Connectionists: Brain-like computing fanfare and big data fanfare

james bower bower at uthscsa.edu
Sun Jan 26 21:38:46 EST 2014


Sorry,

History tells us that it is simply not the case, in the vast number of cases, that anythings goes and every approach has equal probability of success.  While we are certainly in a age of cultural “inclusion” (at least supposedly) and many of us were raised with the idea that everyone’s opinion and effort is as worth while as everyone else’s.  And, importantly, aggressive mammals (primates) adopt all kinds of strategies to avoid direct conflict (less risky), it remains the case that history tells us that some approaches lead to more progress than others.  

30 years ago, it was only my hunch that the abstract form of computing structures designated “Neural Networks” despite the fact that they bore no real resemblance to anything neural - would not, in the end, tell us very much about the nervous system.  For sure, lots of smart and dedicated people, in part with funding provided (largely) by military agencies who were sold on the brain-like nature of these devices (I know, I was there when the sales pitches were being made - grumbling in the corner), have done remarkable things - in engineering.

But, illuminate the function of the real nervous system, they have not.  To my great relief, very few were even making that claim this year at NIPS - didn’t need to, Google is now the justification for the work.

Now, some version of the idea that the connections are THE key (thus: “connectionism’) has sold the white house on the idea that we should spend a large amount of money on further developing the “look see”, stamp collecting approach to trying to understand brain function.  Who cares that the connections are made on DENDRITES, which are still ignored by almost all abstract brain modelers, despite the fact that they clearly do the computation!!!  Perhaps not surprisingly, the principle proponent of this approach in the OSTP is the son of a neuroanatomist.  (and, I should say, a good one).  Furthermore, we can assure that we will have lots of pretty colors to show, which sadly now stands for progress in neuroscience, whether we know what they mean or not.

I restate for the last time on this list, in this half century:

1) ultimately, how the nervous system really works will depend on understanding its real circuitry and physiology - just like is also the case for cars, TV sets, computers, AND neural networks.  (how many times in NN meetings were there can be a lot of bizarre claims made, have I heard serious scientists say (including some contributing to this discussion) - “I don’t care what you claim it does, I want to know how it really does it”  why doesn’t that apply to the brain as well??) 

2) Over and over again in history, paying attention to the actual physical structure of a system has lead to unexpected breakthroughs - while theories and models rife with ad hoc assumptions and convenient abstractions have been a distraction.  THEY ARE NOT TESTABLE either.

So, yes, I stand by my conviction - with 30 years of Neural Network history (and abstract brain models as well) now behind us, that unless your theory or model is or explicitly can be linked to the actual physical structure of the nervous system, you are unlikely to make much useful progress in figuring out how the brain works.

In the end, it is the structure of the bran itself that determines how it works - only the religious disagree with that.

What that means, sorry to say, is that you actually DO have to REALLY learn about that structure, better yet, study it yourself. 

And I absolutely stand by the conviction that collecting data, free of any real (read rendered in mathematics) theory or model is NOT the way to go.  For Tyco Brahe the precise motion of the planets was clearly the right data to collect.  Nucleotide sequences, were clearly the structural basis for DNA - but who says that the most important information we need now to advance neuroscience is a complete map of neural connectivity?  As Bard pointed out, in effect, this enterprise doesn’t even have a defined end point - nor could it.  I predict now that many millions perhaps even billions of dollars later, we will have lots of pretty pictures, and that most of what everyone now thinks is true, they will still think is true - just like has happened in the vast majority of cases with the other major recent neuroscience operation based on the “analysis" of pictures, human brain imaging.  That is not progress.

However, the BRAIN project as currently conceived certainly has a lower barrier to entry (i.e. coefficient of  “inclusiveness”) than the requirement that you actually do the hard work of establishing a computational framework based on the real structure of brains first.  Remarkably enough, some have even stated its lack of computational structure as an asset!  

One for all and all for one I guess.


Jim



 
 

 

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/20140126/ac957ae9/attachment.html>


More information about the Connectionists mailing list