Announcement: ANML Web Site (SPOILER)

Louis Savain savainl at earthlink.net
Wed Jan 21 16:48:56 EST 1998


In article <En5BED.G95 at freenet.victoria.bc.ca>, mentifex at scn.org
wrote:

>As the Associative Network for Machine Learning (ANML) website at
>http://home.earthlink.net/~savainl/ ushers in the temporal neuron
>revolution, counterrevolutionary cadres are duty-bound to present
>a Unified Theory of Cognition (UTC) which synthesises the salient
>features of ANML into a grander and more comprehensive structure:

  Why counter revolutionary?

>  /^^^^^^^^^^^\  Channels of Temporal Succession  /^^^^^^^^^^^\
> /visual memory\           ________    semantic  /  auditory   \
>|      /--------|-------\ / syntax \/  memory   |   memory      |
>|      |  recog-|nition | \________/------------|-------------\ |
>|   ___|___     |       |     |flush-vector     |    _______  | |
>|  /image  \    |     __|__  / \  ________      |   /stored \ | |
>| / percept \   |    /     \/   \/ Verbs  \-----|--/ phonemes\| |
>| \ engrams /---|---/ Nouns \    \________/     |  \ of words/  |
>|  \_______/    |   \_______/----/-/-/-/-/------|---\_______/   |
>|               |   ////////    /////////       |               |

  In the near future, accurate knowledge of the anatomy and/or
topology of the brain will be most important especially with regard to
feedback pathways.  The rest will emerge.  The above may well have
great significance in the way things are organized but I would place
the major part of my confidence in the self-organizing and associative
power of temporal neurons.

>Temporal events in the brain-mind are of the utmost significance,
>but only as SUCCESSIVE taggings to a long, neuronal memory fiber,
>and NOT as nearly simultaneous spike-firings which may or may not
>cause variations in the strength of the likelihood that a synapse
>will transmit a signal after having LEARNED to transmit the spike.

  I am having a hard time understanding the above, probably due to my
lack of familiarity with your terminology and your ideas.

>In the vernacular, the mindgrid does not CARE about close-but-no-
>cigar instances of almost-but-not-quite LEARNING.  The crucial e-
>vent in synaptic learning is that, at BOTH ends of an associative
>tag, synapses shall be so solidly strengthened that they MAINTAIN
>in perpetuity the tendency to fire and thus recall an association.

  I don't disagree with this.  In my own model, neurons can and do
reach a certain maturity beyond which their synapses cannot be
weakened or disconnected.

>So much of what Mr. Louis Savain posts to the AI newsgroups is in
>accordance with the mind-model of Mentifex AI, that he is owed an
>advisory post on the core differences between his ANML theory and
>http://www.scn.org/~mentifex/ the proposed Standard Model of Mind.

  I'm not entirely sure where we really disagree. Thanks for including
a link to ANML on your site.  I have returned the favor.

Regards,

Louis Savain

President, Marengo Media, Inc.



More information about the Neur-sci mailing list