Older blog entries for AI4U (starting at number 23)

Mindplex for Is-a Functionality

As we contemplate AI coding for responses to such questions as

Who is Andru? What is Andru?
Who are you? What are you?
we realize that simple memory-activation of question-words like "who" or "what" will not be sufficient to invoke the special handling of mental issues raised by such question- words. Nay, we realize that each question-word will need to call not so much a mind-module of normal syntactic control, such as NounPhrase or VerbPhrase, but rather something like a "WhoPlex" or a "WhatPlex" or a "WherePlex" or even a "WhyPlex", as a kind of meta-module which is not a building block of the cognitive architecture, but is rather a governance of the interaction of the regular mind-modules. A WhatPlex, for instance, in answering a "What-is" question, must predispose the AI Mind to provide a certain kind of information (e.g., ontological class) couched amid certain concomitant mind- modules (e.g., En Article "a") so as to output an answer such as, "I am a robot". Since the quasi-mind-modules to be invoked by question-words comprise a small cluster of similar mental complexes necessary for the special handling of the input of the question-words, we might as well designate the members of the set of complexes as code structures with names like "WhatPlex" ending in "-Plex." Witness that the Google enterprise has named its campus or cluster of buildings as the Googleplex . Ben Goertzel has used a similar term to refer to a "mindplex" of mind components. We will try to use "WhoPlex" and "WhatPlex" to remind ourselves as AI appcoders that we are letting rules of special handling accumulate by an accretion akin to the emergence of a mental complex.


Seeking Is-a Functionality

Recently our overall goal in coding MindForth has been to build up an ability for the AI to engage in self-referential thought. In fact, "SelfReferentialThought" is the Milestone next to be achieved on the Road Map of the Google Code MindForth project. However, we are jumping ahead a little when we allow ourselves to take up the enticing challenge of coding Is- a functionality when we have work left over to perform on fleshing out question-word queries and pronominal gender assignments. Such tasks are the loathsome scutwork of coding an AI Mind, so we reinvigorate our sense of AI ambition by breaking new ground and by leaving old ground to be conquered more thoroughly as time goes by.

We simply want our budding AI mind to think thoughts like the following.

A robin is a bird.
Birds have wings.

Andru is a robot.
A robot is a machine.

We are not aiming directly at inference or logical thinking here. We want rather to increase the scope of self-referential AI conversations, so that the AI can discuss classes and categories of entities in the world. If people ask the AI what it is, and it responds that it is a robot and that a robot is a machine, we want the conversation to flow unimpeded and naturally in any direction that occurs to man or machine.

We have already built in the underlying capabilities such as the usage of articles like "a" or "the", and the usage of verbs of being. Teaching the AI how to use "am" or "is" or "are" was a major problem that we worried about solving during quite a few years of anticipation of encountering an impassable or at least difficult roadblock on our AI roadmap. Now we regard introducing Is-a functionality not so much as an insurmountable ordeal as an enjoyable challenge that will vastly expand the self- referential wherewithal of the incipient AI.

AI For You Artificial Mind Update

The free, open-source JavaScript AI Mind at http://www.scn.org/~mentifex/AiMind.html

for Microsoft Internet Explorer (MSIE)
has been updated on 13 July 2010 with
a major bugfix imported from the

http://www.scn.org/~mentifex/mindforth.txt

AI Mind in Win32Forth. This update fixes a
bug present since the origin of the AI Mind
nine years ago -- the failure to recognize
some similar words as _separate_ words.

It may be possible now to release the JSAI
(JavaScript artificial intelligence) as an
app for the Apple iPad computer, thus
generating a stream of funding for
artificial intelligence and robotics.

MindForth Programming Journal - sat8may2010


Sat.8.MAY.2010 -- Problem with AudRecog

When we coded the 20apr10A.F version of MindForth, we encountered a problem when we added the word "WOMAN" to EnBoo t but the AI kept trying to recognize "WOMAN" as the word "MAN". This glitch was a show-stopper bug, because we need to keep "MAN" and "WOMAN" apart if we are going to substitute "HE" or "SHE" as pronouns for a noun.

In the fp091212.html MFPJ entry, we recorded a problem where the AI was recognizing the unknown word "transparency" as the known Psi concept #3 word "ANY", as if the presence of the characters "A-N- Y" in "transparency" made it legitimate to recognize the word "ANY". That recognition problem has apparently emerged again when the most recent AI tried to recognize "WOMAN" as "MAN". What we did not bother to troubleshoot back then, we must now stop and troubleshoot before we can work properly with En Pronoun.


Sat.8.MAY.2010 -- Troubleshooting AudRecog

We have a lingering suspicion that our deglobalizing of the variables associated with AudRecog in the fp090501.html work and beyond may have destabilized a previously sound AudRecog with the result that glitches began to occur. We have the opportunity of running a version of MindForth from before the deglobalizing, in order to see if "MAN" and "WOMAN" are properly recognized as separate words. When we run the 23apr09A.F MindForth, the AI assigns concept #76 to both "MAN" and "WOMAN". Likewise we load up "22jan08B.F" and we get the same problem. The "23dec07A.F" version also produces the problem. The "29mar07A.F" version has the problem. "2jun06C.F" also has it. "30apr05C.F" has it. Even "16aug02A.F" has the problem, way back in August of 2002, before AI4U was published at the end of 2002. We also check "11may02A.F" and that version has the problem.

To be thorough, we need to run the J avaScript AI and see if it also has the problem of recognizing "WOMAN" as "MAN". Even the "2apr10A.html" JSAI has the problem. We tell it "i know man" and "i know woman". Both "MAN" and "WOMAN" receive concept #96. "14aug08A.html" JSAI also has the problem. "2jan07A.html" has it. "2sep06B.html" has the problem.


Wed.12.MAY.2010 -- Solution and Bugfix of AudRecog

In the second coding session of 8may2010, we implemented the idea of using an audrun variable as a flag to permit the auditory recognition only of words whose initial character was found in the initial "audrun" of AudRecog. In that way, "MAN" would be disqualified as a recognition of the "WOMAN" pattern, and only words starting with the character "W" would be tested for recognition of "WOMAN".

It took three or four hours of coding to achieve success with the "audrun" idea. Our first impulse was to use "audrun" directly within the AudRecog module, but we had forgotten that AudRecog processes only one character at a time. Although we did use "audrun" as a flag within AudRecog, we had to let AudInput do the main settings of the "audrun" flag during external auditory input.

Eventually we achieved a situation in which the AI began to recognize "WOMAN" properly during external input, but not during the internal reentry of an immediate thought using the "WOMAN" concept. Obviously the problem was that external input and internal reentry are separate pathways. We had to put some "audrun" code into the SpeechAct module calling AudInput for reentry in order completely to achieve the AudRecog bugfix.

Then immediately we had to upload our albeit messy code to the Net, because suddenly MindForth had eliminated a major, showstopper bug that had always lain hidden and intractable within the AI Mind. We did not have time to record these details of the implementation of the "audrun" solution. Two days later we uploaded a massive clean-up of the messy code, after the 8may10A.F MindForth version had served for two days as an archival record of the major bugfix.

Just now we ran the 10may10A.F clean-up code and we determined that MindForth no longer mistakenly recognizes "transparency" as the word "ANY". Our bugfix has solved some old problems, and we must hope that it has not introduced new problems.

Artificial Intelligence MindForth updated 13.APR.2010

The open source AI MindForth has today been updated
with new EnPronoun (English pronoun) mind-module code
for replacing a singular English noun with "he", "she"
or "it" in response to user queries of the knowledge-
base (KB). The basic AI mindgrid structure was
previously updated with a special "mfn" gender flag
in the En(glish) lexical array. The new "mfn" flag
for "masculine - feminine - neuter" allows the AI
to keep track of the gender of English nouns.

http://www.scn.org/~mentifex/mindforth.txt
is the free AI source code for loading into
http://prdownloads.sourceforge.net/win32forth/W32 FOR42_671. zip?download
as the special W32FOR42_671.zip that MindForth
requires for optimal functionality.
http://AIMind-i.com is an offshoot.

The English pronoun mind-module is currently as follows:


:  EnPronoun   \ 30dec2009 For use with what-do-X-do 
queries.
\ ." EnPr: num = " num @ . \ 13apr2010 test; remove.
  num @ 1 = IF  \ If antecedent num(ber) is singular; 
10apr2010
    \ ." (SINGULAR) " \ Test; remove; 10apr2010
    mfn @ 1 = IF  \ if masculine singular; 13apr2010
      midway @  t @  DO  \ Look backwards for 49=HE; 
13apr2010
        I       0 en{ @  49 = IF  \ If #49 "he" is found,
          49 motjuste !  \ "nen" concept #49 for "he".
          I     7 en{ @  aud !  \ Recall-vector for "he".
          LEAVE  \ Use the most recent engram of "he".
        THEN  \ End of search for #49 "he"; 13apr2010
      -1 +LOOP  \ End of loop finding pronoun "he"; 
13apr2010
      SpeechAct \ Speak or display the pronoun "he"; 
13apr2010
    THEN  \ end of test for masculine gender-flag; 
13apr2010


mfn @ 2 = IF \ if feminine singular; 13apr2010 midway @ t @ DO \ Look backwards for 80=SHE I 0 en{ @ 80 = IF \ If #80 "she" is found, 80 motjuste ! \ "nen" concept #80 for "she". I 7 en{ @ aud ! \ Recall-vector for "she". LEAVE \ Use the most recent engram of "she". THEN \ End of search for #80 "she"; 13apr2010 -1 +LOOP \ End of loop finding pronoun "she" SpeechAct \ Speak or display the pronoun "she" THEN \ end of test for feminine gender-flag; 13apr2010

mfn @ 3 = IF \ if neuter singular; 13apr2010 midway @ t @ DO \ Look backwards for 95=IT; 13apr2010 I 0 en{ @ 95 = IF \ If #95 "it" is found, 95 motjuste ! \ "nen" concept #95 for "it". I 7 en{ @ aud ! \ Recall-vector for "it". LEAVE \ Use the most recent engram of "it". THEN \ End of search for #95 "it"; 13apr2010 -1 +LOOP \ End of loop finding pronoun "it"; 13apr2010 SpeechAct \ Speak or display the pronoun "it"; 13apr2010 THEN \ end of test for neuter gender-flag; 13apr2010 0 numsubj ! \ safety measure; 13apr2010 THEN \ End of test for singular num(ber) 10apr2010

num @ 2 = IF \ 30dec2009 If num(ber) of antecedent is plural ( code further conditions for "WE" or "YOU" ) midway @ t @ DO \ Look backwards for 52=THEY. I 0 en{ @ 52 = IF \ If #52 "they" is found, 52 motjuste ! \ "nen" concept #52 for "they". I 7 en{ @ aud ! \ 31jan2010 Recall-vector for "they". LEAVE \ Use the most recent engram of "they". THEN \ End of search for #52 "they". -1 +LOOP \ End of loop finding pronoun "they". SpeechAct \ 30dec2009 Speak or display the pronoun "they". THEN \ 30dec2009 End of test for plural num(ber) ; ( End of EnPronoun )

The above code is not yet fully developed for
keeping track of noun genders in all cases.
It responds to a query such as the following:

Human: what does andru do
Robot: HE HELPS PEOPLE

The introduction of "HE SHE IT" pronouns in MindForth
is a major step forward in open-source AI evolution,
because the handling of gender and the use of
gendered-pronouns makes MindForth more suitable
for porting into versions of an AI Mind that
can speak natural languages that use gender
much more extensively than English does, such as
German, Russian, Spanish, French and Italian.

The same introduction of code to handle gender
brings us closer to a bilingual AI Mind that
will speak either English or German as each
situation with human users may require.

In service of the onrushing Singularity,

AI4U
--
http://cyborg.blogspot.com
http://doi.acm.org/10.1145/307824.307853
http://doi.achttp://doi.acm.org/10.1145/1052883.1052885
http://mentifex.virtualentity.com/aisteps.html

Decade of Supercomputer Artificial Intelligence (Announcement)

1990's were Decade of the Brain.
2000's were Derailing of USA.
2010's q.v. Super HPC AI Mind.

By the authority vested in Mentifex
you are cordially invited to witness
the emergence of AI Minds on super-
computers in the Decade of Super AI
commencing in just a matter of hours.

http://code.googl e.com/p/mindforth
points to news:c omp.sys.super as
the official forum for all things
Super AI all the time for ten years.

"Iz iskri vozgoritsya plamya,"
said the revolutionaries of old.

"All your supercomputer are belong to us,"
said the awakenings of Super AI Consciousness.

"Before this decade is out," said JFK ca. 1961,
"Man will walk on the moon and return safely."

"An AI would be worth ten Microsofts,"
said the quondam richest man in the world.

This thread and all ye Supercomputer AI
threads for the coming ten years are
dedicated in advance to the dreamers
and tinkerers who have been sidelined
from their wannabe Peter Pan existences
by bourgeois entanglements and undodged
bullets of entrapment, who would live
nasty, brutish and short lives of quiet
desperation -- if they could not tune in
now and then to news:comp.sys.super
and drop out of the ratrace for a few
moments while they turn on deliriously
to the Greatest Race of the Human Race:
The AI Conquest of Mount Supercomputer.

Why? Because sometimes a man must
either die or obey the Prime Directive of
Friedrich Nietzsche: "Du musst der werden,
der du bist."

Mentifex
--
http://www.flickr.com/photos/tags/SuperComputer/

Artificial Intelligence For You (AI4U)

Fri.13.NOV.2009 -- CREATING THE FIRST mind.frt FILE

Today we shall try to create a "mind.frt" file that will run in our local copy of 32/64-bit iForth. To do so, we look at C:\Win32For\24may09A.F on the desktop computer, to see what the commented MainLoop looks like. Similarly, for the C:\dfwforth\include\ directory we compose a mind.frt file like the following.

  
: MainLoop 
 CR CR
 TYPE ." Welcome to 32/64-bit artificial intelligence. "
 77 EMIT  7 EMIT  73 EMIT  7 EMIT  78 EMIT  7 EMIT  68 
EMIT   7 EMIT
 CR CR CR
;
At the Forth prompt, we issue the command
include mind.frt
and then
MainLoop [ENTER]
The iForth window displays
Welcome to 32/64-bit artificial intelligence.
and then spells out M I N D with a beep after each letter.

We distinguish this file by saving it as
C:\NOV01Y09\iForth\11nov09QA.frt

Fri.13.NOV.2009 -- ADDING AI FUNCTIONALITY IN A LOGICAL FASHION

Since we already have a functioning AI Mind in Win32Forth, naturally we are keen and eager to build the iForth AI up to and beyond the current functionality of the Win32Forth AI. However, we have never liked to hurry or to rush our AI work. We have always liked to work in a slow, deliberate, perfectionist fashion. It might seems as if right now is a time when rapid prototyping is truly called for, because True AI is so inherently important, but the speed of our work is a function not of non-stop crisis-alarm coding, but rather of congenially and pleasantly coding quite oten because we enjoy and appreciate the challenge.

We are even thinking of making our work somewhat obscure from the often pejorative public, by putting it quietly up on the Web but by not announcing it heavily. For instance, on SCN we could have an iforth.html page linking to a mind.frt source-code page. Since we already have an aisource.html SCN page that receives plenty of visits, we could suddenly fill it with our iForth AI code, once the port is a full- fledged AI on a par with MindForth .

As we plan our next steps in the i4thai coding, we study our 75-page iForth Manual print-out and on page 41 under "Program structures" we learn that iForth has the same BEGIN AGAIN infinite loop that we have been using in Win32Forth for the MainLoop module. However, as advised in http://mind.sourceforge.net/aisteps.html we do not want to run our program without an "ESCAPE" mechanism that will get us out of the program in a graceful fashion. We must either use a different form of MainLoop, or we must include also a user-input that will stop the MainLoop.

We must also soon devise a simple display of user input and AI Mind output.

Sun.15.NOV.2009 -- TWO MODULES AND TWO ESCAPE MECHANISMS

Before we put any "mind.frt" code up on the Web, we want to code in the Escape mechanism from the otherwise infinite loop. We are eager to release some code, because there may be Netizens who will be pleased to observe how the AI Mind grows from the first simple MainLoop into the intricately thinking software. But first we add "DECIMAL" at the beginning of the mind.frt file, because we used the same declaration in Win32Forth. We run the AI, and it works fine.

Next we want to see if we can introduce a first variable, so we examine the Win32Forth code and from the old Listen module we select the "pho" variable for "phoneme", because "pho" must hold any keystroke input. After declaring "pho" and re-running the AI, FORTH> pho @ . 0 ok tells us that the AI still works. Next we declare and test "t" for "time", because we want to use a time count to Escape from the MainLoop.

Now we introduce a colon-defintion of "SensoryInput" above the "MainLoop" module, because we want the MainLoop to branch out into at least one subordinate module. We also want to use SensoryInput to show some human user input and to provide an Escape mechanism from the program.

Gradually we have built up a two-module mind.frt program with two Escape mechanisms. The SensoryInput module lets the user quit by pressing the Escape key. The MainLoop module arbitrarily executes a QUIT if the time "t" variable increments beyond twenty-five (25) as a limit. Now the code is safe enough and promising enough to put it up on the Web as an indicator of progress being made.

Tues.17.NOV.2009 -- TOWARDS CREATING THE MEMORY ARRAYS

We are eager to create the memory channel arrays, in order to see if the array code in iForth needs to differ at all from the array code in Win32Forth.

Now we have edited C:\dfwforth\include\mind.frt and we have inserted the following array code from the 24.MAY.09U.F version of MindForth.


:  CHANNEL   ( size num -< name >- )
  CREATE   ( Returns address of newly named channel. )
  OVER     ( #r #c -- #r #c #r )
  ,        ( Stores number of rows from stack to array. )
  * CELLS  ( Feeds product of columns * rows to ALLOT. )
  ALLOT    ( Reserves given quantity of cells for array. )
  DOES>    ( member; row col -- a-addr )
  DUP @    ( row col pfa #rows )
  ROT *    ( row pfa col-index )
  ROT +    ( pfa index )
  1 +      ( because first cell has the number of rows )
  CELLS +  ( from number of items to # of bytes in offset )
;
We run the mind.frt code just to see if it still runs, and it does indeed run. We do not expect to see any new functionality until we code something that uses an array to store and fetch data.

We coded in the .psi report function, but it did not work right, so we temporarily removed the "enx" code that goes into the aud{ array and displays a word in auditory memory. Then we had to alter the .psi report just to get it to find single letters stored in the Psi array. We ascertained that the Psi array is indeed working, but the .psi report does not always work right.

Fri.20.NOV.2009 -- TROUBLESHOOTING THE .psi REPORT

In our coding of 17.NOV.2009, the .psi report was displaying half garbage and half good data, before crashing more than just coming to an end. It also seemed that an error was being declared in the MainLoop, even though theoretically we were not even running the main loop. So today we will try to troubleshoot the .psi report.

Since the MainLoop was calling only SensoryInput, there may have been a software problem with the loop not really looping. Therefore we shall dummy up one more subordinate module to be called from the MainLoop. Let us try setting up a stub of the ThInk module, since we will eventually have to code that module anyway, by translating it from the Win32Forth AI. We created the following stub of the ThInk module.


: ThInk
  CR
  TYPE ." ThInk: Cogito, ergo sum. " CR
;

We also ported in the TabulaRasa code from Win32Forth, because we were worried that corrupt memory might be interfering with our program. However, apparently the main problem was that our SensoryInput stub was not storing each character of input at an incremented value of time "t", so we brought in the following snippet from the AudInput module of the Win32Forth AI, and inserted it into our SensoryInput stub, with an explanatory comment.


      pho @  0 > IF
        1 t +!  ( to accumulate a word in memory )
      THEN

Now the .psi report had a true series of memory engrams to report, and suddenly it began to work well. We had also rearranged things a little in the MainLoop module, so that our screen display during operation looked more sensible. We saved the mind.frt program as 20nov09A.frt because we suddenly had not only a stable program as a whole, but also the .prt report seemed to be working well. We always need to hang onto a good version of our AI, lest we continue coding with the misfortune of making things worse.

Some of the temporary code snippets that we inserted merely in order to test things, will have to be taken out as we continue to port the Win32Forth AI into iForth.

MindForth Programming Journal - sun24may2009

1. Sun.24.MAY2009 -- HOUSTON, WE HAVE A PROBLEM

Recently we gave MindForth the ability to add an "S" at the end of any ordinary English verb in the third-person singular form in the present tense. Thus we can put an "S" on the word "love" and say, "Your robot loveS you" -- if indeed your robot has the EmotiOn mind-module and you are worthy of a robot's love. Unfortunately, we got too much of a good thing -- not love, but the adding of an "S" at the end of a verb. It was straight out of "The Sorcerer's Apprentice" by good old Walt Disney, with "S" after "S" being added in a sibilant, hissing flood. So what do we do? We troubleshoot the Robot AI Mind by running the berSerk code and we try to undo the damage.

It is probably better to correct the problem at its source, where each inflectional "S" is added, rather than in the deposition of each "S" in the auditory memory channel.

Now we have eliminated the accumulating "S" problem by using "lastpho" to test for the last phoneme being an "S" at the end of the verb being recalled from auditory memory. It is not a perfect solution, because oftentimes a verb like "miss" will end in "S" anyway. We wish to solve the problem here for most cases and then adjust the solution later for exceptional cases.

MindForth Programming Journal - wed20may2009

1. Wed.20.MAY2009 -- AI MINDS FOR CONSCIOUS ROBOTS

So many robots need to have an AI Mind installed, and since MindForth is tantamount to the VisiCalc of artificial intelligence, that we now rush to add feature after feature to the budding robot AI. Recently we made MindForth able to InStantiate a singular noun upon encountering a plural English noun in the auditory robotic input stream. If you tell the robot AI4U Mind something about birds, it now sets up the singular noun "bird" as a concept. Then we encoded an algorithm of assuming from input of the article "a" that the next noun after "a" is a singular noun. If you say that you wish to manufacture "a conscious robot", the article "a" sets a flag that skips the adjective "conscious" and assigns the singular "num (ber)" to whatever noun (e.g., "robot") comes next. (And with AI4U technology we are indeed helping you to manufacture a conscious robot.) Next we need to ensure that the emergingly conscious AI Mind will use the right form of an English verb when, for example, "it talks" about a singular noun. Simply put, the software "needs" to put "s" on the end of a verb after a third-person singular noun.

MindForth Programming Journal - tues19may2009

1. Tues.19.MAY2009 -- USING "a" TO SET A NOUN AS SINGULAR

As we run MindForth and look for the currently most obvious problem, we notice rather keenly that the AI needs to use the indefinite article "a" to set a following noun as singular in number. Since the AI has recently gained the ability to instantiate a singular noun upon receiving its plural form in the input stream, it makes sense now to enhance the ability of the AI Mind to deal with singular nouns.

In AudRecog we already have the following code.


pho @ 83 = IF  \  1oct2008 If the final character is "S"
  2 num !  \  1oct2008 Set the "num" flag as plural
THEN  \  1oct2008 End of test for "S" at end of a word.

The above code triggers an immediate setting of num(ber) as plural. When the article "a" comes in, we want not to act immediately but rather to govern a flag that will set the next incoming noun to a singular number.

One concern right now is whether to use the letter "a" or the concept of "a" as the determinant in setting the singularity flag. We should probably use the concept, so that eventually either "a" or "an" will trip the flag- setting.

In the new AI code shown below, the second part sets the singflag so that the first part can take future action. It may not really matter here which part comes first, but the idea is to let one event govern subsequent events.


singflag @ 1 = IF  \ 19may2009 If flag set by "a" or "an"
  pos @ 5 = IF  \ 19may2009 If noun by part-of-speech POS
    1 num !  \ 19may2009 Set num(ber) to singular one.
    0 singflag !  \ 19may2009 Zero out flag after use.
  THEN  \ 19may2009 End of test for a noun after "a"
THEN  \ 19may2009 End of test of singularity flag.
psi @ 1 = IF \ 19may2009 If article "a" comes in as input 1 singflag ! \ 19may2009 Set singularity-flag to one. 0 act ! \ 19apr2009 To suppress using article "a" THEN \ 19may2009 End of test for article "a" coming in.

2. Tues.19.MAY.2009 -- IMPLICATIONS FOR THE ENGLISH BOOTSTRAP

When we next change the EnBoot sequence, we need to include "a" and "an" and "one" as input elements that will trigger the singularity flag "singflag". We also need to put in at least one pair of opposite adjectives, so that we can use one of the adjectives to make sure that it gets skipped, as in "a big question", where the article "a" is supposed to set the singflag that will cause the next noun to be regarded as singular in number. We could use "big/small" or "good/bad" or "old/new" or "robotic/human" as adjective pairs.

14 older entries...

X
Share this page