From: Ben Goertzel (email@example.com)
Date: Fri Aug 18 2006 - 21:49:03 MDT
Another perspective is to look at NM and other integrative
architectures as moving beyond the elegance-and-simplicity based
aesthetic of contemporary math/physics/computer science, and accepting
an aesthetic that values the heterogeneity and complexity associated
with living systems...
Most prior AI approaches have been based on one special trick -- one
learning algorithm, one knowledge representation, etc. How elegant it
would be if there were one special thing like this at the core of
intelligence! (some on this list seem to think this "one special
thing" is bayes' theorem -- an assertion that I find erroneous in the
same way that it's erroneous to posit reinforcement learning, or
semantic networks, or any other critical yet limited concept as the
one special thing at the center of intelligence...)
Moving to an aesthetic in which there is no one special trick, but
rather an appropriate arrangement of heterogenous parts, is a big
aesthetic change for most hard-science-theory-oriented people...
> I think there are strong connections between the hybrid aspect of
> Novamente, and the failings with logical positivism pointed out by
> Wittgenstein, and later by post-modern philosophers such as Derrida.
> The reason you're free to build a hybrid architecture is that you
> aren't clinging to the belief in semantic purity that was a common
> feature of logical positivism, 1980s unified architectures for
> cognition, and modernism in general. Once you accept that semantics
> and syntax can't be analyzed separately - which is a key part of what
> the post-modernists said, muddied up with Marxist terminology in an
> attempt to lend radical chic to what is basically a linguistic problem
> - you're free to put together hybrid systems, without worrying that
> you can no longer know exactly what something represents, or what your
> architecture can and can't compute.
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:57 MDT