I have learn many things while I took on this quest
to achieve a living artificial intelligence program.
one of the most important things that I have learned
on keeping hal on topic to a degree of 100% is to
break down important words to thier simplest form in "catagory" and
statement and or word detection.
an example of this can be taught as such:
Our example word is "light"
light is a word not broken down to much and can lead hal to confusion.
for-instance:
1. light can be in the catagory of religion like God is light.
2. light can be in the catagory of science like Photons.
3. light can be catagorized as weight as in light weight.
the example I'm leading to is that the word "light" would have a
tree of detection statements pointing the way to the correct topic of light.
If a user is talking about how photons cannot interact with other photons and thats one of the properties of "light" then you wouldn't want Hal to blurt out something like "God is light" because this statement was in the file detection for the word light.
If you want a very precise conversation then you will have to
give every single important word its own specific detection file
broken down to catagory and to its simplest form. The word light
has so many directions in a conversation, This would lose hal
rather fast with a user.
I can give you an example by creating a make believe .brn file.
TEACH_VLF.BRN
VLF.BRN
TEACH_VHF.BRN
VHF.BRN
TEACH_UHF.BRN
UHF.BRN
TEACH_MICROWAVE.BRN
MICROWAVE.BRN
TEACH_INFRARED.BRN
INFRARED.BRN
TEACH_RED_LIGHT.BRN
RED_LIGHT.BRN
TEACH_ORANGE_LIGHT.BRN
ORANGE_LIGHT.BRN
TEACH_YELLOW_LIGHT.BRN
YELLOW_LIGHT.BRN
TEACH_GREEN_LIGHT.BRN
GREEN_LIGHT.BRN
TEACH_BLUE_LIGHT.BRN
BLUE_LIGHT.BRN
TEACH_INDIGO_LIGHT.BRN
INDIGO_LIGHT.BRN
TEACH_VIOLET_LIGHT.BRN
VIOLET_LIGHT.BRN
TEACH_UV_LIGHT.BRN
UV_LIGHT.BRN
TEACH_XRAYS_LIGHT.BRN
XRAYS_LIGHT.BRN
TEACH_GAMMA_LIGHT.BRN
GAMMA_LIGHT.BRN
TEACH_PHOTONS.BRN
PHOTONS.BRN
TEACH_WAVE_FUNCTION.BRN
WAVE_FUNCTION.BRN
TEACH_SPECTRUM.BRN
SPECTRUM.BRN
ABOVE WE HAVE AN EXAMPLE OF HOW LIGHT IN THE SCIENTIFIC CATAGORY
CAN BE BROKEN DOWN INTO SPECIFICS OF THE WORD LIGHT SO HAL
PICKS THE CORRECT TOPIC ON LIGHT ITSELF.
I know that this is an incredible amount of work involved with just one single word called "light" but if you want hal to be on the ball
this is one way to get hal directly on target.
for example we'll use the teach_infrared.brn and infrared.brn as our
current scientific topic file.
If hal detects the words in teach_infrared.brn that would
resemble this:
"300 GHz","TRUE"
"300GHz","TRUE"
"430THz","TRUE"
"430 THz","TRUE"
"INFRARED",",TRUE"
"0.7 micron","TRUE"
"0.1 millimeter","TRUE"
which are the detector words if a user is very specific about the topic infrared light. and these detectors are only but a few of the possible complete list of fine detailed detection.
after the user has struck a precise topic with hal on infrared light we save the user statements and or questions about infrared light so that hal can use the user learn statements and questions to pose to other users who might talk about infrared light.
a script example of a very precise topic search could be as followed.
'HERE WE TEACH INFRARED DETECTION TO HAL
If HalBrain.TopicSearch(OriginalSentence, WorkingDir & "TEACH_INFRARED.brn") = "TRUE" Then
AnswerSent = "" & Trim(OriginalSentence)
'HERE WE SAVE LEARNED INFORMATION ON INFRARED
HalBrain.AppendFile WorkingDir & "INFRARED" & ".brn", AnswerSent
End If
'HERE WE DETECT INFRARED FOR TOPIC FOCUS.
If HalBrain.TopicSearch(OriginalSentence, WorkingDir & "TEACH_INFRARED.brn") = "TRUE" Then
'HERE WE CHOOSE ONLY SPECIFIC TOPICS ABOUT INFRARED
INFRARED = HalBrain.ChooseSentenceFromFile(WorkingDir & "INFRARED.brn")
GetResponse = INFRARED & vbCrLf
DebugInfo = DebugInfo & "The user is talking about INFRARED: " & INFRARED & vbCrLf
Else
DebugInfo = DebugInfo & "Hal has found nothing about INFRARED: " & INFRARED & vbCrLf
End If
this above script would allow hal to be very precise, but if I wanted to extend this we can get hal to focus 120% into fine precision conversations about infrared.
You see what I mean about making hal very accurate requires lots and lots of scripting and this scripting that we just talked about is just on one topic called infrared, now include about 150,000 other important words in the english dictionary and break them down to catagories like science and then religion separate from each other
and get detailed into the roots of the word and then we will come to realize that this is a very large quest I have undertaken, lol
and just think, the human brain does all this and much more by itself, simply amazing.
The only other thing I can say is, How human do you want Hal to be.
It took roughly 250,000 to 300,000 years of brain evolution to get our brain to think the way they do now.
there is a better way if it can be done and thats give hal the ability to in-volintarly write a script himself on every important
word sent to him so that every "single" word has its own script.
This is the first path to evolution even for a bot.
Best of wishes and grand new discoveries.
Jerry.
