Ben: It seems that discussions of deep issues on this list rarely go very
far, because the
diversity of perspectives is so great that few of the vocal list
members agree even on the right way
to think about the deep AGI issues at the high level.... Whereas in
discussions within a
team focused on a particular approach, one can plunge much deeper into
specific issues
Ben,
Thanks again for helpful reply. Surely, it would help everyone to
discuss/identify the "diversity of perspectives" on practical grounding even
if briefly? Has anyone else any views here? How about Alan/ Bob M et al
identifying the basic principles of their/different perspectives here? You
don't have to go into great detail - just something similar to what Ben has
done, for a start.
-----Original Message-----
From: Ben Goertzel
Sent: Tuesday, August 28, 2012 4:42 PM
To: AGI
Subject: Re: [agi] Finding the "Right" Computational Model to Support
Occam's Razor
On Tue, Aug 28, 2012 at 10:20 AM, Mike Tintner <[email protected]>
wrote:
Ben,
Thanks for reply.
What then is the idea[s] for grounding this system?
The two papers
Perception Processing for General Intelligence, Part I:
Representationally Transparent Deep Learning
Perception Processing for General Intelligence, Part II: Bridging the
Symbolic/Subsymbolic Gap
linked here
http://wp.goertzel.org/?p=404
explain how we are intending to link DeSTIN in with OpenCog, to serve
as OpenCog's visual cortex...
When the Opencog brain says "pick up the box" to the robot, what will
"box"
and "pick up" connect to in the robot system?
"Box" is a fuzzy set of attractor patterns in the DeSTIN hierarchy (I
know you often interpret the word
"pattern" in an unusual way, but I'll continue to use it in the usual
way anyway ;p)
"Pick up" is also a fuzzy set of attractor patterns in DeSTIN (because
actions can also be perceived),
but also a fuzzy set of paths through configuration-space, represented
as such in the action hierarchy
"Box" and "pick up" are also Atoms in OpenCog's Atomspace, with
multiple semantic links indicating/
constituting their semantic content...
The DeSTIN attractors may be represented as Atoms in OpenCog, and the
(learned) links btw the DeSTIN
centroids involved in the attractors, and the corresponding Atoms in
OpenCog, constitute the
concrete means of the "grounding"
We have not built this yet though, at the moment we're just working
with virtual world agents, and
doing the first steps of tweaking DeSTIN to make it OpenCog-compatible
And what will the robot system have for "pick up"? Again some kind of
variable movement routine?
A fuzzy set of paths through C-space
Is this sort of thing BTW discussed much here/elsewhere? I can't recall
many
if any discussions about the exact principles for grounding - correct me.
We certainly should be discussing ideas for all this a lot.
It has been much discussed privately in the OpenCog HK team, and
privately between
me and David Hansons and Itamar, etc.
It seems that discussions of deep issues on this list rarely go very
far, because the
diversity of perspectives is so great that few of the vocal list
members agree even on the right way
to think about the deep AGI issues at the high level.... Whereas in
discussions within a
team focused on a particular approach, one can plunge much deeper into
specific issues
like this...
-- Ben G
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/6952829-59a2eca5
Modify Your Subscription:
https://www.listbox.com/member/?&;
Powered by Listbox: http://www.listbox.com
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com
