Mentifex AI Minds are a true concept-based artificial general intelligence, simple at first and lacking robot embodiment, but 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.

Wednesday, May 30, 2018

jmpj0530

Solving who-query problems and EnParser bug.

Although the ghost.pl AI responds to a who-query by calling SpreadAct() from the end of AudInput(), the JSAI will call SpreadAct() too many times from AudInput() before the end of the input. Since we need to test for qucon when the Volition() module is not engaged in thinking, we test for quconin the Sensorium() module, which does not call AudInput() but which is called from the MainLoop() after each generation of a thought.

We must also troubleshoot why the JSAI eventually outputs "ME ME ME". We discover that EnNounPhrase() is sending an aud=726 into Speech() while there is a false verblock=727. Then we learn that the concept-row at the end of "ROBOTS NEED ME" for "ME" at t=727 has an unwarranted tkb psi13=727, as if the concept 701=I had a tkb. Apparently we need to prevent a false tkb from being stored. An inspection of the diagnostic display shows that the tkb properly set for each verb is improperly being retained and set for the object of the verb. We then notice that the EnParser() module is properly setting the time-of-direct-object "tdo" to be the tkb of a verb and leaving the tkb value set to the "tdo" value. So we insert into EnParser() a line of code to reset tkb immediately back to zero after storing the tkb of a verb, and the erroneous "ME ME ME" output no longer appears.