Connectionists: How the brain works

james bower bower at uthscsa.edu
Mon Jan 27 08:56:37 EST 2014


A couple of points I have actually been asked (under the wire) to comment on:

Do you need to model quantum mechanics?  No (sorry Penrose) - One of the straw men raised when talking about realistic models is always:   “at what level do you stop, quantum mechanics?”.  The answer is really quite simple, you need to model biological systems at the level that evolution (selection) operates and not lower.  In some sense, what all biological investigation is about, is how evolution has patterned matter.  Therefore, if evolution doesn’t manipulate at a particular level, it is not necessary  to understand how the machine works.

Thomas’ original post regarding radios is actually illuminating on this point.  It starts by saying:  "I guess it uses an antenna to sense an electromagnetic wave that is then  amplified so that an electromagnet can drive a membrane to produce an airwave that can be sensed by our ear. Hope this captures some essential aspects.”  This level of description actually, captures the essential structure of a radio - i.e. the level of function of the radio at which the radio’s designers chose the components and their properties.  So, how proteins interact is important, the fact that that interaction depends on the behavior of electrons is not.  The behavior of electrons of course constrains how a particular molecule might interact - but evolution does not change the structure of electrons - (at least as far as I know).

A other question that has been raised, somewhat defensively, has to do with the other level of modeling - and whether abstract models with no relationship to the actual structure of the nervous system, by definition, can not capture how the brain works?  there are several answers to this question.  The first and most obvious has to do with scientific process - how would you ever know?  If a model is not testable at the level of the machine and its parts - then there is no way to know and the model is not useful for what I am trying to do, which is to understand the machinery.    Put another way, if a model does not help in understanding “the engineering” (if you allow me the short hand), then it is also not useful in figuring out how the brain really works.  

There is another issue here as well - and that has to do with the likelihood that the model is correct.  This gets into issues of complexity theory, a subject many number of members of this list serve know better than I do.  However, I believe that one of the insights attributed to Kolmogorov  (who has been mentioned previously) is that there is a relationship between the complexity of a problem and its solution.  If there is a solution to a problem the brain solves that is simpler than the brain itself, then there is some constraint that has forced the brain to be more complex than it needs to be:  examples usually given include the component parts it has been ‘forced’ to work with, for example, or constraints imposed by the supposed sequential nature of evolutionary processes (making an arm from a fin).  

While it is at least worth considering whether the arm from fin argument applies to the nervous system, because we don’t understand how the brain works, we can’t really answer the question whether there is some simpler version that would have worked just as well.  Accordingly, as with the radio analogy, in principle, asking whether a simpler version would work as well, depends on first figuring it out how the actual system works.  As I have said, abstract models are less likely to be helpful there, because they don’t directly address the components.  However, while there is now finally some actual scientific work on this - still, most neurobiologists and even a lot of Neural network types, don’t seem to take into account how expensive brains are to run and the extreme pressure that has likely put on the brain to reach a ridiculous level of efficiency.  Accordingly, I am betting on the likelihood that the brain is not just some hacked solution, but in fact, may be an optimal solution to the problems it solves (remember, species also “pick” (again forgive the short hand) the problems they solve based on the structures they already have.  So, I myself assume that if there were a simpler physical solution, evolution would have found it.  At least, I can say with certainty that where ever it has been possible to measure the physical “ sophistication" of the nervous system, it is operating a very close to the constraints posed by physics (single photon detection by the eye, the ear operating just above the level of brownian noise, etc).  No reason, therefore not to assume until shown otherwise that the insides aren’t also “optimized”.  And, as I have said, we won’t be able to even address the question really, until we know how the thing works.


That said, even if a simpler solution is as effective as the brain, I am interested in the brain.  Engineers are interested in building other devices and they would likely always prefer (for obvious reasons) to be as simple as possible. As I have said, I believe evolution is under the same constraint, actually, but in any event the question I have chosen to try to understand (perhaps foolishly) is how brains really work - not if something else could do the same thing using a simpler form.

Finally, the above argument also relates to the issue of simple redundancy  - brain’s almost certainly can’t afford redundancy in the way that engineers have traditionally built in redundancy. To give credit where credit is due - one of the lessons that I did take from neural networks is that there are many more sophisticated ways to achieve fault tolerance than simple redundancy. Fault tolerance (this is how I actually think about learning, in fact) is a key requirement of brains. But almost certainly not accomplished by simple redundancy.  

Jim











On Jan 27, 2014, at 3:09 AM, Axel Hutt <axel.hutt at inria.fr> wrote:

> I fully agree with Thomas and appreciate, once more, the comparison with physics. 
> 
> Neuroscientists want to understand how the brain works, but maybe this is the wrong question. In physics,
> researchers do not ask this question, since, honestly, who knows how electrons REALLY move around 
> the atom core ? Even if we know it, this knowledge is not necessary, since we have quantum theory that tells uns something about
> probabilities in certain states and this more abstract (almost mesoscopic) description is sufficient to 
> explain large scale phenomena. Even on a smaller scale, quantum theory works fine, but only since physicists 
> DO NOT ASK what electrons do in detail, but accept the concept. In contrary, todays neuroscience ask 
> questions on all the details, what why ? Probably it is better to come up with a "concept" or more abstract
> model on neural coding, for sure on multiple description levels. But I guess, looking into too much much 
> (neurophysiological) detail slows us down  and we need to ask other questions, maybe more directing 
> towards experimental phenomena. 
> 
> A good example is the work of Hubel and Wiesel and the concept of columns (I know Jim and others do not like
> the concept), based on experimental data and deriving such a kind of concept. Of course these columns are not there
> 'physically' (there are no borders) but they represent more abstract functional units which allow to explain certain
> dynamical features on the level of neural populations (e.g. in the work on visual hallucinations). Today this concept
> is largely attacked since biologists 'do not see it'. But, again going back to physics, the trajetories of single electrons in 
> atoms have not been measured yet and so the probability density of their location has not been computed yet from 
> the single trajectories, but the resulting concept of probability orbits of electrons is well established today since it 
> works well.
> 
> Another analogy from physics (sorry to bore you, but I find the comparison important): do you believe that an 
> object changes when you look at it (quantum theory says so) ? No, sure not, since you do not experience/measure it.
> But, hey, the underlying quantum theory is a good description of things. What I want to say: in neuroscience we 
> need more theory based on physiological (multi-scale) experiments that describes the found data and permit to accept
> more (apparently) abstract models and get rid of our dogmatic view on how to do research. If an abstract description 
> explains well several different phenomena, then per se it is a good concept (e.g. like the neural columns concept).
> 
> Well, I have to go back to theoretical work, but it was very nice and stimulating attending this discussion.
> 
> Axel
> 
> Some of our discussion seems to be about 'How the brain works'. I am of course not smart enough to answer this question. So let me try another system.
> How does a radio work? I guess it uses an antenna to sense an electromagnetic wave that is then  amplified so that an electromagnet can drive a membrane to produce an airwave that can be sensed by our ear. Hope this captures some essential aspects.
> Now that you know, can you repair it when it doesn't work?
> I believe that there can be explanations on different levels, and I think they can be useful in different circumstances. Maybe my above explanation is good for generally curious people, but if you want to build a super good sounding radio, you need to know much more about electronics, even quantitatively. And of course, if you want to explain how the electromagnetic force comes about you might need to dig down into quantum theory. And to take my point into the other direction, even knowing all the electronic components in a computer does not tell you how a word processor works.
> A multilayer perception is not the brain, but it captures some interesting insight into how mappings between different representations can be learned from examples. Is this how the brain works? It clearly does not explain everything, and I am not even sure if it really captures much if at all of the brain. But if we want to create smarter drugs than we have to know how ion channels and cell metabolism works. And if we want to help stroke patients, we have to understand how the brain can be reorganized. We need to work on several levels.
> Terry Sejnowski told us that the new Obama initiative is like the moon project. When this program was initiated we had no idea how to accomplish this, but dreams (and money) can be very motivating.
> This is a nice point, but I don't understand what a connection plan would give us. I think without knowing precisely where and how strong connections are made, and how each connection would influence a postsynaptic or glia etc cells, such information is useless. So why not having the goal of finding a cure for epilepsy?
> I do strongly believe we need theory in neuroscience. Only being descriptive is not enough. BTW, theoretical physics is physics. Physics would not be at the level where it is without theory. And of course, theory is meaningless without experiments. I think our point on this list is that theory must find its way into mainstream neuroscience, much more than it currently is.  I have the feeling that we are digging our own grave by infighting and some narrow 'I know it all' mentality. Just try to publish something which is not mainstream even so it has solid experimental backing.
> Cheers, Thomas
> 
> 
> 
> -- 
> 
> Dr. rer. nat. Axel Hutt, HDR 
> INRIA CR Nancy - Grand Est 
> Equipe NEUROSYS (Head)
> 615, rue du Jardin Botanique
> 54603 Villers-les-Nancy Cedex
> France
> http://www.loria.fr/~huttaxel

 

 

Dr. James M. Bower Ph.D.

Professor of Computational Neurobiology

Barshop Institute for Longevity and Aging Studies.

15355 Lambda Drive

University of Texas Health Science Center 

San Antonio, Texas  78245

 

Phone:  210 382 0553

Email: bower at uthscsa.edu

Web: http://www.bower-lab.org

twitter: superid101

linkedin: Jim Bower

 

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.

 


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20140127/5caec36b/attachment.html>


More information about the Connectionists mailing list