Philip Goetz wrote:
Some more notes on cognitive infrastructures:
IKAROS (http://www.lucs.lu.se/IKAROS/index.html)
IKAROS components correspond to brain areas, which are linked to each
other through arrays of real variables that represent neurons. IKAROS
is focused on representing the human brain accurately at a low level,
and duplicating specific brain regions and functions, purely for
research. It has a specified level of detail, as does an
architecture, but is an infrastructure rather than an architecture in
that the user can specify how the components are connected.
BrainStorm/J was an attempt at the Universidad Nacional del Centro in
Argentina to build re-usable Java components for intelligent systems.
The major product of BrainStorm has been JavaLog, an open-source Java
implementation of Prolog suitable for use in a multithreaded Java
application. The project focuses on belief-desire-intention (BDI)
planning architectures.
ABLE is an ongoing effort by the IBM Research Labs to produce a
library of low-level artificial intelligence components (neural
networks, function optimizers, logic engines, finite-state automata,
and so on) that communicate using a standard API. ABLE is thorough
and well-tested. Its API is meant only for components that reside in
the same thread on the same machine, and each component performs a
specific computational task rather than a specific cognitive function.
ABLE is hampered by an overly-restrictive licensing agreement that
allows only 90 days of use.
Alas, I fear that any attempt to build a horse (sorry, an AI) by making
a committee of all the existing AI techniques is going to produce
something that smells exactly as sweet as a camel ....
Richard Loosemore
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?list_id=303
