From: Stephen Reed (reed@cyc.com)
Date: Wed Jun 26 2002 - 22:02:01 MDT
On Wed, 26 Jun 2002, Ben Goertzel wrote:
> I was brainstorming a bit about your P2P project, and came up with some
> ideas. Just food for thought, nothing I'm taking too seriously at this
> point ;)
>
> First, one of the ways to attract funding to the project would be to create
> a particular application with the P2P software, and get a decent-sized pool
> of users.
>
> One thing that people *could* do with P2P is distributed problem-solving.
> You're aware of distributed.net and SETI@home and other similar projects ...
> these are all purely client-server based though. A more innovatively
> structured such application could attract a lot of attention.
I want to write a super-peer network for knowledge exchange among OpenCyc
images. A reference is:
http://dbpubs.stanford.edu:8090/pub/2002-13
We have 23,000 downloads of OpenCyc for Linux since April. Shortly we
will introduce a Win32 version, which I expect to have triple the download
rate. The majority of these downloaders are put off by the fact that
OpenCyc does not have an English chat interface. If that aspect were
sufficiently addressed I imagine that by this time next year we could have
over 100,000 OpenCyc downloads with a majority of somewhat satisfied
users. We will introduce a "knowledge grid adapter" that ties the
downloaded OpenCyc software into the super-peer network that we control.
The user gives us some machine cycles as the entry price for knowledge
grid inclusion. This is similar in operation to "spy-ware" but ethically
different in that the user must download and install the knowledge grid
adapter as a deliberate decision.
With a super-peer network, we should have over 10,000 OpenCyc servers
under central/hierarchical control for running Cyc applications. There is
of course high latency among the nodes but many Cyc activities have a
large grain size in comparison to the latency cost. For example,
extracting assertions from text is very time consuming, but if the
documents were distributed among many servers the speed up is almost
linear.
Regarding the details of the super-peer network, I believe that each
joining node could quickly test its available bandwidth and also learn
from the user its availability (always on the internet? static IP
address?) So given knowledge of the computational power and availability,
the central control could designate super peers to be responsible for say
a group of 50 casual nodes.
> There is a potential Novamente tie-in here, in that we could use Novamente
> to do the problem-solving within the distributed architecture... Not a
> super-smart Novamente install of course (we're assuming crappy commodity
> PC's), but select bits of Novamente appropriately chosen...
Over time we can work out ways for our respective projects (including
all on this list) to collaborate. If nothing else, OpenCyc could provide
a transport layer or an implementation pattern to follow.
> Next question as one pursues this line of thinking: What is the particular
> application, in the vein of finding extraterrestrial life or prime numbers
> or protein-folding?
>
> My first thought was the bio work we're now doing -- looking for patterns in
> gene expression data. This is "searching for the cure for cancer" and hence
> has some sex appeal...
>
> Then I thought of terrorism. How about searching public info databases for
> information pertinent to terrorism? There are some useful historical DB's
> out there I believe. "Download this app and use your computer's spare
> cycles to help fight terrorism!" has a certain mass appeal to it, nyet? ;->
Cycorp has several proposals pending in this domain, about which I cannot
yet comment. But we know US government interest is very high on this
topic and I exactly agree with your idea: use open-source materials to combat
terrorism.
> I fully understand that such a project (anything vaguely in this vein) would
> not bring anyone any revenue, but it could get a lot of press (esp. the
> terrorism angle), which would open a lot of doors for funding. In the early
> days of Webmind, our first real round of funding (not seed money) came on
> the heels of a Wall Street Journal article about us...
Agreed.
-Steve
--
===========================================================
Stephen L. Reed phone: 512.342.4036
Cycorp, Suite 100 fax: 512.342.4040
3721 Executive Center Drive email: reed@cyc.com
Austin, TX 78731 web: http://www.cyc.com
download OpenCyc at http://www.opencyc.org
===========================================================
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT