Re: AI architectures in the 2000s

From: Ben Goertzel (
Date: Wed Aug 16 2006 - 20:33:24 MDT


Novamente and LIDA, for instance, are integrative AI architectures
that were NOT created with any particular application in mind...

Furthermore, I wouldn't want to emphasize the "hybrid" aspect of
Novamente too strongly. The hybrid nature is just a matter of
convenience. The design was arrived at by starting top-down and
figuring out what the requirements are for components of an AGI
system. Then, when some existing computer science technology seemed
like it could be modified and extended to fulfill the requirements for
one of the components, it was used in this capacity. OTOH some
components were not matched by any existing computer science
technology and needed to be designed from scratch. So, the hybrid
nature is just a matter of convenience -- it's not as though we
started out by saying "hey, let's glue together some current

I know that LIDA became a hybrid system via the same approach. The
high-level breakdown into components was done first, and then existing
tech was used to fill out the components where possible. But LIDA
really just re-uses existing tech, whereas the components of NM are
filled-in with our own original tech that was often created by varying
and significantly improving on existing tech...

As for "agents", I find that word to have been used so broadly that I
never know what it really means anymore ;-)

-- Ben G

On 8/16/06, Anthony Mak <> wrote:
> It seems there are many custom intergrative approaches for
> AI architectures and it seems they are build for a particular
> application or domain in mind.
> However in the literature, I often hear the words "intelligent
> hybrid system". Is this perhaps the closest thing to a
> discipline that studies AI architecture? In particular,
> I find using agent for intergration quite interesting...
> Anthony Mak
> > -----Original Message-----
> > From: [] On Behalf
> > Of Philip Goetz
> > Sent: Wednesday, August 09, 2006 11:31 PM
> > To:
> > Subject: AI architectures in the 2000s
> >
> >
> > I'm generalizing largely on the basis of the architectures I
> > saw at the AGIRI conference... but would it be fair to call
> > the 2000s the decade of the "kitchen sink" AI? Up through
> > the 1990s, you could look at an AI architecture, and say,
> > "It's rule-based/behavior-based/reactive", or else it would
> > be composed of 2 or 3 cleanly-separated layers, each being in
> > one of those well-known, identifiable architectural genres.
> > Now I find architectures harder to classify. It seems people
> > are not putting as much emphasis on finding the One True
> > Unifying Principle to create AI.
> >
> > What do you think?

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:57 MDT