[Comp-neuro] puzzlement

james bower bower at uthscsa.edu
Wed Aug 13 20:12:54 CEST 2008


just for clarity -- I was not comparing brains to airplanes --

but software systems to engineer planes - to software systems to  
reverse engineer the brain.

As others have pointed out, humans since the dawn of written language  
have always compared the mechanisms of human thinking (brains) to  
whatever is the most sophisticated and powerful technology of the day.

Greek Roman: neural humors - aquifers
Descartes: gears - machines
early 20th century - telegraph systems
middle 20th century - digital computers
later 20th century - parallel distributed analog computing devices

the problem comes when, instead of thinking of current technology as  
perhaps the best, but often bad, metaphor,  we actually come to  
believe that the brain IS one of those things.

On functional and thermodynamic evidence alone, we have never come  
close to building something as sophisticated or complex  or efficient  
as the brain - therefore, we need to be very careful about comparing  
the things we do build to brains.

As a corollary, encouraged by funding agencies (see the post about the  
recent meeting at NSF on the challenges related to understanding the  
neural basis for learning) the claim is often made that we are close  
to applying what we know about the brain to the construction of real  
devices to solve real world (often military - they have more money  
than anyone) problems.  The problem with those claims is that the  
subject and approach usually more reflect the problems as identified  
by the state of our technology ( in this case machine learning), than  
anything much having to do with the brain -- again, metaphor run amuck.

In the particular case of the NSF meeting for example, from the report  
I see little evidence that the question posed on this list earlier -  
as to how much learning the brain actually performs, and how much of  
the performance we attribute to learning is actually a very  
sophisticated application of prior (evolutionary) knowledge already  
built into the networks was even raised as an issue (seems to me it is  
a critical issue).  To their credit, cognitive neuroscientists and of  
course philosophers know about this problem -- but most neural network  
engineers and neurobiologists don't consider it.  Of course, if you  
are building a "learning machine" by definition it mostly 'learns' --  
and thus - the question of prior knowledge is irrelevant, and thus out  
of place in this kind of meeting.

And therefore, again, the problem of the assumptions we make going in  
influencing what comes out.

Jim Bower




On Aug 13, 2008, at 12:26 PM, Bill Lytton wrote:

>
>> as a physicist by education but working in neuroscience, I am  
>> really puzzled about the
>> approach to start from the single neuron to understand information  
>> processing.
>
> I am puzzled by your puzzlement
>
> There is great complexity in a neuron that directly effects its I/O
> -- the internal quantum complexity of a neutral particle doesn't  
> much effect its billiard-ball
> like behavior; the complexity of H2O does start to effect its  
> interactions which makes aqueous
> solutions difficult to model and to reduce cleanly to an analytic form
>
> the brain is often compared to a computer, and sometimes, as  
> implicitly by Jim, to a modern
> aircraft -- complexity arising from the connection of complex  
> components.  Individual
> neuron types differ in their complexity, perhaps lying somewhere  
> between an op-amp (a granule
> cell) and an 8086 CPU (2e4 transistors) -- a Purkinje cell
>
> there are papers by a phys-type that lays this out in a relatively  
> formal way
>  Csete, ME and Doyle, JC, Science 295:1664-1669, 2002
>  Carlson, JM and Doyle, JC, PNAS    99:2538-2545, 2002
> I recall that they also compare brains to airplanes
>
> Bill
>
> -- 
> William W. Lytton, MD
> Professor of Physiology, Pharmacology, Biomedical Engineering,  
> Neurology
> State University of NY, Downstate Medical Center, Brooklyn, NY
> billl at neurosim.downstate.edu http://it.neurosim.downstate.edu/~billl
> ________________________________________________________________
> _______________________________________________
> Comp-neuro mailing list
> Comp-neuro at neuroinf.org
> http://www.neuroinf.org/mailman/listinfo/comp-neuro




==================================

Dr. James M. Bower Ph.D.

Professor of Computational Neuroscience

Research Imaging Center
University of Texas Health Science Center -
-  San Antonio
8403 Floyd Curl Drive
San Antonio Texas  78284-6240

Main Number:  210- 567-8100
Fax: 210 567-8152
Mobile:  210-382-0553

CONFIDENTIAL NOTICE:
The contents of this email and any attachments to it may be privileged  
or
contain privileged and confidential information. This information is  
only
for the viewing or use of the intended recipient. If you have received  
this
e-mail in error or are not the intended recipient, you are hereby  
notified
that any disclosure, copying, distribution or use of, or the taking of  
any
action in reliance upon, any of the information contained in this e- 
mail, or
any of the attachments to this e-mail, is strictly prohibited and that  
this
e-mail and all of the attachments to this e-mail, if any, must be
immediately returned to the sender or destroyed and, in either case,  
this
e-mail and all attachments to this e-mail must be immediately deleted  
from
your computer without making any copies hereof and any and all hard  
copies
made must be destroyed. If you have received this e-mail in error,  
please
notify the sender by e-mail immediately.










More information about the Comp-neuro mailing list