//MEMORY SYSTEMS
//-----------------------------------------------------------------------------------------------------------------------------------
//1) language symbols (guesture&movement and spoken&sfx.) -fell swoop the recognition system in a dumbo way.
//2) rotationally invarient adjacent patch memory. (this has to learn only the front sides of things - but it
// should slice in as it rotates around something if u use a little concurrent memory.)
//3) patch connector, tree id system. - and this also stores the postures, of all the letters of its guesture memory.
//4) language- fixed state predictor! (so this is the big daddy, and its an overlapped combinationally invarient binary tree with substitution symbols.)
//this is a huge load of large key virtual memory- this is where the brain could get larger.
//5) state model, isnt really a memory, but it is so many fixed states.
// they are actually code tho, and they take measurements off the scene to activate.
void roomba_talks_gpu(ROSS& ross) //we had to take a little diversion but now weve got a talking one.
{
//REDY 2 GO!!! SOON!!!!
//put all comments, then ill back it up.
//but u should know it like the design on your hand.
//ACTUAL CODE FOR 3D->
//get multipled vh sobel
//get dynamic list from suppression
//form photo keys of dynamic list
//match history frame to now
//triangulate over all camera offsets 1 rigid body at a time (but only do 1 body, and see if the sift keys give you the rest) -BUT FOV MUST BE CONSTANT.
//get the zbuffer, train the sift descriptors - only on discovered area.
//form screen sift keys (very similar to the suppression for the corners) - that is the scale of the blur we are going to learn.
//match sift keys- and develop dense zbuffer.
//-------------------------------------------------------------------------------------------------------------------------------------------------------
//TREE DISCOVERER
//GET THE SCREEN GRAIN ANGLE -Average the vertices
//REDUCE THE X STRETCH - REDUCE THE Y STRETCH. and produce stretch, using the density of the lines. but xy separate.
//THEN U ROTATE IT BACK BY THESE ANGLES.
//THEN YOU PUT THE ANGLES IN THE KEY.
//then you near match the angles - but how do i tell what point goes on what point? - it should match up to it in nearest proximity to each point.
//SO THE PATCH JUST HAS ANGLES- THEN YOU REPROJECT IT, (using the angle difference of the key coming into it.)
//AND THEN U GET CLOSEST POINTS.
//THEN U HAVE TO INVENT ATTACHMENT TO ITSELF. AND SO IF IM ATTACHED TO SOMETHING ONE DELETES ITS SYMBOL AND THE OTHER ONE GETS IT.
//BUT THATS ONLY WHATS LEARNING ON TOP OF THE ADJACENCY - SO I USE IT LIKE A SHARED DISTRIBUTION - AND ITS ON THIS DISTRIBUTION I SPREAD THE ID.
//YOU ALSO CAN SYNCH WHAT IT LAST SAW TO THE NEW ANIMATION. (and that would require alot less memory cells.)
//THEN YOU GET THE ROTATION DIFFERENCE FROM 1 SUBJECT FRAME - AND THEN U KNOW WHAT THE JOINTS ARE.
//-------------------------------------------------------------------------------------------------------------------------------------------------------
//LEARN ITS LANGUAGE SYMBOLS!!
//GET LIST OF LANGUAGE SYMBOLS- Also, these are trees for the body language. (SO! that means i need the trees!!)
//LANGUAGE SYSTEM - first thing you do, is it has to form nice distant neighbours, to convert to its brain language.
//THIS IS THE EXPERT SYSTEM - COMPUTER VISION CONNECTION.
//it needs to be able to detect all its states no???
//THEN THE MOTIVE DETECTOR HAPPENS - AND ITS LOOKING AT HOW THE TREES ARE TOGETHER,
//KEEP A HISTORY OF LANGUAGE SYMBOLS - A SCENE HISTORY- SO IT CAN PREDICT IN STEPS IN FRONT.
//----------------------------------------------------------------------------------------------------------------------------------------------------
//GET THE LANGUAGE SYMBOLS - AND ROUTE IT TO THE DETECTED MOTIVE
//but does the whole scene cause the whole scene, i dont know-- its the machines mind, its like a virus, except not as hard to come up with.
//WHAT DATA DO U WANT TO SIMULATE??? - its not allowed to be the full flush activity.
// (you can watch him play himself on the computer- after hes gone through all the physics.)
//-------------------------------------------------------------------------------------------------------------------------------------------------------
//THEN ITS JUST SIMULATION CRAP
//get your jump increaser with the substitution. (maybe people will like terrorcom, u dont know. because it definitely sounds a certain way.)
//this happens to be pretty computational, because its only about as fast as its simulation speed, so it has to be very quick, because this
//is more computational than simulating, if you want it to go quickly. -THIS GIVES IT MORE ATTENTION SPAN FOR LANGUAGE, its possibly
// trading shit for manure, but we need to see does it make any difference?
//simulation - pick best thing to say, and best way to move.
//so its the minamax system, and if the computer is constantly predicting the worst, it can always steer its way out of it,
//but it may be an issue, but it might not be, but it can control it to some degree, so just go and dont worry about it.
//so you build something aesthetic so you can get a more human view of what was happening, but the robot works with something even simpler.
//its a super game! after this im quitting programming.
}