We have recently expanded the conceptual flag-panel of four AGI Minds from fifteen associative tags to twenty-one associative tags. With the full complement of associative tags, an AGI Mind such as http://ai.neocities.org/AgiMind.html may think with much more complex ideas than before. Today we would like to work on the Natural Language Understanding (NLU) of user input that includes prepositions, like “for” in the user input “I work for money”. We previously solved the problem of differentiating between nouns used as the object of a verb and nouns used as the object of a preposition, but the AGI Mind did not have the mental functionality for assigning the proper associative tags for both the storage and the retrieval of prepositional phrases. The new tpr tag for “time of preposition” should make it easier for the mid-AGI software to store and recall prepositions.

AGI Roadmap: Thinking with Prepositions

In the ghost309.pl AI we have introduced a new group of transfer-variables designated as $px1 and $px2 and $px3 so that the EnNounPhrase() module may detect linkage from a candidate-noun to a preposition and inspect immediately the flag-panel of the indicated preposition in order to latch onto $px1 as the conceptual time-point of the object of the preposition. Then in the EnPrep() English-preposition module we plan to use the briefly immutable $px1 time-point value to fetch the object of the preposition from memory and speak it as part of an idea being recalled from memory. We were trying to use other variables for the same purpose but they were not immutable; they were loaded with transient values during the though-process of the ghost.pl AGI. So now let us go back into EnPrep() and code the fetching of the direct object of the preposition. We did so, and it worked the first time. We had the following conversation with the ghost.pl AGI Mind.

Human: john writes books for money
Ghost: THE STUDENTS READ THE BOOKS

Human:
Ghost: I AM AN ANDRU

Human: what does john write
Ghost: THE JOHN WRITES THE BOOKS FOR THE MONEY.

We should explain that the ghost.pl AGI knows only that “students read books”, not John’s books in particular. Mentioning books to the AGI causes it to recall its knowledge that “students read books”. When we query the AGI with the input of “what does john write”, the SpreadAct() spreading-activation module inhibits the interrogative prounoun “what” while activating the concepts of “john” and “write”. The response embedded in conceptual memory includes the linkage from the concept of “books” to the prepositional phrase “for money”. The EnArticle() module for the English articles “a” and “the” inserts articles somewhat haphazardly within the output of the ghost.pl AGI.

The new AI functionality of a machine intelligence thinking and conversing with prepositional phrases became possible when we recently expanded the conceptual flag-panel from fifteen associative tags to twenty-one associative tags, including new flags for the control of noun-declensions in Latin or Russian and for thinking with such parts of speech as adjectives, adverbs, conjunctions and prepositions. As we build up the ability to think with these linguistic components, each mid-AGI Mind becomes capable of more and more complex or complicated thought. As we make progress on the AGI RoadMap towards Artificial General Intelligence, we approach a point where Darwinian survival of the fittest comes into play, because among multiple enterprises working on AGI, some will go down the right path and some will enter roads where all hope must be abandoned.