| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1049.1 | | HPSTEK::XIA | In my beginning is my end. | Mon Mar 12 1990 10:37 | 6 |
| Forgive me for being pessimistic, but I doubt there will be a big
market for Neural Networks in the next twenty years. I would be very
interested to learn if you know any marketable applications at the
moment.
Eugene
|
| 1049.2 | a better place to ask | SHIRE::GOLDBLATT | | Tue Mar 13 1990 03:43 | 5 |
| re. .0
Try placing your note in NODEMO::MARKETING. They're the ones who
should look after the future of Digital's product set.
|
| 1049.3 | Here is another pointer, and more... | TKOVOA::MIZUNO | Garfield is my middle name :) | Tue Mar 13 1990 19:38 | 17 |
| Re.0:
You may also post your note in ISTG::AI.
As I recall there are some notes on Neural Net tools.
Personally, I don't see any *useful* commercial implimentations
in Neural Net technology yet. Companies who claim to be using
Neural Net technology in their products are using the techonology
as eye catchers. From many NN/AI researchers point of view, such
products are just bogus! I asked many NN/AI researches at various
international AI conferences and I got the same answer every time.
Well, if that is a consensus amoung NN/AI researches, when will the
NN technology be matured engough to set out as a commercial product?
5 to 10 years in the future is the most commonly accepted view.
*Lucy*
|
| 1049.4 | | EAGLE1::BRUNNER | VAX Architecture | Tue Mar 13 1990 19:57 | 2 |
| I'll start worrying about neural nets when we truly figure out how to
painlessly program for large MIMD arrays.
|
| 1049.5 | D'ont we miss the boat? | RTOISE::SUPPORT_HS | Don't marry, be happy! | Wed Mar 14 1990 06:06 | 17 |
|
Well, I do agree, that's nothing in the market yet, you
can take seriously.
What I'm concerned about is: we can be sure,that's all to
late, when things are proven.Nobody is able to buy know-how
then,whatever it might cost,because you can't buy it for money
at all (that's what I'm pessimistic about).
So, what happens actually within DIGITAL in terms of research
projects,founding and so on?
Helmut
|
| 1049.6 | One example | SCAACT::RESENDE | Just an obsolete child | Wed Mar 14 1990 14:14 | 8 |
| Well, I'll give you one class of application.
We're working with an ISV called Excalibur Technologies that has been working
with neural technology. They market a number of products around pattern
recongition including text retrieval, signal and image recognition. And
they're being quite well received.
Steve
|
| 1049.7 | Real Product | ROXIE::TOGNONI | | Wed Mar 14 1990 18:20 | 14 |
| Handwriting recognizer
I believe it was CNN's Science and Technology week where I heard that
SONY had introduced a handheld computer/device that allowed a person
to handwrite onto a screen/pad and have the information stored in
memory in something like ANSI form. The device utilizes neural nets/
fuzzy logic to recognize English and Japanese characters. It was
targeted at secretaries and court stenographers. I think they said
something about being able to convert short-hand directly to long hand.
Relying on marketing to predict when there will be a market for Neural
nets seems awfully passive. Sometimes you gotta create the market.
Keith
|
| 1049.8 | Yawn | LENO::GRIER | mjg's holistic computing agency | Mon Mar 19 1990 23:42 | 19 |
| Nothing new under the sun (sic).
Neural networks are computationally equivalent to traditional Von
Neumann model computation engines. It's just that they can use more
efficient algorithms for certain computational jobs (feature recognition,
for example), than linear processors.
So, if you have an interesting problem, which has a NN solution which
is efficient, great. If the interesting problem doesn't have a non-NN
solution, it doesn't have a NN solution either. Not quite so interesting
any more.
Just goes with my thinking that more and more dedicated hardware will be
used in the future to address particular computing needs, esp. as the
algorithmic problems are becoming better understood and more applicable to
hardware design techniques...
-mjg
|