IUBio

AI Top-down, Robotics Bottom-up (Was: Bottom-up evolution...)

Arthur T. Murray ba672 at lafn.org
Tue Dec 24 13:16:50 EST 1996


                                 Seattle WA USA, TUES.24.DEC.1996

Hello again, (Professor/Dr./Mr.) Richard hall:

Thanks you for your second e-mail.  I am interspersing my answers.

> The physiologist generally agrees with the latter response
> (although the allusion to the von Trapp family was a stretch)

A British mountain climber in the 1920's was asked why he chose
to climb Mount Everest, and he said, "Because it is there."
Perhaps the von Trapp family made an allusion to the same statement.

> and disagrees with the former response (which was actually a
> rather lame dust off and not a response).

It was not so much a dust-off as an attempt to take the discussion
public on the Internet, where you saw the original statements.

> The Later-with which I sort of agree-
>>
>> The further suggestion is made that AI is a waste of time because
>> we already have an abundance of easy-to-manufacture human brains.
>> As one climbs Mount Everest, we build the AI because it is there.

> Please...

> My point was that the goal of AI can not possible be to recreate
> the animal brain, because that is economic folly, a waste of time.
> AI can use brain theory as a guide to model more efficient
> analysis systems, but we are not talking heart pacemakers or
> cochlear transplants in the same breath as brain transplants.
> My medical insurance would not be amused.  AI seems to be a
> discipline that has the objective of formulating models of information
> processing, period. It stand quite well on thant objective.

But AI is certainly an attempt to create artificial minds.

> The Former reponse-with which I disagree and find to be
> offensively trite-

>> A physiologist questions  "whether AI has any immediate relevance
>> to understanding brain function."   Yes, because theories of mind
>> present an easy target for neuroscientists to attempt to destroy.
>> If the theory indicated below survives attack, it points the way

> I disagree because:

> I cannot imagine (my weakness, perhaps) how presented AI analysis
> could confirm or refute a particular model or theory on brain
> function when brain function is electrical, electrochemical,
> and (bio)chemical in nature. What makes you confident that
> your analysis can refute theories based on experimental results?

I did not mean that AI would refute the theories of neuroscience,
but rather that my publicly stated linguistic theory of mind is a
convenient target for experimental neuroscientists to try to refute.
All along as my theory was being developed, it was done in conformance
with the guidelines deriving from experimental results, such as the
notions on visual feature extraction achieved experimentally by Hubel
and Wiesel in their work on the brains of cats -- work of Nobel stature.

Where my theory and experimental neuroscience come together is not
in the nuts and bolts of chemistry and electric charges, but in the
*logic* carried by the living cells as they interact to mediate a
flow of *information* in the brain-mind.  Mine is a theory of
information-flow, not of biological cell-processes -- which could
just as well be optic switches, or water valves, or quantum states.

> Aqueous electrical circuits are very different from circuit boards
> because the electrical events of the animal brain are intimantely
> linked to metabolic and electrical responses within the cell.
> Electrical or electronic model of brain or nervous function seem
> to miss one critical difference between the world of Ohm and the
> world of mind.  Biological events occur on a time scale that is
> at least 3 to 6 orders of magnitude slower than typical electronic
> switching presumably to accomodate electrophysiological and metabolic
> events within individual cells and circuits.  Otherwise, the brain
> would be totally electronic in nature with no need for
> neurotransmitters or neuromodulators.

And what the brain suffers in the slowness of its signal-transmission,
it makes up for marvelously in its massive parallelism -- still in
accordance with the linguistic theory which I am putting forward.

> It is doubtful the brain will prove to be either a mystical or
> an electrical wonder.  Yet, some AI discussions at this website
> seem to be dominated by physicists and mathematicians who truely
> believe the brain is a) an electrical oscillator or b)
> a massively parallel inductive field which strikes  me as mysticism
> or at least reductionism to the nth degree ;-).

Such strange and unwarranted ideas strike me in the same way.

> My objection is simply that quantum theory and induction field theory
> work swell for electronics and electricity; however, the brain
> is not a circuit board, nor is it a top down designed computer program.
> It is a paramecium or ameoba to the nth power, with emergent
> properties and intrinsic limitations. (This does not mean I take sides
> on the top down or bottom up issue.)  But it does lead me to conclude
> that the difference between simple neuronal synaptic connections and
> brain function is huge. The animal brain is a ponderous system with
> incredible redundancies yet with important subtlies as well.

On the Internet I am very staunchly stating that mine is a top-down
theory of mind, because I want people to critique and test the theory.
The bottom-up approach is, in my view, an admission of failure in AI.

> Clearly humans do not understand how the animal brain functions.
> We do have a good idea of what individual neurons do. We have
> some effective models for some components of brain function.
> We appreciate the importance of inhibition in brain function.
> AI is best when if addresses paradigms for information processing
> and when designing models for artificial control of motor behavior.
> While AI may eventually design analytic systems far superior to
> the animal brain, I have not yet seen AI discredit any particular
> theory of brain function that was based on experimental results.
> At best, AI has helped investigators frame questions for experimentation,
> but AI is a parallel (and valuable) field of investigation.

> In 1996, it is still seems to be apples and oranges. I am an interested
> and informed audience, so please prove me wrong, give me evidencs that
> AI is biologically relevant to the study of the animal brain.  Prove
> that I am incorrect and save snide retoric for your Wednesday poker game.

AI is biologically relevant because AI designs models which can be
implemented so as to mimic biological brains.  If the AI model works,
it points the way for what to look for in deeper scrutiny of the brain.

> rlh

> Richard Hall
> Comparative Animal Physiologist
> Division of Sciences and Mathematics
> University of the Virgin Islands
> St. Thomas, USVI  00802

> 809-693-1386
> rhall at uvi.edu

I am glad to respond to your inquiries, and I am not trying to engage
in snide rhetoric.

I would certainly prefer to conduct the discussion out in the open on
the Internet, so that others may join in, and so the ideas which I am
trying to publish will reach a greater audience.  Because I am merely
an independent scholar trying to make an individual contribution to AI,
I do not have the normal avenues of peer-review publication open to a
scholar such as yourself, and therefore I need as much Internet exposure
as possible.  That is why I paraphrased your e-mail and responded to it
in a public post on the Internet.  Please ask any further questions or
make any further comments by joining in the public Internet forums, but
please do not do as some do, who engage in personal ad hominem attacks.

Merry Christmas 1996 and Happy New Year 1997!

                                     Sincerely,

                                     Arthur T. Murray
                                     ba672 at lafn.org
                                     mentifex at scn.org
                                     uj797 at freenet.victoria.bc.ca
--
  /^^^^^^^^^^^\ Standard Theory of the Brain-Mind /^^^^^^^^^^^\
 /visual memory\           ________   semantic   /  auditory   \
|      /--------|-------\ / syntax \  memory    |episodic memory|
|      |  recog-|nition | \________/<-----------|-------------\ |
|   ___|___     |       |     |flush-vector     |    _______  | |
|  /image  \    |    ___V_____V__  word-fetch   |   /stored \ | |
| / percept \<--|-->/ library of \--------------|->/ phonemes\| |
| \ engrams /   |   \  concepts  / for thinking |  \ of words/  |
|  \_______/    |    \__________/  in language  |   \_______/   |

http://www.newciv.org/Mentifex/ run periodic Mentifex Web search.
http://www.complex.com.pl/~venom/science.html five English files.
û



More information about the Neur-sci mailing list

Send comments to us at biosci-help [At] net.bio.net