye...@gmail.com> wrote in message
news:1119111622....@f14g2000cwb.googlegroups.com...
snip
>
> It would be interesting to know what the grant proposal looked like. I
> suspect it was a little more than: we're going to put some chemicals
> into a reactor and hit them with UV and electrical discharges to see
> what happens.
>
> RM
There was a hypothesis about the composition of the atmosphere back in
the day, so to speak. The guys involved put a version of that atmosphere
in a chamber like a bell jar and zapped away with electricity and uv to
simulate lightning and sunshine. Ran an analysis of the contents later.
I doubt that there was a grant proposal involved.
del cecchi
>
Joe Legris
>
> Colonel Forbin wrote:
>
>>In article <1119034498....@g47g2000cwa.googlegroups.com>,
>>dan michaels <feedbac...@yahoo.com> wrote:
>>
>>>However, what would be the nature of "loose-coupling" in a system with
>>>100B neurons and 100T synapses? How do we actually connect all of those
>>>cells together in a practical manner to attain this goal, rather than
>>>just saying that loose-coupling theoretically implies stability?
>>
>>neurons are equivalent. Furthermore, there are a large variety of
>>alternate signaling paths besides synaptic transmission. There
>>are negative feedback loops everywhere. Neurons aren't simple
>>binary switching elements. They are complex analog devices.
>
>
>
> Yes, I know all this general backgnd info. Lots of different pathways,
> which if anything, only makes the problem even more difficult.
>
> Doesn't really help much in trying to understand how to build stability
> into such a massively complex system - esp when we have trouble
> stabilizing even a single FB loop in engineering. I guess maybe the
> answer is ... if you greatly overdamp the system using massive amounts
> of negative-FB, then it will generally be stable.
>
> OTOH, massive overdamping tends to slow down the responsivity greatly,
> and this would be counter-productive regarding survival
> of the organism. Let's take a running man. His FB-loops are pretty
> finely tuned, and probably not over-damped. We aren't sloths. We can
> react quickly to terrain changes, etc. Let's take a guy driving a car,
> and a child runs out in front. The guy reacts pretty fast, given all of
> the processing which goes on to produce an avoidance response.
> Overdamping due to massive negative-FB is probably counter-productive
> to survival. It doesn't seem likely to be the case.
>
Consider the action of the cerebellum; it employs tuned *feedforward*
networks to achieve the necessary motor response speed. Negative
feedback is used over the long-term (off-line, so to speak) to tune the
network, but the instantaneous response is open loop.
--
Joe Legris
Glen M. Sizemore
"dan michaels" <feedbac...@yahoo.com> wrote in message
news:1119237640.4...@z14g2000cwz.googlegroups.com...
> Hi Joe. If I read this correctly, your suggestion is that the
> "initial" response is open-loop or otherwise uncontrolled, but then
> "after" a bit of time, the negative feedback processes come into play,
> and stabilize the action. This could conceivably give both rapid
> responses and also overall longer-term stability. This is good, I
> think! There are a lot of variations you could play on this scheme. It
> could be refined by having the initial O-L responses being more
> localized and the later negative-FB signals more global-acting. IOW, a
> strong local action in a area of neural tissue could initiate
> inhibition of similar action from taking place in near temporal
> proximity in nearby tissue. Etc. Etc. Hmmm. This scheme bears some
> resemblance to the concept of refractory periods in action-potential
> formation. IOW, the refractory period imposes a natural temporality and
> directionality on the flow activity emanating from an individual
> neuron. Likewise, the scheme
> above could impose something similar on the larger scale of activity in
> neural nets. The longer, slower-acting negative-FB in effect makes the
> wider area of tissue refractory.
>
Glen M. Sizemore
systems at equilibrium, but they are not damped (sort of) with respect to
events that produce rapid changes (A-process). These events initially
produce a slow-onset, somewhat long acting opposing process (B-process) that
pushes toward the old equilibrium level. This opposing process eventually
becomes more rapid in its onset and increases in duration. But the main part
is that its equilibrium state is caused by different processes than what
goes on when something purturbs the state. I thought it seemed relevant to
the issue of highly damped systems showing slow response.
"dan michaels" <feedbac...@yahoo.com> wrote in message
news:1119282358.1...@g43g2000cwa.googlegroups.com...
>
>
>
> > Opponent process theory
> >
>
>
> Green vs red cones, also forward excitation and inhibition, also
> lateral excitation and inhibition, also positive and negative-FB, also
> area one interacting reciprocally with area 2, etc, on and on, are all
> opponent processes. But as already discussed, such facts alone don't
> solve the problem regards building stable working mechanisms. The
> solution is in how you connect them up, spatially and temporally,
> locally and globally, and whatever else.
Joe Legris
> "initial" response is open-loop or otherwise uncontrolled, but then
> "after" a bit of time, the negative feedback processes come into play,
> and stabilize the action. This could conceivably give both rapid
> responses and also overall longer-term stability. This is good, I
> think! There are a lot of variations you could play on this scheme. It
> could be refined by having the initial O-L responses being more
> localized and the later negative-FB signals more global-acting. IOW, a
> strong local action in a area of neural tissue could initiate
> inhibition of similar action from taking place in near temporal
> proximity in nearby tissue. Etc. Etc. Hmmm. This scheme bears some
> resemblance to the concept of refractory periods in action-potential
> formation. IOW, the refractory period imposes a natural temporality and
> directionality on the flow activity emanating from an individual
> neuron. Likewise, the scheme
> above could impose something similar on the larger scale of activity in
> neural nets. The longer, slower-acting negative-FB in effect makes the
> wider area of tissue refractory.
>
Experiments have demonstrated the stabilizing influence of the
cerebellum over motor control. Monkeys trained to resist the movements
of a motorized handle did so quite effectively and with little
overshoot. When the outputs from the cerebellum were temporarily
deactivated (through localized cooling), pronounced overshoot and and
oscillation occurred - the system became unstable.
The interesting result is that such damping is *not* the result of
real-time negative feedback. Muscles have so-called stretch receptors
that are activated when a muscle is placed under a tensile load. When
the monkey's biceps are initially stretched by the handle, negative
feedback via the stretch-receptors allows him to resist the motion, but
the fine-tuning and damping of that response is open-loop. The triceps
act antagonistically to dampen the biceps response but the triceps are
activated *before* they are stretched - they actvin open-loop
feed-forward mode, controlled by the cerebellum. Fine motor control
depends largely the conditions that existed when the response was
trained. Of course, training never really ends, so if the dynamic
properties of the task change, for example the force of the motorized
handle is increased, instability will occur until the new "damping
factor" is trained.
--
Joe Legris
Frithiof Andreas Jensen
"Guy Macon" <_see.web.page_@_www.guymacon.com_> skrev i en meddelelse
news:11ap166...@corp.supernews.com...
>
>
>
> SioL wrote:
>>
>>Wouldn't it be cool to have AI in a AVR, you'd just connect a mic
>>and tell it:
>>
>>Howdy there partner, would you flash that LED once a second for me?
>
> Then you would have to listen to it complain about how all the diodes
> on it's left have begun to ache...
Stanislav Lem has some interesting "robot stories" along that line - like a
guy needs legal advice, so he build a robot lawyer which immediately
requires more hardware to even look at his problem; which of course the
"lawyer" then cannot solve - but the lawyer was clever enough to make the
designer sign a leagally binding document promising to keep the lawyer
operational after learning the answer to his problem ;-)
Albert van der Horst
Rich The Philosophizer <r...@example.net> wrote:
<SNIP>
>
>I submit, what happens is fragmentation. A spirit, or consciousness, or
>mind, or whatever you wish to call it, literally gets torn into two
>"pieces", one of which is animating each body, albeit with only half the
>energy of the whole mind. And, yes, chances are, they'd be polar opposites
Okay. So the only thing missing is the pope declaring that splitting
a soul two-ways is as mortal a sin as using a condom.
What you say implies an inherently methaphysical world view,
shared by almost nobody in this group, certainly not me.
>--
>Cheers!
>Rich
Groetjes Albert
--
--
Albert van der Horst,Oranjestr 8,3511 RA UTRECHT,THE NETHERLANDS
Economic growth -- like all pyramid schemes -- ultimately falters.
alb...@spenarnc.xs4all.nl http://home.hccnet.nl/a.w.m.van.der.horst
Curt Welch
> However, my question regards stability really has more to do with how
> do you connect together billions of interacting elements in various
> neural nets, and still make the operation stable. You have to balance
> off influences against each other. Too much excitation and the net [and
> resulting output behavior] erupts, too much inhibition and net activity
> dies off [and the organism/behavior goes torpid]. Coming up with a
> system with the right mix, I think, is not a trivial problem. And as
> mentioned earlier, I think you clearly need stabilizing mechanisms at
> "every" level of nervous activity for the best result. Also, I am
> thinking of all brains, not just the highly-evolved motor systems of
> primates/etc, where a high degree of learning and practice takes place.
There are many simple solutions to that problem. In the networks I've
created and played with, I've had to solve that problem many times.
To start with, if you solve it at the lowest levels, then you have no
network level problems. For example, if each node simply trains itself to
maintain an average activity level by adjusting its behavior over time to
be more sensitive to fire when it's activity becomes to low and less
sensitive when the activity becomes to high, then each node is simply
self-regulating itself, and the network as a whole can never get out of
control.
I've also used techniques to regulate the network activity at the network
level. I simply track the total activity of the network, and then adjust
all the nodes evenly to be more, or less, sensitive to activation based on
total network activity level.
In my latest design, I eliminated the problem completely by working with
these feed-forward pulse routing networks. Each node that "fires" will
always activate one, and only one, "downstream" node, so it's impossible
for the network to over, or under, activate.
How the brain manages to regulate the nodes I have no clue, but I don't see
it as a hard problem. Any network design that has the power to amplify
activity (one node, causes 2 or more downstream nodes to fire), or damp
activity (one node fires and it doesn't cause any other nodes to fire),
must have something in place to regulate activity. But it can be done by
individual nodes regulating their own activity, or it can be done at the
network level. The entire brain could easily have systems to monitor total
brain activity (based on consumption of whatever chemicals are needed to
supply energy to the neurons for example), and then use some chemical
controls to increase or decrease the sensitivity of all the neurons as
required.
Or, if there's an energy flow required to power the cells, it can regulate
activity by simply limiting the energy available to allow the cells to
fire.
The brain probably uses a bunch of these techniques at the same time.
--
Curt Welch http://CurtWelch.Com/
cu...@kcwc.com http://NewsReader.Com/
Joe Legris
> at a much higher level than my question about general
> stability was aimed. Up at the organism-interacting-with-environment
> level, motor control of arm movements/etc, the system definitely needs
> all manner of feedforward/feedback control, anonist-antagonist units,
> reflex arcs, etc, to control the movements. What you describe with the
> triceps sounds like an example of using what is, in essence,
> feedforward "prediction" overlaid atop the background reflex and
> control processes.
>
> However, my question regards stability really has more to do with how
> do you connect together billions of interacting elements in various
> neural nets, and still make the operation stable. You have to balance
> off influences against each other. Too much excitation and the net [and
> resulting output behavior] erupts, too much inhibition and net activity
> dies off [and the organism/behavior goes torpid]. Coming up with a
> system with the right mix, I think, is not a trivial problem. And as
> mentioned earlier, I think you clearly need stabilizing mechanisms at
> "every" level of nervous activity for the best result. Also, I am
> thinking of all brains, not just the highly-evolved motor systems of
> primates/etc, where a high degree of learning and practice takes place.
>
We do know that some things are conserved across vertebrate species:
genetic control of neural development and the subsequent thickening and
thinning of axonal arborizations and neural connections depending on
experience during youthful critical periods.
Organisms with a genetic predisposition toward unstable nervous systems
would have been weeded out long ago. Also, the neural refinement that
occurs during critical periods is sensitive to *synchronous* neural
activity, which is generally the result of sensory stimulation - the
organism becomes tuned to the order inherent in its environment.
Interconnected neurons that fire synchronously tend to grow bushier,
forming more presynaptic interconnections while those that fire
asynchronously tend to wither. I am guessing that excitation due to
local instability would tend to be disorganized, leading to withering of
the participating presynaptic terminals and a correction towards stability.
--
Joe Legris
Stephen Fuld
snip
> The real NS sounds a little more complex
> that your current systems. Each cell has 5,000-10,000 inputs and sends
> outputs to several 1000 others.
I have seen numbers like this before, and I never understood something about
them. IT implies that the total number of inputs is greater than the total
number of outputs. Since each "connection" involves one output to one
input, how can total number of inputs exceed the number of outputs? Of
course each neuron can have quite varying and different numbers of inputs
and outputs, but if the total number has to be equal, doesn't the average
have to be equal?
--
- Stephen Fuld
e-mail address disguised to prevent spam
Curt Welch
Yeah good point, if you talk about the average number of interconnects, you
shouldn't try to talk about the inputs being different than the outputs.
It just doesn't make sense to do that.
But, if the neurons which tend to have large numbers of inputs aren't the
ones with large numbers of outputs, then talking like that might be valid.
I don't know enough about the brain to know if that might be true even in a
limited case.
Maynard Handley
I just came across this very interesting press release from Caltech a
few days ago. The important paragraphs are
Now a research team of neuroscientists from the California Institute of
Technology and UCLA has found that a single neuron can recognize people,
landmarks, and objects--even letter strings of names
("H-A-L-L-E-B-E-R-R-Y"). The findings, reported in the current issue of
the journal Nature, suggest that a consistent, sparse, and explicit code
may play a role in transforming complex visual representations into
long-term and more abstract memories.
and
"Our findings fly in the face of conventional thinking about how brain
cells function," adds Christof Koch, the Lois and Victor Troendle
Professor of Cognitive and Behavioral Biology and professor of
computation and neural systems at Caltech, and the other co-senior
investigator. "Conventional wisdom views individual brain cells as
simple switches or relays. In fact, we are finding that neurons are able
to function more like a sophisticated computer."
IBM claims
http://domino.research.ibm.com/comm/pr.nsf/pages/news.20050606_CognitiveI
ntelligence.html
that their simulation is not a neural net but at the cellular/molecular
level. Good luck to them, but it seems awfully optimistic what they are
doing given the incredibly unexpected stuff we're still learning.
Maynard
Curt Welch
> The real NS sounds a little more complex
> that your current systems. Each cell has 5,000-10,000 inputs and sends
> outputs to several 1000 others.
I don't see that as an issue because I think that anything you do with 10,
000 inputs can also be done with two inputs and more nodes. I think it's
just a tradeoff between number of nodes vs number of inputs. I think with
the hardware we have to work with it will always be easier to use more
nodes with less connections.
> And feedback, both positive and
> negative, is everywheres present.
Yeah, the feeback question is one I have not yet realy explored. It's
needed for sure and I have various ideas about how to best build it into
this type of network, but I've not explored those ideas yet (well except
for some simple experiments).
Rich Grise
> "dan michaels" <feedbac...@yahoo.com> wrote:
>> Curt Welch wrote:
>> > In my latest design, I eliminated the problem completely by working
>> > with these feed-forward pulse routing networks. Each node that "fires"
>> > will always activate one, and only one, "downstream" node, so it's
>> > impossible for the network to over, or under, activate.
>
>> The real NS sounds a little more complex
>> that your current systems. Each cell has 5,000-10,000 inputs and sends
>> outputs to several 1000 others.
>
> I don't see that as an issue because I think that anything you do with 10,
> 000 inputs can also be done with two inputs and more nodes. I think it's
> just a tradeoff between number of nodes vs number of inputs. I think with
> the hardware we have to work with it will always be easier to use more
> nodes with less connections.
>
>> And feedback, both positive and
>> negative, is everywheres present.
>
> Yeah, the feeback question is one I have not yet realy explored. It's
> needed for sure and I have various ideas about how to best build it into
> this type of network, but I've not explored those ideas yet (well except
> for some simple experiments).
I think that before we start analyzing a human brain, we should work on
how an ameba knows what's an enemy and what's food.
Cheers!
Rich
Joe Legris
Exactly. What people REALLY want comes down to what their cells want:
ventilation, nourishment, a productive occupation, a community and a
chance to reproduce - motivation and intelligence are utterly
distributed right down to the bottom level. Networks of processing
elements have no need for and ultimately no capacity for intelligence. A
network element slogs on because it cannot die. A cell persists because
it must live.
--
Joe Legris
Stephen Fuld
> than the #cells it receives from.
OK, but if the number of outputs is also synapses, not neurons, then we are
in the same spot.
> And each output axon bifucates many
> times to form synapses.
Yes, of course.
> It all works out.
Not unless you count inputs as synapses and outputs as neuron's connected
to, independent of the number of synapses between a single pair of neurons.
I guess it is possible, but it sure seems odd to count inputs and outputs
differently.
Jan Vorbrüggen
> how an ameba knows what's an enemy and what's food.
The problem with that approach is that it won't help you understand the
human, or a mammalian, or even just a vertebrate brain one iota. Insects,
for instance, have a totally different architecture (in the sense of
"computer architecture") of their brains that mammalian (which includes
human) brains. The least complex organism to learn something about our-
selves is a mouse.
Jan
Jan Vorbrüggen
> Technology and UCLA has found that a single neuron can recognize people,
> landmarks, and objects--even letter strings of names
> ("H-A-L-L-E-B-E-R-R-Y"). The findings, reported in the current issue of
> the journal Nature, suggest that a consistent, sparse, and explicit code
> may play a role in transforming complex visual representations into
> long-term and more abstract memories.
"...a single neuron can recognize people, landmarks, and objects" is
rubbish. You can use a single neuron as a representative of an ensemble
that does all the work and that signals recognition of some specific
objects, yes - but that much has been known for at least fifteen years.
> "Our findings fly in the face of conventional thinking about how brain
> cells function," adds Christof Koch, the Lois and Victor Troendle
> Professor of Cognitive and Behavioral Biology and professor of
> computation and neural systems at Caltech, ...
Certain people from MIT, Caltech and some other places have too much public
visibility for science's, and their own, good.
Jan
Robert Myers
> http://pr.caltech.edu/media/Press_Releases/PR12710.htmlJoe Legris
>
> I just came across this very interesting press release from Caltech a
> few days ago. The important paragraphs are
>
> Now a research team of neuroscientists from the California Institute of
> Technology and UCLA has found that a single neuron can recognize people,
> landmarks, and objects--even letter strings of names
> ("H-A-L-L-E-B-E-R-R-Y"). The findings, reported in the current issue of
> the journal Nature, suggest that a consistent, sparse, and explicit code
> may play a role in transforming complex visual representations into
> long-term and more abstract memories.
>
Sounds like a reincarnation of the grandmother cell to me.
RM