Ben Goertzel wrote: > My PhD is in math and I used to be a math prof, but I have > found no opportunity yet to use really advanced math in AI
My B.A. is in Latin and Greek and I used to be a teacher of Latin and German. In Mentifex AI, I use very little math and tons of linguistics. A brief progress report follows now. Earlier last week I finally got some true AI thinking with http://mind.sourceforge.net/Mind.html in MSIE JavaScript. By "true AI thinking" I mean that the Mind.html AI was able to respond correctly to simple knowledge-base queries. Prior to my work of 20 March 2007, the AI Mind would always veer off into spurious associations of spewing gibberish. Now for the first time I fully understand the broad outlines of the function of my own AI software, whereas previously the AI would malfunction but I did not know what was wrong. To tie this progress report in with recent discussion on the AGI mail list, I would like to report that the AI Mind may seem to do what a database could do, but uses arrays of associative memory instead of a database, and uses "spreading activation" (q.v.) to propagate spikes of excitation from concept to concept and from a linguistic superstructure down into the building blocks of thought -- nouns as a class; verbs as a class; other parts of speech. Mentifex AI is neuroscience and not evolution. The AI Mind has always been based on what I could learn over years of independent scholarship in the study of neuroscience. See http://mind.sourceforge.net/theory5.html for my AI theory. It is time for evolution to take over from neuroscience. Just before composing this post, I checked the user logs of the last one hundred visits to my AI project and saw the following instances of Netizens downloading the AI: C:/Documents and Settings/cazub/Desktop/Mind.html C:/Documents and Settings/Lucian_Twilight/My Documents/Ai_source/Mind.html C:/WINDOWS/Desktop/Abhi's Fun Stuff!!/NLP/Mind.html Prior to last week Mentifex AI sported a large user base of _malfunctioning_ AI. From now on, there is a basic, solid functionality which I may hope only to improve and not to worsen. Even if my own future coding efforts take a wrong direction and render the AI Mind inferior, the current diaspora of working code may take root in the environment of other AI programmers who surpass me. I feel psychological pressure to code Mind.Forth because http://aimind-i.com is where Frank J. Russo has devoted a new Web domain to his re-implementation of my Forth AI. Before I resume Mind.Forth coding (critiqued by the way in http://doi.acm.org/10.1145/307824.307853 ACM SIGPLAN 1998; and http://doi.acm.org/10.1145/1052883.1052885 ACM SIGPLAN 2004), Mind.html needs further tweaking and fine-tuning, so that the AI will not only think a single thought in response to user input but will exhaustively think through all the thoughts lodged in its associative knowledge base. Then in Mind.Forth I will aim for machine reasoning. Respectfully submitted, A.T. Murray -- http://www.technorati.com/wtf/mentifex ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=303
