Derek,
I'm interested in why "lines" create difficulties (of acceptance) for you.
What do you think *maps* which you seem to have no problem accepting, are
made of ... but lines? What do you think image schemas, on which cognitive
embodied science is founded, are made of ... but lines?
The key difficulty here is that you seem to be still thinking of *computer
programs* - rather than robots. I take it for granted that the only real AGI
will be a robot, and, in that case I suggest - but by all means object -
that a robot that directs its own thoughts and actions - and is directed
from outside by others - by means of lines (goal-directed lines) is not at
all hard to accept.
It is arguable - and is argued by many linguists - that language began and
begins with signs/sign language - people pointing each other along certain
lines - to look, move and act along certain lines. Even one of Ben's demos
is based on lines - a dog being pointed to a ball - though it wasn't
properly instantiated.
The larger problem is that our whole culture treats language as if it were
purely a means of *depicting* the world - a set of *pictures*. Actually, it
is primarily a system of guiding actions. But culturally we haven't realised
this yet. So I don't think you and others are attuned to seeing AGI - and
the "language" of an AGI - as centrally about guiding action.
I accept that a good illustration of how an "outline" language will work
with a robot is needed to make all this clear to you and others - and I have
one - but again it requires a good deal of time & space to set out.
-----Original Message-----
From: Derek Zahn
Sent: Saturday, January 26, 2013 12:42 AM
To: AGI
Subject: RE: [agi] The many different types of embodiment
Mike Tintner wrote:
We are continually following (and imposing) lines of action and thought.
It's hardly a big stretch to see those lines as represented in the brain -
there to guide the planning of our movements and thoughts.
It is a big stretch, to me. You apply the word "line" in all sorts of
creative and metaphorical ways to various activities, which is fine for
us -- that is how humans think -- but the challenge of AGI is figuring out
how to get computers to do those things. It doesn't do any good to suggest
mechanisms that require general intelligence just to use...
And besides it is far from clear to me that implementing an artificial
intelligence using some literal sense of "lines" is necessary or even a good
idea. To be convincing you have to at least give some indication about how
a computer program would go about doing it.
but presumably, animals - and any AGI robot - must have some internal
equivalent, however crude, of our external maps
Certainly! This is much more sensible. Brains do in fact have at least one
extensive and complicated mapping system (google "place cells" for more
info, and I think the various native coordinate systems used in the brain
could possibly be put to similar generative use). My personal tentative
belief is that at least some of the "basic" concepts underlying our
conceptual system -- image schemas and other foundational general/powerful
things -- are laboriously abstracted in early childhood from these hardwired
special-purpose brain circuits (possibly some have, in a rudimentary way,
been "compiled" into the genome to make that easier). Another example could
involve abstracting the so-called "object file" which our brains use to keep
track of important objects in our environment... perhaps the whole "object"
concept starts from there... I don't know how this would work (yet)... Jean
Matter Mandler has a theory she calls Perceptual Meaning Analysis which
looks promising and is pretty close -- but too vague for engineering use.
Derek Zahn
-------------------------------------------
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-f452e424
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com
