On Tue, Apr 9, 2013 at 6:48 PM, Piaget Modeler <[email protected]>wrote: Any representation which offers reification (i.e., hypostatization) would address your complexity concerns.
I disagree with your literal statement although I agree that is part of it. Jim On Tue, Apr 9, 2013 at 6:48 PM, Piaget Modeler <[email protected]>wrote: > Any representation which offers reification (i.e., hypostatization) would > address your complexity concerns. > > ~PM > > ------------------------------ > Date: Tue, 9 Apr 2013 18:38:02 -0400 > Subject: [agi] Composite Data Languages Are Necessary for AGI > From: [email protected] > To: [email protected] > > > Many complicated computer programs may have something that acts much like > what I am calling a composite data language, but I am arguing that an AGI > program has to explicitly rely on such a thing to show sufficient > scalability to act as a demo. Furthermore, I am arguing that natural > languages are composite data languages as well. > I believe that an AGI program needs to be able to refer to the 'objects' > of a 'situation' using combinations of referential 'data objects'. Once > these data objects are defined, the AGI program has to be able to generate > new ways of referring to various composites of the subject matter in order > to create simpler references for future use. So then a composite data > referent can be developed to refer to a combination of aspects of a > 'situation' when it is effective to do so. I believe that this system is > absolutely necessary for AGI because the fundamental problem that faces us > right now is one of AGI complexity. By being able to create composite > referents we might be able to significantly reduce the complexity problem. > What I am saying here is that by carefully considering this point of view, > I believe that some AGI programmers might be able to enhance their > programming. > A composite reference language has to be decomposed efficiently. Neural > Networks, for example, lack that ability and I believe that is part of the > reason why they failed to live up to their early promise and hype. The > AGI program should typically be able to find some way to discover and look > at the parts of a composite reference when it seems efficacious to do so. > However, some composite references won't be easily decomposed. The parts > of a reference (the real world object being referred to, for example,) > might be very difficult to discover. But most of the referents should be > partially decomposable. > The fact that a referent is easily decomposable does not mean that a > referent is easily composed from the parts. If you know what the > composite reference is supposed to mean, you can discover or decompose the > parts from which it was composed by keeping the original reference in mind. > However, when you begin drawing references (like 'words') together to form > a new composite reference, the 'words' that you will be using will be taken > from other previously existing composite references. There is bound to > be some confusion in this process but those issues can be reduced with > experience. But even if the composite references are easily > decomposable, good compositions are more difficult. > If my theory is right I should be able to create a good AGI demo within 2 > years. > Jim Bromer > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/19999924-5cfde295> | > Modify <https://www.listbox.com/member/?&;> Your Subscription > <http://www.listbox.com> > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/10561250-470149cf> | > Modify<https://www.listbox.com/member/?&;>Your Subscription > <http://www.listbox.com> > ------------------------------------------- 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
