Richard Loosemore said:
To answer your question, the complexity is so deeply embedded in the
thing that the AGI is supposed to be doing, that it is not at all clear
if there will ever be a way to build an AGI without it being complex.
Remember: that is the point of the argument - that it is not clear that
it can be avoided, and that therefore we must proceed on the assumption
that it cannot.
--------------------------------
What sort of effects of the kind of complexity that you are
considering do you think might occur?

My point is this.  The AGI program would be a reference program of
some kind.  The knowledge, or knowledge of its IO data environment,
would be represented in the form of references (of some kind).  These
references may refer to referents (in the 'outside world') that were
themselves complex without the program suffering some kind of runaway
complexity.  But I do think the system of references themselves would
exhibit some computational complexity that is crudely analogous to
that which may be exhibited in a program that models emergent behavior
(like an emergent model of a flock of birds), but that does not mean
that the program would go out of control anymore than the more
familiar computational models of emergent behavior go out of control.

So, my feeling is that emergent complexity in the referent universe
may be crudely represented by a system of references, and a system of
references may themselves exhibit emergent complexity but these are
not totally uncontrollable even though they can produce unexpected
results.  To me, the real problem is finding effective means of
learning when complexity and complications constitute the standard
problem.

Is your definition of complexity different?  (I cannot accurately
recall the precise terms that you previously used, but I doubt that
our definitions are that different.)  Do you think there is something
more to it?  Do you see some sort of problem with the kind of computer
program that would be needed to deal with this kind of complexity?
Are you saying that the contemporary computer program may not be
adequate to deal with this kind of complexity regardless of the amount
of memory, speed and parallelism that can be brought in?

Jim Bromer

-------------------------------------------
agi
Archives: http://www.listbox.com/member/archive/303/=now
RSS Feed: http://www.listbox.com/member/archive/rss/303/
Modify Your Subscription: 
http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4
Powered by Listbox: http://www.listbox.com

Reply via email to