RE: AI in <what?>

From: Ben Goertzel (ben@goertzel.org)
Date: Sun May 19 2002 - 17:49:30 MDT


> Well, quite so. My worry is not that human spatial perception and
> organization is better/best. It's that the net kind of sucks.
> Now, dont' get
> me wrong, it's possible that a mindsystem with it as it's native
> environment
> might see emergent patterns and find this very enlightening and
> whatnot, but
> the Net is largely a repository of knowledge about the physical
> world. And
> without a way to relate this data, much of what it finds is going to be
> useless.
>
> So I worry.

I agree and I do share the worry, just to a lesser extent...

My work with net data has spanned 3 areas: text analysis, bio db analysis,
and financial time series analysis...

Intuitively, I get the feeling a lot of symbol grounding can be done in the
context of financial and bio db's. The language used to discuss the data in
the db's can be grounded in the db's itself. The feasibility of this is
suggested by projects like the Gene Ontology Project, which tries to
systematize the language used in describing bio db's ...

One vision I have here is of a bunch of simpler Alife organisms -- internet
agent/animals -- interacting with each other and mining financial and bio
and weather and commercial data etc. This digital biota becomes its own
little world; Novamente is to these little agents as humans are to other
animals.... And, it's not true that *all* this data is primarily grounded
in the physical world -- finance these days has as much to do with the net
as with the physical world for example.... Economic data as well...

I want Novamente to have feelers into various part of the Net, plus camera
eyes and artificial ears. It must understand music so I can jam with it ;>

In terms of Novamente, I don't think it's important that there be agreement
on what info. sources are necessary to what degree, not at the current stage
of the project. It is more necessary in Eliezer's AI approach because he
grounds cognition so tightly in "mental imagery." I think cognitive
algorithmics is pretty much sensory-modality and actuator independent...
The *emergent dynamics* on gets in a mind are highly dependent upon
sensorimotor particulars, but the *ability of a codebase to give rise to
emergent mind dynamics* in my view depends on cognitive stuff rather than
sensorimotor particulars. So I've chosen to focus on cognitive stuff for
starters...

These certainly are deep and interesting issues and I wish I had more to say
about them than "we'll see" but I guess "we'll see" is my opinion ;->

ben



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:38 MDT