My talking about computational arithmetic might be a little misleading. Arithmetic is fully defined theoretically whereas AI is not. The possibilities for a new way to analyze or interpret (and process) a situation in general are probably endless. But, the point that I have tried to make in the past is that n-ary mathematics uses a highly efficient compressed form of representation but it also uses a highly efficient compressed form of computation. I am not sure if generalization using discrete reasoning (like categorical-substitution methods) and weight-based reasoning (like statistical methods) have to be the basis of all computational reasoning but they do have to be supplemented by variations of their application that can only come from learning. So my pointing to computational arithmetic to illustrate what is needed for AI may be misleading. But, what I was trying to say is that there are probably different computational ways to combine different kinds of concepts (or concept like knowledge) and the emphasis of the old argument of the distinction between neural networks (or statistical tools) and rule-based AI is not indicative of an awareness of the potential of these methods which I believe are missing from contemporary AI.
One conclusion that I came to as I wrote this was that an AGI program (even a weak prototype of an AGI program) probably does need to encode knowledge in the terms of discrete methods and weight-based methods in order to take advantage of a potential for computational efficiency. And as I continue to think about this I end up back in my usual loop. Weight based methods will have some advantage over discrete methods until better methods to deal with np problems are developed. Because while there are very efficient methods for binary arithmetic there are not efficient general methods for combining elemental logic where multiple literals (variables) are combined across AND-OR variations. Now I am starting to think that I have to reexamine the basic problem of Boolean Logic. Jim Bromer On Wed, Nov 25, 2015 at 10:37 PM, Jim Bromer <[email protected]> wrote: > I guess it takes a lot of work and the use of a few different frames > of reference to understand why n-ary arithmetic is so powerful. That > knowledge does not come to every programmer just because they can use > binary arithmetic. However, using a thought-experiment to try to > understand what it would take for a program to learn to develop binary > (or n-ary) arithmetic might be a frame of reference some programmers > could use to understand what I am talking about. ------------------------------------------- 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
