On Fri, Dec 28, 2012 at 9:30 PM, Jim Bromer <[email protected]> wrote:
> On Fri, Dec 28, 2012 at 6:09 PM, Logan Streondj <[email protected]>wrote: > >> It's good to get the ideas into something more tangible like programs. >> > > Once you start to create an actual program your original options close in > on your plans. One of the problems in developing a viable AGI program is > that it has to be kept simple. You might have different methods that are > called in for separate cases but that does not always work so well. > > >> >> If you find that after whichever period of time you aren't getting much >> anywhere with your chosen route, perhaps you'll choose to contribute to >> another AGI project, perhaps my own. >> I did a version release today, now have support for primitive variables >> :-). By next version release quite possibly will be able to do factorial or >> some other simple procedures. >> And likely by next year will have English grammar, >> allowing for easier verification by others with smaller learning curve >> > > What does that mean? How does it support primitive variables. > It's an incremental support to full support of variables: Here I made a blog post: http://weyounet.info/2012/12/0-4-8-3-1-varname-su-value-be-ya-sysh-hspl/ > And how would you change your plans if something did not work. For > instance, if it did not have English grammar how would that affect your > concepts about AGI? The most important thing is to identify problems that > can be solved and problems that you don't have an answer for. After > working on logical satisfiability I have come to the conclusion that I > don't have a solution to logical complexity. So then in order to make my > AGI program work it would have to work by finding a way to overcome the > problem of logical complexity by some other means. It would have to acquire > a great deal of information by serendipity and then make > 'intuitive' guesses about relations that can only be structured through > correlation and the recognition that if process X could be applied to > situation Y then it suggest a path toward finding a solution. > > But what if my ideas did not work? Then it would tell me that I had been > making some mistake. If I could find good candidates that kept the program > from working then I should be able to test them pretty quickly. Perhaps > basic correlation is not good enough. Perhaps the program has to rely on > some kind of enhanced correlation where there are numerous reasons to > believe the process X *can* be applied to situation Y and that it *will* > lead to a path toward a solution. > Jim Bromer > > > > > On Fri, Dec 28, 2012 at 6:09 PM, Logan Streondj <[email protected]>wrote: > >> Hey I'm just offering you support to do some real coding. >> It's good to get the ideas into something more tangible like programs. >> >> If you find that after whichever period of time you aren't getting much >> anywhere with your chosen route, perhaps you'll choose to contribute to >> another AGI project, perhaps my own. >> >> I did a version release today, now have support for primitive variables >> :-). By next version release quite possibly will be able to do factorial >> or some other simple procedures. >> And likely by next year will have English grammar, >> allowing for easier verification by others with smaller learning curve. >> >> I'm programming in Assembly, but it is quite simple, >> only 16 assembly commands used, all register-machine, >> makes it easy to port and that kinda stuff. >> >> You would certainly have the capacity to improve upon current AGI >> programs, can look at the current roadmap and see where your ideas might >> fit in >> https://sourceforge.net/p/rpoku/code/ci/dc0d7886965d5cab645a4d5a220391b316c7c388/tree/roadmap.txt?format=raw >> >> >> On Fri, Dec 28, 2012 at 10:53 AM, Jim Bromer <[email protected]> wrote: >> >>> The most important case would be the one where it does show some >>> capability of learning a crude simplistic language but where it either >>> lacks subtlety or where it shows a wide variation of depth. In some cases, >>> for example, it might seem to be working but then it just cannot continue >>> to learn new things about a particular subject or where other subjects >>> which are comparably as easy seem to be totally beyond it. This is along >>> the lines of how other AI projects have fared. Let's say that my project >>> did turn out like this. Then in order to show that it was a valid concept >>> I would have to advance the program so that it was able to go further than >>> it had. The thing is that although the various AI methods are able to do >>> some tasks better than others they all fail at a level below what we need >>> to see in order to compare them to children. So being human like is not >>> the immediate goal, and being really smart is not the immediate goal. But I >>> would need to show that I could improve on contemporary AGI programs in >>> order to demonstrate that my ideas were workable and since my program would >>> be limited I would need to show that some improvements could be made to my >>> program. >>> >>> Jim Bromer >>> *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> >>> <https://www.listbox.com/member/archive/rss/303/5037279-a88c7a6d> | >>> 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> > <https://www.listbox.com/member/archive/rss/303/5037279-a88c7a6d> | > 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
