This sounds like a flexible, multiple-inheritance class hierarchy to me. Or maybe a map (over conceptual space), with the transitive "part of" relation in Ryszard Michalski's papers that PM already linked us to. Or any other combined hierarchical/spatial structure. I think the key novel observation you're pointing to is that neighboring regions in conceptual space can be grouped *dynamically *to define a larger region that makes sense for the current reasoning context.
On Mon, Dec 10, 2012 at 5:56 PM, Jim Bromer <[email protected]> wrote: > I originally came up with the image of overlapping transcendent boundaries > while thinking about how we use logic. When we are considering a group of > related logical propositions we are putting boundaries around those > propositions and imagining them as existing as a complete universe. But if > were to learn or derived some new logical propositions, we do not need to > combine them to use them. We can consider the new propositions as if they > existed as a separate universe of their own. But if we realized that some > of the propositions of one group might have an impact on the other we could > transcend the original boundaries separating the groups. We could combine > them into a larger group or we could take those propositions from the two > groups that directly relate to each other and create a transcendent > boundary around them to consider them as a universe or a > domain to themselves. The boundaries do exist, there may be some overlap > in them but we may take a number of means to transcend those boundaries to > use them in consideration of other situations. This idea of transcendent > boundaries can be extended to any systems of ideas. > > The outline of the next steps is not important, it is more of a guiding > simile. > Jim Bromer > On Mon, Dec 10, 2012 at 4:43 PM, Piaget Modeler <[email protected] > > wrote: > >> I agree, you're only done with the architecture at that point. >> >> You still need to implement and get results. >> >> Can you (1) reiterate what you meant so that it is comprehensible to the >> rest of the world, >> and then (2) outline the next steps? >> >> ~PM. >> >> ------------------------------ >> Date: Mon, 10 Dec 2012 15:37:31 -0500 >> >> Subject: Re: [agi] Internal Representation >> From: [email protected] >> To: [email protected] >> >> No, produce something that actually works. Then you are done. Not simple. >> >> On Mon, Dec 10, 2012 at 2:43 PM, Piaget Modeler < >> [email protected]> wrote: >> >> >> You'd have to repeat that in Math, or Algorithms and Data Structures to >> make it comprehensible (to me). >> >> There are too many concepts such as "boundaries with transcendent >> overlaps" that are private to you alone. >> >> But I suppose you're understanding my method. List the requirements, >> constraints, dependencies, and >> assumptions. Then find a solution to address them all. Iterate until >> done. Very simple. >> >> Cheers! >> >> ------------------------------ >> Date: Mon, 10 Dec 2012 14:30:28 -0500 >> >> Subject: Re: [agi] Internal Representation >> From: [email protected] >> To: [email protected] >> >> So the program needs to be able to form 'ideas' but then it needs to be >> able to use them in different ways. For example you may know something and >> use that knowledge in your thinking without necessarily acting on that >> particular piece of knowledge. We can talk about an explosion without >> actually exploding something. >> Constraining the search space and using boundaries with transcendent >> overlaps are ways which our programs might avoid combinatorial explosions. >> And we can use branching indexes as a means to find more detailed >> information. Then, if particular 'ideas' are more closely associated with >> particular index branches (or with other 'ideas' that are associatively >> distributed in a similar way) they might not come to mind unless there is >> some other reason to be looking at that particular group of 'ideas'. If >> 'ideas' should be associated with other groups of ideas (or at other points >> on the index) then they can be cross associated. However, the index >> associations should be categorized or sub indexed in some way so that the >> system does not become overwhelmed by associating all 'ideas' at the root >> or initial path into the index. >> >> *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/23050605-2da819ff> | > 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
