Wednesday, May 19, 2004

TIGGR & MIKE (AI): Artificial Intelligence - Solved!...Well Maybe. {Holographic Correlational Opponent Theory}


AFTER A FEW BRIEF CHATS AND 2 LOOOONG NITES, I have decide that a little Discussion Mike and I were having on Consciousness is worth further investigating...maybe we have it licked - maybe AI is easy!?.....

Read on for our continued discussion on Correlational Opponent Theory.

>> Damn you and your damn theory!

*teehee*...this is a topic that my brain has been chewing on for about 10 years....glad to see it is causing someone ELSE to "lose some sleep"

>>Ever since you told me you believe that each person simply "reacts" to his/her surroundings, I've been watching people, and thinking hard for some evidence it's not true. And so far, I haven't found it - But - I still haven't found any evidence to disprove my preferable theory that we are unique, blah blah....

AHH Yes, and that is the beauty of this "Little" theory - I have not been able to debunk it so far..however nor have I really been able to PROVE it "beyond a shadow"...thats what the whole Co-OpTheory Worksheet is for - trying to formalize the idea and get any wrinkles in the logic ironed out with some factual (albeit mathematical) data.

>>Unique in appearance - there is no question there. Unique in mind, well - yes - I believe that (tho I'm still considering the development of the mind [in relation to experiences] as people grow from being a little toddler) - unique in behaviour - well, I would say yes also, if I didn't see so many people behaving in extremely predictable ways every day. I could argue that one's determination to think for themselves creates individuality and uniqueness....But the rebuttal to that would be that throughout the course of that person's life, they have been exposed to situations, or attitudes that have given them the attitude towards thinking.

Well, this is a much more complex thought process than initially it seems....MY Argument is that human behaviour is a systematic respose to the current input stimuli that the Human receives COMBINED with ALL of the responses to ALL PREVIOUS inputs to the system....In this way, the "Learned", "Programmed" or "Aquired" behavioural mechanisms that humans develop are essentially the "Least amount of change required" to make their output (actions/Behaviours) EQUALLY OPPOSE the combined inputs (and learned adjustments to those inputs from the past)..our basic motivation then becomes Laziness.."What is the least amount of work (output/action I have to take) I have to do in order to balance this equation:

Current Inputs + Behaviour = Sum of all Inputs and outputs provided to the system over time

_IF_ perchance, my past actions created an internal "Engram" that means NEXT time the same scenario comes around I don't see it as being "out of the ordinary" or interesting (i.e If the value of my brainwaves PLUS the input scenario is CLOSE to what I remember) then the input can be considered fully learned...How then, can we "LEARN" behaviour's in such a way that we exhibit "Unique" personal behaviours? Easy...when we encounter the input stimuli, something ELSE (other than the INPUT and our OUTPUT) "unbalances" the system...for instance, if a parent ADD's input the a baby's "Processor" during and after the presentation of an input then the INPUT will NOT equal the memory encoded in the baby's brain and therefore the scenario becomes "Interesting" to the baby and the baby's mind encodes the difference between the input and the expected value back into the system so that next time the memory is closer to the inputs...

Which leads to your next comment really well:

>>Raising a child will provide some decent evidence one way or another towards this topic - how they learn, what they learn for themselves and why - for example - I have a cousin who is a couple of years old - he can open and close sliding glass doors, and understands that getting his fingers jammed would be bad - yet he has never jammed his fingers..... (I don't know if this is something his parents tried to teach him...)

YES..Spot on!...{maybe Kat and I should get onto the baby-making-bandwagon and "Build" ourselves our very own Human-AI experiment *LOL*}...i am actually looking forwad to Fatherhood partly for this (not exact) reason..i think I understand the complexity of the layering of learned experience enough to be a "Useful" father...also though, because the system is LOGARITHMICALLY complex (i.e. a previous learned response cannot be undone, it must be "learned out" and the "Whole of Brain" experiece is fully encoded - Every input and response EVER EXPERIENCED is incoded in the brain engram all at the same time)..i am aware that it is a pretty oenerous task and MY Mistakes compound once they are learned by my child......

The Sliding door conundrum.....not so hard to explain....TWO bits to it:

Going back to ..."Least amount of change required" to make their output (actions/Behaviours) EQUALLY OPPOSE the combined inputs"... If the inputs to this "feedback loop" include metrics such as current amount of pain, current amount of force being exerted as WELL as visual cues such as the location of the child's hand in relation to other objects, velocity of the door etc... then this whole "least amount of resistance" theory covers this learned behaviour by trying to balance the _potential_ for pain (or bad experience) with the current siutation...However, The child must have been made aware of their own mortality for this all to work. At some point the input to the system (Pain, blood, shouting at by mum and dad) increased above the "remembered" value and caused a learning experience....this learning experience then gets "reused" in different scenarios..HOW?...well you see this whole process exhibits a sort of INERTIA....like this: Child puts hand near something (other than a door) MUM or DAD yells at them to stop (protecting them from hurting themselves) and the Input stuation now does not match what the child expects (an INTERESTING situation occurs) so the child learns to expect shouting when their hand nears something...next time mum and dad DON't yell (situation is not dangerous)..this in turn Becomes an INTERESTING situation because the child was expecting to be shouted at when their hand goes near something but weren't (and, as such their inputs don't match their memory)....they Un-Learn the hand/shout scenario a little bit...this naturally encodes the "Middle ground" between the two scenarios and, unknowingly, has encoded the same situation with varying inputs as a Safe/Not Safe engram....As soon as the child goes to touch something and doesn't get shouted at, they LEARN the situation to be safe (because there was no shouting and the inputs _still_ do not exactly match the memory). In this way, although MUM & DAD did not "Teach" the "Doors can hurt your hand" Meme to the child, they have learned the process for evaulating safe stuations because they were told about an unsafe one (I think that all makes sense)....THIS DOES NOT WORK IF THE CHILD HAS NOT FELT PAIN! - if the child has not felt pain then there is no difference between the inputs of:

* the child's hand touching a HOT item that could hurt them (or sharp-hurty door)

* the child's hand touching a soft cushion that could not

If there is no difference in these two inputs from the child's point of view then there could be no learning...thus we can prove that in order for this theory to work, a feedback allowing data about the Host/Human/Child/Robot needs to be included as an input to the system...which is ok because this is exactly what humans have - 5 RAW sensory inputs (Smell, touch, taste, sight sound) plus multiple internal inputs (heart rate, temperature, pain, emotive response, hunger etc).

Secondly (an briefly coz this is oficcially a long-winded email now), various regions of the brain are used for differnet tasks and trained specifically for those tasks accordingly..the various different areas of the brain are then "Joined" together (at the Mendula Oblongata) and the total output is sent down the spinal cord to complete the feedback loop (action is taken, the response from the action is the next input value..e.g turn your head, the visual [and audio] input changes, potentially producing an output). Over time, (and continuously over our life), sections of the brain become increasingly and slowly diconnected from each other.... At a time near conception, our brains are almost "fully connectionist" that is; each neuron is (almost) connected to every other neuron in the brain, therefore the brain is completely aware of the other bits of the brain....as time goes by (rapidly when young, slowly as we get older) connections between nerons are "dropped" this results in sections of the brain becoming diconnected from other sections at least partially. The effect of this is that sections of the brain that are "Trained" to respond to certain inputs no longer have the "Complete" view of the world that they used to and are required to provide an output based upon on the "partial picture" of the world that it has....when many sections make "Guesses" like this, it results in novel reponses that still are applicable to the scenario they are in response to but based upon incomplete or inaccurate data and, therefore NOT QUITE RIGHT" which we see as being "Creative" or "Novel" thought....this is kinda like the "Tower Of Babel" story of how "God" introduced the Languagfe Barrier (Different languages did not exist in th ebible before this time) overnight in order to stop humans building a tower all the way to heaven...the brain, after some initial training in the "one big same language"...suddenly starts speaking different dialects in different areas and cant fully understand the dialect of other areas...this results in a kinda neural "chinese whispers" which we see as Creativity or uniqueness - hrmm maybe this is a good name for creative thought (Neural Chinese whispers)...

(Also links between neurons that have no direct link to the outside world are created over time in the brain ...that is, one part of the brain may take its input from the output of another part of the brain...so these decision making links are heirarchical not flat!)

Soooo, from 3 simple theorys (Co-Op theory + Neural Learning + Disconnectionist evolution) we get intelligence and creative thought....

The next question is, how does this apply to evolution?

Could, over time, our physiology change so that certain areas of the brain are pre-connected (or pre-disconnected) from each other to produce a more efficent brain? - is this what happened between Homo Sapiens (neanderthal man) and Homo Sapien-Sapiens (Human man)??

Could, over time, the "Driver Signal" used to start this "wave theory thing I have going" (in this case the 28Hz driver signal on the Excel worksheet) be customized to pre-contain data on it rather than being a blank driver signal...could, perhaps, we be copied with the waveform (or part of) that our parent/mother has and, therefore, could we be born with A Priori knowledge of the world?...the answer would seem at first look to be YES - after all, we are exactly the same biological species as our ancestors but we are "Smarter" or seem to be because we "Know more about more"..we "remember" what our ancestors learned to some degree...now, sure, some of this is because we have ancestors alive to change our inputs and therfore condition our responses as well as other artifacts such as the technology available today as opposed to before which change our inputs compared to those of our ancestors and, thus change our behaviours BUT, how can WE (our generation) learn everything our previous generations knew and then use that knowledge to "Create" ne knowledge and then continue the cycle?...surely this is only possible if, every generation, we Pre-Prime the driver waveform of the brain with some knowledge (that is, it is not a simple Sine wave but a complex wavelett similar to that left over after a waveform has already "Learned" something)???

 

Ahhh..well I would say you have just touched the tip of a 10-years-worth-of-thought iceburg...maybe I should finish this email here and continue my rant at another time....

;)

Hey, By the way, what time u knocking of today?...chances of a lift?

Lastly, I am enjoying discussing this with someone (for once) so much I have created an online blog for it..you will get an invite email soon, we can continue this thread on there (and, thus refer our notes or any links we get) from home.

The Blog will be at:

http://bitkraft.blogspot.com/

My IT Blog is at

http://tiggrbitz.blogspot.com

Kat and I have a personal one at

http://tiggrlyfe.blogspot.com tho it is blank at the mo

Cheers.

T.


No comments: