I have spent about 12 hours on my AI / AGI project so far this year, so I decided to end my self-imposed ban from the group since participating in this group seems to motivate me to work on my project.
I think that I can use language to instruct an AI program to learn about the things that I can communicate through language. The program would be able to pick up factoids via simple language (similar to regular strings and context-free strings) but it would also be able to try to utilize knowledge and to acquire more insight through more complicated language (similar to context-sensitive strings). For example, (an abstract example), if it was trying to understand something that did not conform to some facts that it was familiar with (using higher level sentential structures that it was familiar with) it could use progressively more simple sentences to try to fit the new facts into the knowledge that it had previously acquired. Since I would be able to detect this, I, (as programmer-teacher), should be able to be able to make a good guess about the kind of knowledge that it might use effectively at that point. One problem is that word-concepts may change, not only in their application but in the level of abstraction and particularization. A word or a phrase can even be both more particular and more general at the same time. An exemplar is an example of this. For a program to benefit from this kind of abstraction-generalization polymorphism (many shapes-not OOP polymorphism) it has to be capable of both context-free (like) communication and context-sensitive (like) communication. It has to be able to plug new ideas into preexisting knowledge in simple ways, but then it has to try to derive some guesses about these new ideas in more sophisticated ways. This is a very simple explanation of how my program should work. I feel that AGI has failed just because knowledge is too complicated. But, I want to show that a basic strategy which includes the kind of thing that I mentioned here makes sense and it should work - at least until the knowledge base becomes too complicated. 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
