On Tue, 2002-04-02 at 08:46, ben goertzel wrote:
> 
> 1)  A (we hope!) passably efficient way to do "linguistic feature structure 
> unification" (a kind of language parsing) as a case of general logical 
> unification... this basically eliminates a separate NLP module, reducing it 
> to a set of special parameter settings for general cognition methods... as 
> it should be...


I've had good luck taking this approach to NLP, although it is too early
to say how good that luck actually is. This makes sense in theory, but
it really taxes the (human) mind when trying to analyze how this would
work in excruciating detail.

 
> 2)  Complex inference (inference dealing with quantifiers, variables, etc.) 
> is reduced to "simple inference on inheritance links" + "complex 
> procedures".  This reduction is carried out using combinatory logic.  I 
> suspect the human brain carries out a similar reduction in a very different 
> way
> 
> 3) Causal inference is reduced to "predictive implication" plus a special 
> case of "procedure learning" (the latter being the learning of "plausible 
> causal mechanisms")
> 
> 4)  Association-finding is reduced to simple probabilistic inference plus a 
> Hebb rule variant


These are all essentially reducible to the same structure, though with
slightly different traversal behaviors.  Doing the work to support one
should essentially provide the structure to support the rest.

 
> Overall, what we found is that a lot of mental functions are obtainable as 
> combinations or specializations of other mental functions, but not in 
> immediately obvious ways.


The "not in immediately obvious ways" is a killer. Part of this is
because people frequently define intelligence in terms of peculiar
behaviors that human minds tend to exhibit that may or may not be
relevant to the actual problem. The "yeah, but can it do <insert some
quirky human characteristic>?" question can be a problematic one, and
often requires more a lot more thought than the question may merit
simply because that is how people relate to "intelligence". There are
still a couple things that I would rather prove by doing rather than
completely verifying on paper first.

If we consider the human brain as a model, it is clear that limited
specialization of functionality at the lower levels must be true, at
least in the sense that it is one way to do it.  The human mind is an
extraordinarily complex construct built upon a very small number of
primitives. I don't think the primitives themselves are particularly
important beyond the fact that they can be put together efficiently to
support the basic functional structures.

I've been meaning to provide substantially more documentation for a
number of things, but these days I don't have time to provide much
written documentation for anything beyond the code I write (code is
self-documenting, right? :-) for my own engineering team, never mind the
rest of the world. Operating a startup has never been conducive to
having lots of free time... :-)

Cheers,

-James Rogers
 jamesr@best.com