Article View: comp.ai
Article #9331Announcement: The Temporal Neuron Revolution
From: Louis Savain
Date: Tue, 23 Sep 1997 00:00
Date: Tue, 23 Sep 1997 00:00
96 lines
5682 bytes
5682 bytes
Announcement: The Temporal Neuron Revolution [English] RESEARCH RESULTS ANNOUNCEMENT September 23, 1997 Marengo Media, Inc. is a software development firm located in Southern California. We have been involved in fundamental research in connectionist approaches to AI for over three years. We have made little progress until quite recently. In the beginning of this year, we decided to investigate a novel theory of intelligence, one which maintains that learning in biological nervous systems depends entirely on the search and discovery of temporal or causal associations between events. Essentially, our work stems from the hypothesis that, at the level of abstraction which interests us, i.e., emergent intelligence and complex signal processing, neurons are nothing more than simple causality processors and that the brain is but a network for the detection of causal correlations. We think that classical conditioning is directly linked to the ability of neurons to make sequential associations. We were elated to find out later that recent discoveries in neurophysiology having to do with the temporal aspect of synaptic plasticity (H. Markram, Science, 1/10/97), were in perfect agreement with part of our model. In April, we became sufficiently convinced of the viability of our thesis to begin work on the design and implementation of a simple and powerful neuron for the detection of temporal order and the creation of temporal associations. Today, we are happy to announce that early results from tests conducted using our chess playing hierarchical network, have exceeded all our expectations. Watch this newsgroup for the next announcement from Marengo Media regarding the coming AI revolution, that is, the temporal neuron revolution. The announcement will be made under the title "ANML: Associative Network for Machine Learning." There will be software and source code available for free downloading. After that, all follow-up messages should be posted to the newsgroup "comp.ai." Thank you for your attention. Sincerely, Louis Savain President, Marengo Media, Inc. PS. Please, do not send email. [Fran�ais] ANNONCE DE R�SULTATS DE RECHERCHE 23 Septembre, 1997 Marengo Media, Inc. est une maison d��dition de logiciel situ�e en Sud Californie. Il y a plus de trois ans que nous conduisons des recherches fondamentales dans les domaines connexionistes de l'IA. Jusqu'� ces derniers temps, nous n'avons pas eu de succ�s. Au d�but de cette ann�e, nous avons d�cid� d�essayer une nouvelle th�orie de l�intelligence. Cette th�orie maintient que l�apprentissage dans les syst�mes nerveux biologiques d�pend enti�rement de la recherche et d�couverte des associations temporelles ou causales entre les �v�nements. L'essence de notre travail nait de l'hypoth�se que, au niveau d'abstraction qui nous int�resse, i.e., l'intelligence �mergente et le traitement des signaux complexes, les neurones ne sont que de simples processeurs de causalit� et que le cerveau n'est qu'un r�seau hi�rarchique pour la d�tection des corr�lations causales. Nous pensons que le conditionnement classique est directement li� � la capacit� des neurones de faire des associations sequentielles. Ainsi, nous �tions ravis d'apprendre plus tard que les r�centes d�couvertes en neurophysiologie concernant l'aspect temporel de la plasticit� synaptique (H. Markram, Science, 10/1/97), �taient en parfait accord avec une partie de notre mod�le. En avril, nous �tions devenus suffisamement convaincus de la viabilit� de notre th�se pour entamer un projet ayant pour objectif la mise au point d�un simple et puissant neurone con�u pour la d�tection de l�ordre temporel et la cr�ation des associations temporelles. Aujourd�hui, c�est avec plaisir que nous annon�ons que les premiers r�sultats des essais ex�cut�s sur notre r�seau hi�rarchique, construit sp�cialement pour jouer aux �checs, ont d�pass� toutes nos pr�visions. Surveillez ce newsgroup pour le prochain communiqu� de Marengo Media concernant la r�volution d�IA � venir, i.e., celle du neurone temporel. L�annonce se fera seulement en Anglais (sorry) sous le titre "ANML: Associative Network for Machine Learning." Il y aura un programme avec code d�origine disponible pour t�l�chargement gratuit. Apr�s cela, tout message "follow-up" devra �tre post�e au newsgroup "comp.ai." Nous vous remercions de votre attention. Sinc�rement, Louis Savain Pr�sident, Marengo Media, Inc. PS. Pas de messages electroniques s'il vous plait.
Technical Details
Message-ID:
<342b36de.10039603@news.earthlink.net>Path:
arc1-ca.pugleaf.net!unknown!not-for-mailShow Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII X-Google-Thread: 10d5b4,10bd0b81cba64349,start X-Google-Attributes: gid10d5b4,public X-Google-Thread: 10ca9b,10bd0b81cba64349,start X-Google-Attributes: gid10ca9b,public X-Google-Thread: 10d04f,10bd0b81cba64349,start X-Google-Attributes: gid10d04f,public X-Google-Thread: 101fc6,10bd0b81cba64349,start X-Google-Attributes: gid101fc6,public X-Google-Thread: f6b1a,10bd0b81cba64349,start X-Google-Attributes: gidf6b1a,public X-Google-Thread: 10461e,10bd0b81cba64349,start X-Google-Attributes: gid10461e,public From: savainl@earthlink.net (Louis Savain) Subject: Announcement: The Temporal Neuron Revolution Date: 1997/09/23 Message-ID: <342b36de.10039603@news.earthlink.net> X-Deja-AN: 274977920 Distribution: inet Organization: EarthLink Network, Inc. Newsgroups: comp.ai,comp.ai.philosophy,comp.theory.self-org-sys,comp.theory.dynamic-sys,comp.ai.neural-nets,fr.comp.ia