I wrote that the reason computers, physics and computational
mathematics have advanced so far is because the n-ary numerical
representation, where n>1, is an effective compression of the unary
system and because computational mathematics are able to use those
representations without decompressing them (into unary form for
example). I was thinking about it today and I realized that all the
n-ary numbers are able to represent every number uniquely so I
wondered what was lost going from unary to binary. A one-to-one
correspondence between any equal parts of two or more numbers is
possible with unary representations but it is not with binary
representations. So, for example, if you want to subtract an equal
amount from two n-ary numbers you won't know -for all possible cases-
how many columns of digits will be involved. Therefore you cannot use
that information as an absolute in a mathematical abstraction. This
means that many great functions and mathematical methods, some of
which are quite ancient, cannot be efficiently used for
multiplicities. So, for another example, even if you carefully laid
out a sophisticated vector system that included individuals and
relational information as well, you wouldn't be able to get very far
with it because of the complications that are created by the lack of a
one-to-one correspondence at the base representations. You would have
to do a great many operations to put all the information in the
appropriate form to use it for a complex analysis of a situation and
if you tried to change one small thing it could require a complete
recalculation. If you want to be able to keep track of corresponding
bits you have to use a flag system in which each columnar bit is meant
to categorically represent something. I don't know what the
mathematical name for this kind of representation is but it is very
inefficient for complicated systems because it is essentially a unary
system. What happens with a logical formula is that multiple
operations create a distribution of these corresponding (bit)
references by recombining them. In some ways logic is like a neural
network. The summary or result of logical functions are represented in
a system that is distributed and combined. NNs are usually thought of
as fields and the result of a logical formula as a line of bits but
there is something in common. Once the polynomial time solution for
Boolean Satisfiability is worked out a more insightful examination of
the similarities and differences of these systems may lead to novel
learning and representational systems.
Everything is finally starting to become very clear to me. I don't
mean that I have it all figured out - that would be absurd - but I do
mean that I can finally start to see the path. I can't describe it
very well but I can see it.
Jim Bromer


-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to