 |
Network Programming - The main goal of webneurons is to develop a better way of programming computers. Demonstration of a simple webneurons network. Also includes a minimal javascript XML parser of 37 lines.
|
|
|
|
|
Parallel Processing. An email to National Academies
|
13 Jan 2011
To Samuel H Fuller and Lynette I Millett
National Research Council
In reading your prepublication book "The Future of Computing
Performance" (Game over or next level) I notice some features of my
project at http://webneurons.com may help towards a general parallel
programming system, or even provide one.
Imagine a collection of webpages, each page has programming within,
say javascript or whatever. Each page runs continually on it's own
small processor and has inputs from other pages and outputs to other
pages. So you can build up a network of communicating pages. Each page
has it's own memory in the form of variables, object, array or
whatever attached to the page's processor. So memory is distributed
rather than shared.
The whole thing is inherently parallel using a distributed memory
system and a huge number of small processors each with attached
memory. I suppose you could group several pages on one processor if
you wanted.
With this system I think you could program any computing problem from
small ones even up to supercomputer level. I notice that many
supercomputers use massive small processor collections with local
memory. Shared memory was never a good idea I thought.
Now look at it from another angle. Neurons in the brain each have a
continually running processor and memory within and they have inputs
from other neurons and outputs to other neurons. That is all it takes
to get massive parallel operation and all the ingredients for AI too!
The point is that if you substitute "neuron" for "webpage" in this,
then you have my system as described at the start.
I cannot see why nobody has tried this before.
I think it's been staring us in the face for years, just model the way
the brain does it, but not with inflexible neural networks, rather put
some decent programming language inside the "neurons" and allow
flexible inputs and outputs just like the brain does, then we can work
from there. You have to be able to write any program and neural
networks are too inflexible for that. The brain is the original and
best parallel processor system tried and tested, why look elsewhere.
Sincerely,
John Middlemas
|
|
|
|
|
|
Back
Site written by John Middlemas