- summary Selection of verbs of being for inclusion in thought
- labels module
Function -- what the BeVerb module does
The BeVerb module is part of the MindExpansion beyond the ProofOfConcept AI, which could think only in the format of subject-verb-object (SVO). In the SVO syntax, a transitive verb mediates the relationship between a subject operating on a direct object. The KnowledgeBase of even a simple AI may contain numerous propositions stated in the SVO format. In the ProofOfConcept Forthmind, a chain of SVO ideas may cause a looping sequence of repetitive thoughts, such as the following.
Cats eat fish.
Fish eat bugs.
Bugs eat germs.
Germs kill cats.
Cats eat fish...
The performance of the AI in learning the above ideas and in thinking them in a potentially infinite loop was a useful tool in debugging the AI software and completing the ProofOfConcept phase.
In the MindExpansion phase of AI development, the VerbPhrase module must permit the AI Mind to think not only with transitive verbs of action but also with intransitive verbs of being. It is more difficult for an AI programmer to get the AI to think with verbs of being rather than with verbs of action, because it is hard to implement the factors and reasons for VerbPhrase to use a verb of being. Typically a human user may cause the AI to think with a verb of being by stating a fact or asking a question like, "Who are you?" Then it is a simple choice for the AI software to respond with a verb of being in a sentence like "I am Andru".
However, it is not so simple for the AI to get the correct verb-form right. Like the proverbial engineering student who said, "Last week I did not know what an engineer was; now I are one," the emerging AI mentality was often saying "I are Andru" instead of "I am Andru."
The problem was ameliorated by the coding of a minor intervention in the BeVerb module, where any attempt by the AI to use the wrong form of "to be" after a personal pronoun was thwarted by the software substitution of the correct form instead of the wrong form. Since there are only a few personal pronouns and only a few (but highly irregular) forms of "to be", it is not hard to intercept the linkage of the pronoun with the verb-form and force the thinking of the correct verb-form. However, the practice of not letting the AI use the first form it comes up with smacks of MindControl and ThoughtPolice. How is it justified for a MindMaker to intervene in the FreeWill of an AI to choose whatever VerbForm it feels like -- even a wrong one?
The answer is that the choice of a verb-form is a two-step process. In the first step, SpreadingActivation leads the software to select a verb such as "to be." In the second step, which is selecting the verb-form that will express an idea, it is legitimate for the MindMaker to code an intervention that steers the AI to the correct verb-form, because arriving first at the wrong verb-form in conceptual memory was nevertheless a successful step in the right direction. It only needed straightening out so that the AI would not think like the proverbial engineer but rather like a well-educated liberal arts major.