Fwd: [tt] Darpa’s Simulated Cat Brain Project a ‘Scam’: Top Scientist
7 views
Skip to first unread message
Bryan Bishop
Nov 24, 2009, 11:08:18 AM11/24/09
to diybio, diytrans...@googlegroups.com, ExI chat list, World Transhumanist Association Discussion List, kan...@gmail.com
(the spelling is Markram, IIRC.)
---------- Forwarded message ----------
From: Eugen Leitl <eu...@leitl.org>
Date: Tue, Nov 24, 2009 at 5:00 AM
Subject: [tt] Darpa’s Simulated Cat Brain Project a ‘Scam’: Top Scientist
To: t...@postbiota.org, in...@postbiota.org, ne...@postbiota.org
Darpa’s Simulated Cat Brain Project a ‘Scam’: Top Scientist
By Noah Shachtman * November 23, 2009 | * 11:49 am | * Categories:
DarpaWatch
Last year, the Pentagon’s premiere research arm gave IBM nearly $5 million to
make electronics that mimic the “function, size and power consumption” of a
cat’s brain. Last week, IBM’s lead researcher on the project, Dharmendra
Modha, announced that he had made major progress toward that goal, simulating
on a supercomputer the number of neurons and synapses inside a feline mind.
Now, a leading neuroscientist is blasting the whole project as a “scam” and a
“hoax.”
Modha told a supercomputing conference that his cortical simulator had
generated the digital equivalent of a billion neurons connected by 10
trillion individual synapses. It was, apparently, the first baby step toward
Pentagon mad science division Darpa’s goal of re-creating a brain that’s as
compact, as efficient and as power-smart as the one inside a house pet. Some
colleagues went so far as to compare it to the Large Hadron Collider.
But neuroscientist Henry Markham is considerably less impressed. “What IBM
reported is a scam — no where near a cat-scale brain simulation,” he writes
in an open letter to Bernard Myerson, IBM’s Chief Technology Officer. “I am
absolutely shocked at this announcement. Not because it is any kind of
technical feat, but because of the mass deception of the public.”
Markham isn’t exactly a disinterested observer, as IEEE Spectrum’s Sally Adee
notes. He’s got his own ersatz mind project, called Blue Brain, that’s also
affiliated with IBM. So perhaps it’s not surprising that Markham claims Modha
has simply put together a “PR stunt here to ride on Blue Brain.”
Still, such public criticism is unusual — especially when you consider that
it’s also an indirect indictment of Darpa, one of the leading funders of
artificial intelligence research.
"All these kinds of simulations are trivial and have been around for
decades — simply called artificial neural network (ANN) simulations. We even
stooped to doing these kinds of simulations as benchmark tests four years ago
with 10’s of millions of such points…. If we (or anyone else) wanted to we
could easily do this for a billion “points,” but we would certainly not call
it a cat-scale simulation. It is really no big deal to simulate a billion
points interacting if you have a big enough computer. The only step here is
that they have at their disposal a big computer. For a grown-up “researcher”
to get excited because one can simulate billions of points interacting is
ludicrous.
…This is light years away from a cat brain, not even close to an ant’s
brain in complexity. It is highly unethical of Mohda to mislead the public in
making people believe they have actually simulated a cat’s brain…. That IBM
and DARPA would support such deceptive announcements is even more shocking."
I think the proper response here is: mee-yow! I can’t wait for round three of
this cat-brain cat fight.
_______________________________________________
tt mailing list
t...@postbiota.org
http://postbiota.org/mailman/listinfo/tt
--
- Bryan
http://heybryan.org/
1 512 203 0507
---------- Forwarded message ----------
From: Eugen Leitl <eu...@leitl.org>
Date: Tue, Nov 24, 2009 at 5:00 AM
Subject: [tt] Darpa’s Simulated Cat Brain Project a ‘Scam’: Top Scientist
To: t...@postbiota.org, in...@postbiota.org, ne...@postbiota.org
Darpa’s Simulated Cat Brain Project a ‘Scam’: Top Scientist
By Noah Shachtman * November 23, 2009 | * 11:49 am | * Categories:
DarpaWatch
Last year, the Pentagon’s premiere research arm gave IBM nearly $5 million to
make electronics that mimic the “function, size and power consumption” of a
cat’s brain. Last week, IBM’s lead researcher on the project, Dharmendra
Modha, announced that he had made major progress toward that goal, simulating
on a supercomputer the number of neurons and synapses inside a feline mind.
Now, a leading neuroscientist is blasting the whole project as a “scam” and a
“hoax.”
Modha told a supercomputing conference that his cortical simulator had
generated the digital equivalent of a billion neurons connected by 10
trillion individual synapses. It was, apparently, the first baby step toward
Pentagon mad science division Darpa’s goal of re-creating a brain that’s as
compact, as efficient and as power-smart as the one inside a house pet. Some
colleagues went so far as to compare it to the Large Hadron Collider.
But neuroscientist Henry Markham is considerably less impressed. “What IBM
reported is a scam — no where near a cat-scale brain simulation,” he writes
in an open letter to Bernard Myerson, IBM’s Chief Technology Officer. “I am
absolutely shocked at this announcement. Not because it is any kind of
technical feat, but because of the mass deception of the public.”
Markham isn’t exactly a disinterested observer, as IEEE Spectrum’s Sally Adee
notes. He’s got his own ersatz mind project, called Blue Brain, that’s also
affiliated with IBM. So perhaps it’s not surprising that Markham claims Modha
has simply put together a “PR stunt here to ride on Blue Brain.”
Still, such public criticism is unusual — especially when you consider that
it’s also an indirect indictment of Darpa, one of the leading funders of
artificial intelligence research.
"All these kinds of simulations are trivial and have been around for
decades — simply called artificial neural network (ANN) simulations. We even
stooped to doing these kinds of simulations as benchmark tests four years ago
with 10’s of millions of such points…. If we (or anyone else) wanted to we
could easily do this for a billion “points,” but we would certainly not call
it a cat-scale simulation. It is really no big deal to simulate a billion
points interacting if you have a big enough computer. The only step here is
that they have at their disposal a big computer. For a grown-up “researcher”
to get excited because one can simulate billions of points interacting is
ludicrous.
…This is light years away from a cat brain, not even close to an ant’s
brain in complexity. It is highly unethical of Mohda to mislead the public in
making people believe they have actually simulated a cat’s brain…. That IBM
and DARPA would support such deceptive announcements is even more shocking."
I think the proper response here is: mee-yow! I can’t wait for round three of
this cat-brain cat fight.
_______________________________________________
tt mailing list
t...@postbiota.org
http://postbiota.org/mailman/listinfo/tt
--
- Bryan
http://heybryan.org/
1 512 203 0507
Daniel C.
Nov 24, 2009, 12:02:05 PM11/24/09
to diy...@googlegroups.com
On Tue, Nov 24, 2009 at 9:08 AM, Bryan Bishop <kan...@gmail.com> wrote:
> I think the proper response here is: mee-yow! I can’t wait for round three of
> this cat-brain cat fight.
Next up: hair pulling and playground-name calling.
> I think the proper response here is: mee-yow! I can’t wait for round three of
> this cat-brain cat fight.
Does an ant's brain really have a billion neurons in it? Or is the
"point":neuron correlation not 1:1?
-Dan
Bryan Bishop
Nov 24, 2009, 12:06:58 PM11/24/09
to diy...@googlegroups.com, kan...@gmail.com
http://heybryan.org/books/Biology/neuro/neuron_counts.txt
whale: 200,000,000,000 neurons
human brain: 100,000,000,000 neurons
chimpanzee: 6,200,000,000 neurons
cat: 300,000,000 neurons
octopus: 300,000,000 neurons
elephant: 200,000,000 neurons
dog: 160,000,000 neurons
frog: 16,000,000 neurons
rat: 15,000,000 neurons
mouse: 4,000,000 neurons
cockroach: 1,000,000 neurons
drosophila melanogaster: 100,000 neurons
honey bee: 850,000 neurons
pond snail: 11,000 neurons
ant: 10,000 neurons (100,000 neurons in some species)
sea slug: 7,000 neurons
caenorhabditis elegans: 302 neurons
What Markram was saying was that an ant has the same level of
complexity. Markram's research is centered around the idea that the
brain is not simply a giant ANN, that there is actual structure that
matters. Apparently the ant brain doesn't have much structure to it
other than layers and layers of neurons. But no, it doesn't have a
billion neurons.
Eugen Leitl
Nov 24, 2009, 12:20:43 PM11/24/09
to diy...@googlegroups.com
On Tue, Nov 24, 2009 at 10:02:05AM -0700, Daniel C. wrote:
> Next up: hair pulling and playground-name calling.
Blame that Schachtner dude.
> Next up: hair pulling and playground-name calling.
> Does an ant's brain really have a billion neurons in it? Or is the
> "point":neuron correlation not 1:1?
--
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE
Cathal Garvey
Nov 25, 2009, 5:15:46 AM11/25/09
to diy...@googlegroups.com
Neurons are far from "points".
Although a well-designed Neural Network with a well designed *context* can be an excellent simulation of life (see the Creatures series of games for a good, fun-to-try example of this), having a floating cloud of neurons without predetermined structure or context is useless.
Even then, you can't simply imprint the "shape" of a cat's brain on your neurons and expect much, because in nature there are a host of different neuron types, which support a variety of different communications and exchanges, each of which "learns" from its own history and contributes this "memory" to its behaviour toward other cells. If you want to mimic a brain as it's found in nature, "points" won't cut it.
However, that's not to say that, with the proper architecture, you couldn't use a billions-scale set of "points and synapses" to create an intelligent system. It's just that biomimetics is off the table, because you don't have the complexity you need to mimic nature.
Although a well-designed Neural Network with a well designed *context* can be an excellent simulation of life (see the Creatures series of games for a good, fun-to-try example of this), having a floating cloud of neurons without predetermined structure or context is useless.
Even then, you can't simply imprint the "shape" of a cat's brain on your neurons and expect much, because in nature there are a host of different neuron types, which support a variety of different communications and exchanges, each of which "learns" from its own history and contributes this "memory" to its behaviour toward other cells. If you want to mimic a brain as it's found in nature, "points" won't cut it.
However, that's not to say that, with the proper architecture, you couldn't use a billions-scale set of "points and synapses" to create an intelligent system. It's just that biomimetics is off the table, because you don't have the complexity you need to mimic nature.
Eugen Leitl
Nov 25, 2009, 3:42:53 PM11/25/09
to diy...@googlegroups.com
On Wed, Nov 25, 2009 at 10:15:46AM +0000, Cathal Garvey wrote:
> Neurons are far from "points".
Indeed. Run, don't walk to purchase
> Neurons are far from "points".
http://www.amazon.com/Biophysics-Computation-Information-Computational-Neuroscience/dp/0195181999/
and
http://www.amazon.com/Methods-Neuronal-Modeling-Computational-Neuroscience/dp/0262112310/
> Although a well-designed Neural Network with a well designed *context* can
> be an excellent simulation of life (see the Creatures series of games for a
> good, fun-to-try example of this), having a floating cloud of neurons
> without predetermined structure or context is useless.
>
> Even then, you can't simply imprint the "shape" of a cat's brain on your
> neurons and expect much, because in nature there are a host of different
> neuron types, which support a variety of different communications and
> exchanges, each of which "learns" from its own history and contributes this
> "memory" to its behaviour toward other cells. If you want to mimic a brain
> as it's found in nature, "points" won't cut it.
http://www.philosophy.ox.ac.uk/__data/assets/pdf_file/0019/3853/brain-emulation-roadmap-report.pdf
> However, that's not to say that, with the proper architecture, you couldn't
> use a billions-scale set of "points and synapses" to create an intelligent
> system. It's just that biomimetics is off the table, because you don't have
> the complexity you need to mimic nature.
Cathal Garvey
Nov 26, 2009, 7:00:24 AM11/26/09
to diy...@googlegroups.com
AI and whole brain emulation are different problems.
Exactly my point!
Ich bin Singularitarian
Nov 29, 2009, 2:04:39 AM11/29/09
to DIYbio
Here's a good link:
http://www.youtube.com/watch?v=1QPiF4-iu6g&feature=channel
World's first cyborg :))
Also, I dont think you would need to simulate all the neurons of the
brain to achieve its functionality.
Almost more than half would be dedicated for survival tactics which
may not be needed at all in the simulated brain (atleast the first
version)
http://www.youtube.com/watch?v=1QPiF4-iu6g&feature=channel
World's first cyborg :))
Also, I dont think you would need to simulate all the neurons of the
brain to achieve its functionality.
Almost more than half would be dedicated for survival tactics which
may not be needed at all in the simulated brain (atleast the first
version)
Cathal Garvey
Dec 1, 2009, 7:19:37 AM12/1/09
to diy...@googlegroups.com
Some would argue that "Survival Tactics" underpin almost all recognisably intelligent functions of an animal.
Humans and other higher-order sentient animals are unusual in bothering with anything above the necessities of food, water and sleep. that's not to say that animals aren't all sentient to some degree, but their thoughts, emotions and drives are shaped by their experiences, and those experiences are primarily about survival.
We're only slightly different. Without the "survival tactic" of demanding food, warmth and protection, babies mightn't bond as well to their parents as they need to to become balanced and functioning people. Before they ever get to the "interesting stuff" beyond the basics of surviving, they may stray away from reasonable thought because of a deficiency in their basic urges.
I don't believe we're at the point where we can pick-and-choose what bits of a brain to emulate. Emulate the whole thing, and it gives you a chance to better understand nature's finest invention. Hold off on the hubris until you know you know better.
Humans and other higher-order sentient animals are unusual in bothering with anything above the necessities of food, water and sleep. that's not to say that animals aren't all sentient to some degree, but their thoughts, emotions and drives are shaped by their experiences, and those experiences are primarily about survival.
We're only slightly different. Without the "survival tactic" of demanding food, warmth and protection, babies mightn't bond as well to their parents as they need to to become balanced and functioning people. Before they ever get to the "interesting stuff" beyond the basics of surviving, they may stray away from reasonable thought because of a deficiency in their basic urges.
I don't believe we're at the point where we can pick-and-choose what bits of a brain to emulate. Emulate the whole thing, and it gives you a chance to better understand nature's finest invention. Hold off on the hubris until you know you know better.
Eugen Leitl
Dec 1, 2009, 7:26:45 AM12/1/09
to diy...@googlegroups.com
On Tue, Dec 01, 2009 at 12:19:37PM +0000, Cathal Garvey wrote:
> I don't believe we're at the point where we can pick-and-choose what bits of
> a brain to emulate. Emulate the whole thing, and it gives you a chance to
> better understand nature's finest invention. Hold off on the hubris until
> you know you know better.
You cannot omit parts when you do a simulation from first principles.
> I don't believe we're at the point where we can pick-and-choose what bits of
> a brain to emulate. Emulate the whole thing, and it gives you a chance to
> better understand nature's finest invention. Hold off on the hubris until
> you know you know better.
Moreover, you don't want a generic cat, you want a specific cat.
The only way to get there is by ablative scanner, and plenty of time.
(Current scanners could do a mouse in a few years).
Cathal Garvey
Dec 1, 2009, 7:48:56 AM12/1/09
to diy...@googlegroups.com
If you could do an ablative full-brain scan and simulate the results, then it would give you a physical model of the brain's structure, but it won't give you results on its own.
Take a rat, teach it recognisable tricks and thought patterns, then euthanise in such a way as to perfectly preserve all brain structures in a replicable (if not resuscitate-able) form..perhaps anaesthesia, saline replacement of blood with sulphur dioxide to preserve brain tissues and some kind of anti-freeze, followed by careful freezing. Microtome sections could be scanned with exceptional clarity and a computer script could identify cell innards and "stack" the sections to come up with a map of the brain.
Once you have structure, the rest of your work (arguably the hardest) is to "annotate" all the cell forms you've found and identified, assign variables that determine their behaviour, and hope you know enough to bootstrap anything resembling a working brain. Then you can try to elicit the trained responses you taught the rat, to see if the structure and chemistry model are up to scratch.
It's the last step they're probably on; MRI, CAT and Tomography scans have probably given as much information as you'd get from a sectioned brain already, but the variables needed to prompt a meaningful result from the physical brain are practically impossible to determine in their native state with our current technology. The best results I've heard of are with transgenic mice whose cells release a measurable signal when their neurons fire or whatever, which is a good first step. Even with technology like this to help elucidate the natural state of the brain, much will remain guesswork.
It's gonna take a long time I think.
--
letters.cunningprojects.com
twitter.com/onetruecathal
Take a rat, teach it recognisable tricks and thought patterns, then euthanise in such a way as to perfectly preserve all brain structures in a replicable (if not resuscitate-able) form..perhaps anaesthesia, saline replacement of blood with sulphur dioxide to preserve brain tissues and some kind of anti-freeze, followed by careful freezing. Microtome sections could be scanned with exceptional clarity and a computer script could identify cell innards and "stack" the sections to come up with a map of the brain.
Once you have structure, the rest of your work (arguably the hardest) is to "annotate" all the cell forms you've found and identified, assign variables that determine their behaviour, and hope you know enough to bootstrap anything resembling a working brain. Then you can try to elicit the trained responses you taught the rat, to see if the structure and chemistry model are up to scratch.
It's the last step they're probably on; MRI, CAT and Tomography scans have probably given as much information as you'd get from a sectioned brain already, but the variables needed to prompt a meaningful result from the physical brain are practically impossible to determine in their native state with our current technology. The best results I've heard of are with transgenic mice whose cells release a measurable signal when their neurons fire or whatever, which is a good first step. Even with technology like this to help elucidate the natural state of the brain, much will remain guesswork.
It's gonna take a long time I think.
2009/12/1 Eugen Leitl <eu...@leitl.org>
--
You received this message because you are subscribed to the Google Groups "DIYbio" group.
To post to this group, send email to diy...@googlegroups.com.
To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/diybio?hl=en.
--
letters.cunningprojects.com
twitter.com/onetruecathal
Eugen Leitl
Dec 1, 2009, 8:25:03 AM12/1/09
to diy...@googlegroups.com
On Tue, Dec 01, 2009 at 12:48:56PM +0000, Cathal Garvey wrote:
> If you could do an ablative full-brain scan and simulate the results, then
> it would give you a physical model of the brain's structure, but it won't
> give you results on its own.
from that I see
> If you could do an ablative full-brain scan and simulate the results, then
> it would give you a physical model of the brain's structure, but it won't
> give you results on its own.
> Take a rat, teach it recognisable tricks and thought patterns, then
> euthanise in such a way as to perfectly preserve all brain structures in a
> replicable (if not resuscitate-able) form..perhaps anaesthesia, saline
> replacement of blood with sulphur dioxide to preserve brain tissues and some
> kind of anti-freeze, followed by careful freezing. Microtome sections could
> be scanned with exceptional clarity and a computer script could identify
> cell innards and "stack" the sections to come up with a map of the brain.
> Once you have structure, the rest of your work (arguably the hardest) is to
> "annotate" all the cell forms you've found and identified, assign variables
> that determine their behaviour, and hope you know enough to bootstrap
> anything resembling a working brain. Then you can try to elicit the trained
> responses you taught the rat, to see if the structure and chemistry model
> are up to scratch.
> It's the last step they're probably on; MRI, CAT and Tomography scans have
> probably given as much information as you'd get from a sectioned brain
> already, but the variables needed to prompt a meaningful result from the
> physical brain are practically impossible to determine in their native state
> with our current technology. The best results I've heard of are with
> transgenic mice whose cells release a measurable signal when their neurons
> fire or whatever, which is a good first step. Even with technology like this
> to help elucidate the natural state of the brain, much will remain
> guesswork.
> It's gonna take a long time I think.
>
> 2009/12/1 Eugen Leitl <eu...@leitl.org>
>
> > On Tue, Dec 01, 2009 at 12:19:37PM +0000, Cathal Garvey wrote:
> >
> > > I don't believe we're at the point where we can pick-and-choose what bits
> > of
> > > a brain to emulate. Emulate the whole thing, and it gives you a chance to
> > > better understand nature's finest invention. Hold off on the hubris until
> > > you know you know better.
> >
> > You cannot omit parts when you do a simulation from first principles.
> > Moreover, you don't want a generic cat, you want a specific cat.
> > The only way to get there is by ablative scanner, and plenty of time.
> > (Current scanners could do a mouse in a few years).
> >
> > --
> > Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
> > ______________________________________________________________
> > ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
> > 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE
> >
> > --
> >
> > You received this message because you are subscribed to the Google Groups
> > "DIYbio" group.
> > To post to this group, send email to diy...@googlegroups.com.
> > To unsubscribe from this group, send email to
> > .
> > For more options, visit this group at
> > http://groups.google.com/group/diybio?hl=en.
> >
> >
> >
>
>
> --
> letters.cunningprojects.com
> twitter.com/onetruecathal
>
> --
>
> You received this message because you are subscribed to the Google Groups "DIYbio" group.
> To post to this group, send email to diy...@googlegroups.com.
> To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/diybio?hl=en.
>
>
--
> > http://groups.google.com/group/diybio?hl=en.
> >
> >
> >
>
>
> --
> letters.cunningprojects.com
> twitter.com/onetruecathal
>
> --
>
> You received this message because you are subscribed to the Google Groups "DIYbio" group.
> To post to this group, send email to diy...@googlegroups.com.
> To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/diybio?hl=en.
>
>
--
Cathal Garvey
Dec 1, 2009, 12:59:08 PM12/1/09
to diy...@googlegroups.com
I think I'm missing a lot of your replies. The posts you've added between quoted sections seem incomplete, unless I'm mistaken?letters.cunningprojects.com
twitter.com/onetruecathal
twitter.com/onetruecathal
Daniel C.
Dec 1, 2009, 1:22:08 PM12/1/09
to diy...@googlegroups.com
On Tue, Dec 1, 2009 at 10:59 AM, Cathal Garvey <cathal...@gmail.com> wrote:
> I think I'm missing a lot of your replies. The posts you've added between
> quoted sections seem incomplete, unless I'm mistaken?
I thought it was a new kind of poetry. With apologies to Eugen, I
> I think I'm missing a lot of your replies. The posts you've added between
> quoted sections seem incomplete, unless I'm mistaken?
didn't find it particularly inspiring.
-Dan
Reply all
Reply to author
Forward
0 new messages