Hello Ben, Yesterday on Tues.21.JUN.2011 I had a conceptual breakthrough in my Mentifex AGI programming and I wanted to tell you and yours about it, but it would have frustrated everybody if I described the new AGI technique without implementing it in software for purposes of demonstration.
Accordingly today I have _just_barely_ implemented the probably patentable but now simply "prior art" idea at http://www.scn.org/~mentifex/AiMind.html for MSIE. Here are the two changelog entries which describe the advance: Wed. "22jun11A.html" implements retroactive KB adjustment for Yes or No. Wed. "22jun11B.html" generates negation of sentence after input of "no". For several years I was worried about how the Mentifex AGI Mind would deal with human user input of simply "Yes" or "No" in response to a question issued from the AGI to the human user, such as, "Do robots play games"? We humans deal constantly with single-word inputs of "Yes" or "No" or "maybe", etc. It's like a toggle of information. But how can an artificial Mind react to Yes or No as we humans do? Yesterday morning it suddenly occurred to me that the question coming from the AGI is actually a skeletal idea in itself, one that needs confirmation with a "Yes" answer, or refutation/negation/denial with a "No" answer. It also was suddenly clear how to implement the exciting innovation in AGI software. So today I rather clumsily coded the new AGI feature in JavaScript, and next I need to code it in Forth so that I can improve the JavaScript. Since a question like "Do robots need food?" is the most immediate thought just generated in the AGI Mind, it is right there ready to be handled _retroactively_ in the software. If the human user answers "No" to a question from the AGI Mind, my software now attaches a negational flag to the verb-concept of the question being dealt with. If the user answers "Yes", right now the affirmative idea is the default condition, but of course I need to pin it down and allow for divergent answers such as "maybe" or "probably". This coding of retroactive knowledge-base adjustment is unusually tricky and complex for me, because the new functionality is actually at least three things happening all at once. Just getting the AGI to pose a Yes/No question is one software dilemma. The train of digesting "No" is a second part, and the default train of "Yes" is a third. Meanwhile I have been following the various AGI and Singularity discussions, and I feel rather strange because I am coding one AGI innovation after another in the AGI Mind. http://www.scn.org/~mentifex/AiMind.html has been getting typically 28 hits /~mentifex/AiMind.html from 21/Jun/2011:14:14:17 to 22/Jun/2011:22:06:59 according to the user logs, from all over the world. It makes me feel motivated to keep working, but the sprawling AGI program is getting almost too complex for one coder. Is it worth while to code AGI? Yes? No? If the AGI asks, I will respond "Yes"! Bye for now. Arthur -- http://cyborg.blogspot.com ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/8660244-d750797a Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=8660244-6e7fb59c Powered by Listbox: http://www.listbox.com
