Cyborg AI Minds are a true concept-based artificial intelligence with natural language understanding, simple at first and lacking robot embodiment, and expandable all the way to human-level intelligence and beyond. Privacy policy: Third parties advertising here may place and read cookies on your browser; and may use web beacons to collect information as a result of ads displayed here.

Monday, December 11, 2017

pmpj1211

How a Mind Maintainer improves a major mind-module.

For anyone interested in a high-tech career as an AI Mind Maintainer, here is a typical debugging session. We run the ghost248.pl AI initially without human input, to see if any mind-bugs manifest themselves. We know from recent coding sessions that there is a bug lurking in the retrieval of the ego-pronoun "701=I" under certain circumstances, but nothing goes especially wrong when we let ghost248.pl engage in meandering thoughts. A few irksome things happen, such as a tendency for the AI to repeat itself in a conjoined thought like "I AM PERSON AND I AM PERSON". We suspect that we must make searches for active ideas start from $krt (knowledge representation time before thinking) instead of current time $t so that the AI does not think a thought and then repeat the very same thought lodged in recent memory. But it is much more irksome that the ghost in the machine is not saying "I AM A PERSON", so we decide now to attempt some refinements in the function of the EnArticle() mind-module for inserting an English article ("a" or "the") before the output of a noun.

We start experimenting with EnArticle() by inserting into it a diagnostic message to tell us whether the module is being called, and what are the values of $subjnum (subject number) and $qv2psi -- the second item, or verb, in a sentence of thought being generated as a combination potentially consisting of subject (qv1), verb (qv2), indirect object (qv3), and direct object (qv4). We run the AI and we inspect for any diagnostic message just before the output of "I AM PERSON", but there is no message. Then we remember that we turned off calls to EnArticle() because it was saying "THE" too frequently. In order to turn it back on, we search the code for "EnArticle", find it, and reinstate the call from EnNounPhrase(). The program goes back to saying the word "THE" too many times, but we also see the values of the variables we are interested in. So we shall try to insert some sensible code.

In the mind-module for English articles, we copy the loop that finds the article "THE" in memory and then, mutatis mutandis, we get the code to find the article "A". We run the fledgling AI and eventually it says, "I HELP A THE KIDS AND KIDS MAKE A THE ROBOTS". The AI is inserting the indefinite article "A" and the definite article "THE". Before we get called on the carpet before the corporate board for not being a good Mind Maintainer, we hasten to pledge that we have been brainstorming some ideas to vastly improve the function of the article-inserting mind-module. We go back into the module and we cook up some code to insert the singular article "A" when the $subjnum is a unitary one ("1") and the $qv2psi verb is "800=BE", in order to cover cases of "I AM...." At the same time we comment out the block of code that inserts "THE" until we are ready to refine the code in a later mind-maintaining session. We run the AI without input and soon it says "I AM A PERSON", so maybe we get to keep our job as an AI Mind Maintainer. Wouldn't you like to have some business cards made up that say you are an "AI Mind Maintainer"? Oh no! The urge has struck. Are we going to be at a 2018 New Year's Eve party and have people say to us, "What on earth is an AI Mind Maintainer?"

We had better put this code and this journal entry up on the Web lickety-split, but first let us mention our plans for how the article-module shall decide algorithmically to insert the definite article "THE". We will create a cluster of variables that briefly hold onto any noun being mentioned in a train of thought or a conversation with a human user. When a particular noun is first mentioned, its identifier goes into the brief-tenure variable. If the same noun gets mentioned again quite soon, in one of the next few sentences, the software will insert "THE" before the new mention of the noun. A human user might say, "I have a book." Then the AI can say something like, "What is the book?", because the place-holder variable prompts the saying of "THE".

Over the past weekend we uploaded a new webpage named MindBoot Module Documentation for Strong AI Mind Maintainers as one of a dozen new uploads with a similar nomenclature emphasizing the career pathway for mind-maintainers. If your corporate information technology (IT) department does not have at least one Mind Maintainer on staff, then (imagine some dire prediction here :-).