Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!oitnews.harvard.edu!purdue!lerc.nasa.gov!magnus.acs.ohio-state.edu!math.ohio-state.edu!howland.reston.ans.net!nntp.crl.com!pacbell.com!amdahl.com!amd!amd.com!txnews.amd.com!news
From: Joseph Bridgewater <bridgwtr@vanzandt.amd.com>
Subject: RE: egers; Thinking Neural Network
Content-Type: text/plain; charset=us-ascii
Message-ID: <DD0JBy.K2p@txnews.amd.com>
To: egers@postman.essex.ac.uk
Sender: news@txnews.amd.com
Nntp-Posting-Host: corgi
Content-Transfer-Encoding: 7bit
Organization: Advanced Micro Devices, Inc.
Mime-Version: 1.0
Date: Tue, 8 Aug 1995 22:19:07 GMT
X-Mailer: Mozilla 1.1 (Windows; U; 16bit)
Lines: 56

>What you say sounds inspirational.
>
>Please do tell me more.
>
>Regards,
>
>Simon.

When our friend recommended we get on the newgroups and start posting 
about our developments, I told him I didn't really know what I would want 
to say.  So, he remcommended posting a "Request For Discussion", and 
basically just tell what we have and see what people write back and 
respond as we see fit (for obvious reasons, most of the details are 
proprietary and cannot be discused).

But for starters, my friend recieved his BS in comp-sci at Sacramento 
State U with a special focus on AI.  He has allways had an interest in 
developing an artificially intelligent program, espicially intent on 
voice-interactive.  He has fairly little experience as a software 
engineer, but a truelly brilliant mind.

I am self-taught.  I also have always had a special interest in 
artificiall intelligence, particullarly voice-interactive.  I have many 
years of diverse experience as a software engineer, but had had little 
exposure to details of how AI is done at that time.

Soon after we met, we found our common interest and began to work 
together on pursuing AI.  We did extensive research on the state of the 
science and much thinking about it.  Our first effort was to see what we 
could do with Xilinx LCAs (FPGAs).  Hardware development at home is not 
very practical and for various reasons, we discontinued that effort.

Later, using all we had learned, we began an effort on software AI, and 
again did much research.  We decided a lofty goal would be to create a NN 
paradigm that would "think".  After much brainstorming and design, we had 
actually come up with a wonderfull design that thinks.  Please see the 
other posts.

I began coding it in highly-portable C.  I completed the initial growth 
routines and many of the functional routines.  However, there was a bug 
somewhere in the generation routines.  For various reasons, we did not 
continue development of it as of about 2 years ago.  Among other 
reasons, we had begun work on a 3D-polyheron game engine with the intent 
to use revenue from that to fund further development.

Since then, I have worked some on it.  I believe I have corrected the 
original severe bug, but seem to have introduced another one.  This code 
is the most complex I have ever seen.  I expect that I will be able to 
hunt that bug soon, find it, and fix it.  I just have little time to 
spend on it (perhaps my motivation level doesn't cause me to make it a 
higher priority).  With running discussions in this newsgroup, I am 
certain to become more motivated and raise the priority level it.

For more info, please post on the newsgroup, or send me email if you are 
strongly compelled to.

