I often wonder if I will need to rely on some sort of simplistic statistical analysis of data (like the way word co occurrence is used in web search) once my 'AGI' program starts producing so much data that the program becomes too slow. I don't think so. (I probably will use some simplistic statistical analyses but I don't think that will be a foundation of my AGI search and retrieval methods.) I am hoping that my structural methods, which will be dependent on finding key relations that tend to resolve many individual relations, should produce the kind of potential that such a program will need. A difference between an AGI program and a web-search program is that the AGI program will be producing all of the data that would comprise its potential for knowledge and reasoning skills, whereas a web search program has to examine data that had been produced by many different individuals and different software. However, even if the binding of knowledge through key structural relations is successful it will eventually consist of so many different objects that it too can cause slowdowns. The possibilities of relative hierarchies of structural relations, used for different purposes might relieve some of that. However, the problem is still similar to a Boolean Satisfiability problem and I don't have a solution for that. But I do have a possible strategy for the structural relation indexing problem. This strategy can be analogously simplified to the game of 5-in-a-row. 5 in a row is like Tic-Tac-Toe but it is played on a Go board and you win the game by getting 5 in a row. The best strategy is to put a piece down where you think that you can maximize the potential for getting possible 5 in a rows in the future. (It can be a metaphor for building key structural relations.) However, it is a mistake to think that an AGI problem is like a two dimensional go board. So what replaces the building on the spatial potential of the board in my metaphor? The potential of generalizations (matrices of cross-generalizations) built to handle similar situations.
-- 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
