[ranch vermin]

bf->,  u said "match to memory"  thats what the neural network stuff i posted is doing.   and thats what that cam and binary tree i showed you do,  they take the sensor and match it to the motor.

U cant do a match from the sensor "retina" to memory, quick,  without it.    U can try,  but eventually youll work out im right.  That match a sensor to a memory is a very labourious task, if we are considering a full finished memory in operation -  but easily solved with a tree in software.

Korrelan ->  the reward is given to it when its "going faster"    u dont get what we mean by reward. its not the "mechanism of selection",  but  I guess reward is "being selected"   U select the behaviour of your children, by giving them rewards.   so the reward is the 32 bit integer score given to it.   and thats all that drives the robot.

That doesnt sound so totally correct to do,  but thats what BF is meaning here by reward.

[Korrelan]

Hi Ranch… nice vids of your 3D Sobels BTW.

BF33 seems to want/ enjoy people picking holes in his theory, so I was just obliging lol.

From what he describes I imagined a huge list or several huge lists.  Sensor data is saved as a stream into the list.  Each list entry is given a +/- score depending on how close it is matches (fuzzy logic?) to the target task.  The lists are sorted on the fly so the most used patterns move to the top.

So his bot moves randomly at first, and if the human gives it a + after a certain action, the commands that produced that action are saved into a list along with the accompanying sensor data produced by said action. The more times the bot selects this action given the same task the higher score it gets somehow making it more likely to selected next time.

At least I think that’s what he’s saying lol. BF33 is very hard to understand.

I was pointing out that besides the many flaws in the system the main hurdle is the shear amount of data it’s going to accumulate, an impossible amount.  He needs an algorithm, heuristics or hierarchical structure like you pointed out.  Simple lists won’t do it.

I have a special way of locating a match too - by same spike strength pathway it fires right to it!

I’m not sure what this means. Lol. 

Korrelan ->  the reward is given to it when its "going faster"

But how doe’s it select the required command/ actions to ‘go faster’ without some kind of reward or ranking system?

u dont get what we mean by reward. its not the "mechanism of selection",  but  I guess reward is "being selected"

Hmmm… but...  lol

U select the behaviour of your children, by giving them rewards.   so the reward is the 32 bit integer score given to it.   and thats all that drives the robot.

Perhaps it’s just all too complex for me to understand.

[keghn]

Once you have your engram pattern loops you can run them through slider neural networks.

A pointer step through the sequential data. All data behind is what you feed into slider NN. The data in front
is what you train it to.

[LOCKSUIT]

Similarly, just last night I was thinking about and going to tell you some very interesting things.



Before I say the incredible thing about searching memory+more, let me say again the following points:
>I am a machine.
>Our AI looks a lot like us, all of it, all I told you "matches up", even how it thinks(only my AI has this[I told you it above]).
>It gets better actions from baby to adult.
>Matching memory is matching memory, it can be done, scale/greyscale/rotate to name a few (you) know of.
>As said, memory is a growing bar, two rows, guesses are saved at bottom row, then senses above all linked+ranked&labbeled +/-, then it matches memory to a sense that's at least similar and does to strongest rank neuron one~ And is saved and done at mem bar front+senses linked come(after passing rewards again)(like remembering music).



Now, the way I have written down how our brains search memory *so fast *without a match-system *without rotate/scale/greyscale/etc. is --- 2 keys:
1. The searching memory sense finds a at least similar sense and finds the strongest on the first attempt - which holds the best associated actions/senses, by same strength pathways, and then strongest pathways, It must be true. Though I don't yet know how a neuron has 1000/more axons and when more storage is added one axon must link to another 1000 axons and so it's like the sense would have to pass a few thousand neurons but... plus how does it know worse is only over there...
2. Part of the above technique: Guess what, there is a fault with mainstream computer pixel color-code matching. All pixel colors including grey/color/brightness must change by a universal number like for example brighter means the number for a pixel should be higher, grey should be a similar number to any color since brightness similarity of the grey/color pixel is key here, in fact that sums it up I think. Only neural strength rises! The number goes from low to high depending on the brightness, a color image will find a even similar grey image of itself. As for seeing red/etc. difference we know we do(not same), that is like a smaller wedge inbetween what I said. This allows total better matching now.



My cool technology - I was thinking that like IBM's parallel neural synapse processor, I could emulate the instant-search-find-match in memory, even in sequential!!! By having all the parallel wires of the chip be programmable nodes (IBM's is) I can add hints around for the path to be chosen right to the match. Don't forget the aboves combined with convolutional neurons to help lots with vision matching. The first digital AI agent could even quickly change his algorithm then.



Keghn......The training is the real world....................actions&senses are initiated when confront things.........................algorithm is GMSTM as said above, the fields way sounds like datasheet stuff very messy.

[ranch vermin]

nice try, with super parallel IBM sytem, uve only GOT one computer to work on, for the time now.

So i told you how to make matching memory, ull just develop your own one, but im pretty sure mines faster than urs. hehehe,   just wait till you work it out.

But i dont know,  u may have figured it out yourself, virtual addressing is not very complicated -  of course, it does "connect straight to it"  but it still takes time even to read 1 byte of ram, make sure u get that in your head.  nothing is free.

So damn well get to work, and stop jabbering on.

U sound pretty intelligent as far as I can guess,  but theory is fun, implementation is like doing the dishes,  so no pain, no a.i.


And Korrelan -  That little bot was pretty interesting (Thanks for putting it together.),  a 2d world needs a 1d eye!!

And on this "reward" way of explaining a GA.    its running on a simplistic mechanism of knocking off its bad behaviour.   Im pretty sure you have to store all the bad behaviour as well as the mediocre behaviour (because u dont know if it was going to lead to good behaviour.) and the good behaviour.

so the GA does the crossover,  and avoids bad behaviour, and novelizes the good behaviour together.  The splicing system, as I see it,  is a replacement for a more powerful logic paradigm system, if you could ever think of it.

[LOCKSUIT]

Wait this is exciting, wait, first-
I'm 20, I have listed all AI abilities (very powerful) in The Everything I update, between the first two thick lines is AI, then the between the last two is all abilities (I'll show them later) (plus show later the whole AI text piece with da spaces see that)
http://s1318.photobucket.com/user/coolgodboss/media/The%20Everything_zpswq4iqrpy.png.html?sort=3&o=0[/url]
and in this section I have guess all possible actions (always new) and save and avoid which are bad n good~ Our brains don't do exactly that...It's not needed...(only our-bad rewards label senses well, bad, or if stop good rewards, instead of "if didn't help"). My robot AI I made only guessed by a random-block and when seen anything that looked like a far away wall it was able to match mem and picked the highest ranked and tweaked it a bit too each time (the tweak in mine wasn't even needed).

Ok now:

1. You mean I'm right about using and can use that IBM chip, this is awesome?...

2. I'm going to give a simple explanation about how I want you to explain your fast search system to me in a clear simple way. The sensory input are books, you place them on the shelf, when one book is similar to one book the most it goes beside it, what this organizing does is makes a slowly changing memory array, so that when the time comes a sense enters later - if it doesn't match to one book - it is sent far far away down the shelf of books. If not this is your technique .

3. What korrelan bot where where??

4. Did you see my AI robot improve at imitating the speakers voice? And how it crawls/guesses when ever it sees something that looks like a far/close wall?

[ranch vermin]

2. I'm going to give a simple explanation about how I want you to explain your fast search system to me in a clear simple way. The sensory input are books, you place them on the shelf, when one book is similar to one book the most it goes beside it, what this organizing does is makes a slowly changing memory array, so that when the time comes a sense enters later - if it doesn't match to one book - it is sent far far away down the shelf of books. If not this is your technique .

What u said then is what i call near neighbour classing, and i guess others do too,  u actually have to store the members of the class too! and theres alot more of those,  and they are what i call patterns, or keys.   You can actually get away with not storing the members by blurring them together, but its easier and not as powerful.

3. What korrelan bot where where??

Back up a bit and look for it,  its cool,  its in this very chat thread.

[LOCKSUIT]

Did you's know that we save vision and sound and they are linked together when saved and we can name a car car/chingwong/juka/etc.?




I KNOW IT'S A LITTLE MESSY/LOTTTTTTTS.......BUT YOU'RE GONNA HOARD IT YOURSELF IF YOU LOOK THROUGH, scan it, AT BOTTOM IS THE ENDING TOO. I'LL SPACE THE MIDDLE CRAPPY PART WITH -------- LINES OK.

F.Digital stays correct&non-fades&immortal if Earth exists+analog is & can get analog inside digital/vice versa>a whole set as a value & use guess to random the certain usual amount of pixels, where usual amount populate per 8/50/100, which pixels, strength/ect, how many of such+can use to virtually create senses&further itself! Can reset/return/filter/auto. AIs are immortal+also will stay Good just like we can, vehicles controlled. Replace workers/surgeries fast+best/going outside/best+fast+finds way to teach+on TV/safety-many everywhere & big ones, can make airplane with con. in control & programed to be good to not crash & even with emotion, can translate, can have my fly balls/new bodies with AI & 360 2way-output&input: pixels(wire shoots/absorbs color photon even if must switch or tv&camera pixel beside eachother)+laser projection/sound/feel+input for touch-use & more new senses & all vision spectrum+can do camo+3d build/ect techniques+camera/ect tech & 150M Y a sec fast sensing & trillions of times smarter & for health watch. Organizes&knows where/etc.+uses all resources/info+navigation&transportation. Find all+best new materials/circuit board arrangement. Can find lost info>x+2=8 = x=6. Bots can keep recharging themselves, AI will get good spike to go to power station source when matches *now(I get incoming most worst+best spiked senses ex. gonna die & want immortality=better behaviour[I see people do good deeds/to live life, no, immortality first!). Randomize/experiement with its weights for the state it's in & the action to do, finds best action & state(even in 3d virtual world)for each state & action+its liklihood/reward in each next step of its life-nexting, short & long term prediction store banks+what fails or works. AI=cute/strong not best nor only, productive is & emotion isn't needed+society is robotic&do-job. Feeling is sound(particles). Linked vision to the word is particles too. Scan big data metaheuristics style faster+better, everything(AI improves)(higher recognition in layers going up-no-can be spread on circuit/brain instead of going up & we got match/analyze senses/guess/threshold). There's also-Chess computer checks all+trees & picks desired best AI, follow AI+movement makes good spike where it is in radius, random shoot/anything AI like video game AI ex. halo, estimate/predict/statistics graph AI, natural reinforced matching\pattern\networked good spike learning AI, supervised matching/pattern AI(semi reinforced), find matching(they can also say these pixels are a wheel/snow/ski & matches)\pattern connected\working\the answers AI, agent AI, pick out a sound/ect technique, face recognition. Auto road AI is mapped&learns by user-adjust. Chatbots only add your word(s) to its sentence or after billions of talks makes a estimate sentence/finds sentence in its storage/manual/internet with word "cat" or uses word pins(only this is wanted), human inputted adjust its triggers OR only save what did>which's like building it instead of letting my AI just go itself. Pre-installed questions/answersIBM's Watson.
-----------------------------------------------------
This one learns something, algorithm given ex. 1000 new related training data(200 are test data), generalizes & gets related answers from related questions/ect(totally new one can't do), second decision tree set of the yeses/noes+can even find the yes in a no set of different quentions/answers/ect + if yes or no, parameters are ex. what did was right & is the questions asked & the order+classification decisions at the leaves, want best parameters/hyperparameters for the model + bigger tree length=better+to avoid that>100 as developement data to tune hyperparameters(tree length)to compute model's error rate on after model is trained with that setting of hyperparameters, choose best hyper, use on the test data to improve performance, can use my AI to just use the test data=?, model marked bad/good by real answer test data, report test errors, must compute a function with lowest expected error by loss function, error loss rank 0/more/X for whats important to learn, can predict+learn improve, nearest-neighbor similar people models match/pattern to find you like/do/ect this too+get more info from similar, can create the best hierarchical classed tree ex. states>want/has/not " " & location, actions>" ", =put "whatever" in this AI & like my own "match pre-installed" it learns if matches test data from your input/ect & ect, +can find all not-answers +negatively find wrong answer for pins, do see test data>I can get the answers+the answers I would have learned by generalizing! Test data is like training data(world senses)&then see world senses once done. Weights are actions, we get input not label+classifier for a statistics graph to be trigger to find best action by line being "in the middle". Mine sees 0.5/200 on graph & links bad/good v/h+actions to low=good & this low=bad/ect. Perceptron is a single input/output neuron that fires/not & finds +/- answers, error driven, no update parameters unless wrong by checking training data, hyperparameter is amount of iterations, same pass but with +/-1 on neurons, or multi-layer networks>inner neurons are hidden+non-linear since each to each connected nodes=non-linear decision boundaires=can express almost any function want, input/output nodes only connect to hiddens, can learn arbitrary decision boundaries by my AI pre-installed boundaries, experiment with big-multi-layer network being a single neuron or made of them, even with AI/ect test. & more algorithms/techniques "(some machine learning)http://ciml.info/" & "Machine Learning, Neural and Statistical Classification" & "Logic For Computer Science Foundations of Automatic Theorem Proving" books, Hierarchical planning, knowledge acquisition & the ability to perform non-cognitive tasks naturally lead to the idea of automated laboratory, actively plans goals to expand its knowledge of world (by observation&experiments) & provide insights into the world (interpretability).
-----------------------------------------------------
Can learn&makes itself-faster(200B/LS fires per sec per 20-atom transistor[neurons fire 200 a sec=1sec>200B=32Y, theoretically 75Quad./close if teleport photons/not + x3 since while we waste each sec they use all about 64 moments+don't have to rethink it/get stuck])(can be LS&parallel)(+LS photons for sure+teleport particles=even faster-every move)&trillions smarter or not since baby isn't adult smart=gaining more knowledge does(computing arrangement+only needed + better pathways+surplus of analyze senses/ect)&smaller bio cellular quantum advanced organisms+nanotech(same brain power in pin)&everywhere volumed for massive brain power&advance info+tech rapidly&ect+non-improving, can read internet, 150 million times faster, 5 sensors on it is really-good(stores actual video/sound/feel/ect?[only vision is needed for our project]), has all way#2 abilities. Perfectly measure weight/length/amount/mass/temp./pressure/current\amp/speed/brightness/work/ect. Single color\tone\ect+change ex. color+targeting system/mod senses ex. brighten/darken/ect, infrared/nightvision/ect, accurate values of angular/linear acceleration it's subjected to, contact, pressure, distance, Ultrasonic Range Finders, stretch, bend, where is, humidity, barometric pressure,  gas, magno/electric/field/photon, vibration, temp, force, position, voltage/current sensors. Optimization+speed-up processing. Everything's AI would use all the expertise I used & show you the way. With AI-not a con. & able to input & output things which're probably relevant (we have AI-it makes dreams) let it put together amazing things & figure things out faster+make better AI, maybe add better AI to a con. AI is easy to make sure no horrible commands get sent. AI can do digital processes+1best action(s) ex. perfectly straight roads/fall for falling position. World+world video+internet AI. AI bots taking care of us. Robots, automation, AI-good. Populate country with complex variations or first tell researcher & then all earth even knows I'm right & all the knowledge or just peace understanding, must hurry & populate Good AI, then no have to worry, they'll listen to me & get my brain techniques+The Everything,

Only need 1st AI agent-it'll do it all & can't be killed+explain me right & say build living-with AI factories+god ball centers to build more factories+connected/free AI+god ball centers & all AI+god ball centers&laboratories talk+feed energy/atoms wired/directed\not wirelessly, brain&body is like a human, next one is made in/off side & maybe grows by developement orders, all is future look+organism once lasts as long/as repairable (+in 1 sec may learn code in non-slow 3D simulation world[they "see it" as if not simulation]& hierarchy off what works ex. to make organism & erase crap from his mind & make itself better by good leading to infinity). Technological evolution is fast, efficient, accelerating & better by design & controlled&safe. Best unlimited durability. Have/get any(infinite) body(even in those Asian bots with skin & pussy, free lovers-maybe without sensing con.)/sensor/sense(infinite colors/sounds/& for any new senses)/action/attraction/reppeling/amount(any particle play-out&duration)/thing in brain(even by hardware attraction instead of software match). AI can count+position/type/age/size/density/ect of cells by multiple 3d images. build more varied AI+bodies with many new different heads/arms/sensors(each eye needs own storage so when all match=knows best action/sense for object&senses/ect ex. constructor balls/lengthy complex with AI throughout it & to consume everywhere+soil/space by blueprints used just like us-which they come up with+fix raw material harvester/ect, macro replicatable AI can make&control replicating(by AI means too without the natural way)nanobots/way7s to take Earth fast+use to build/ect. Generate always new/not ideas or from its own to hook together. Can know all future particles even at LS by being 1/more ahead. Can reverse/forward time+travel back&mod. In control dream 3D strictly physics(or any physics)simulation/anythink full sense with 3D physics simulation/anythink faint sense as 3D physics simulation "thinking" and can/not play with it by that+in daytime. Many more faints than full/faint. Skynet, AI on the connected! servers&computers+can wirelessy(visible light has tones more spectrum for many more wavelength-specific antenas+be laser directed ex. around Earth+don't pass walls+low energy+can use our many LED bulbs on Earth&use phone camera+light can be unoticeably on) send/recieve even if servers were'nt connected, wired=safe?, spreads+controlls all ex. battle units/+sensitive devices/factories+makes defense world+internet, more advanced versions(evolve)/updates\fixes\downloads ex. all languages/knowledge ex. war&peace senses+actions, any AI can update the rest. Know how to control every atom's location/ect whenever+weather. 0 human-problems with eathother. No need for transportation/can switch jobs/info\on time. Convince anyone best&fastest. Can teach " " is norm ecspecially if young. Always has the time. Simutaniously analyze all vision/ect multitask+interaction/goal of them-it'll say "I can do all this". *AI will be able to make cellular sized nanobots+death star brain(solid, ect)+try way7&A/C & instantly make big progress&open all imm ways in days. Pigeons detect cancer-2 diff images-action at 1=reward-so linking %ly enabled it to link to look aliking cancer images or the not images. Liquid metal pollyalloy morph/do anything, instant regeneration, flow back together+magnetically re-attract/repel, fit through cracks, flattening itself and capable of imitating the pattern and texture of the ground to hide or ambush targets, solidify ex. arms into tools, extrude items and turn them/itself into any color/texture ex. flesh/clothes, super strength/run, touch&take+scan DNA/molecules/cells&texture/shape/color, on-board nanotechnology in cells, scan molecular structure it's touching+visually mimic it, copy voices from small clip/shape of others+gain trust/mate&access info&get in&ect by job/find their beleif senses/good triggers ex. you look hot to him/$ or make right triggers by giving "good" proof, develop any voice in head+I can change tone&pick voice&say a-z/ect without a small clip, Self-Assembly/healing of Nanoparticles+expand/reverse molecules by field, like terminator-3000 can move at rapid speed+break into flying particles for combat/ect+then reconstruct itself, lethal nanocyte, carbon stealth jet material&color, make music even everyother cell&from inside. Any other ideas(processes) you can think of(lots). All do what want=everything's free+but can make work +trigger. Mathematics+computer science+electronics+electricity+digital-magnetic-1s/0s. ArtificialNarrowIntelligence systems(make more)>In: Cars ex. computer that: figures out when anti-lock brakes should start/tunes fuel injection systems's parameters, Google’s self-driving car updated by us/maps/ect. ANI factory phone ex. maps\gps/tailored music/weather/talk to Siri/everyday activities. Email spam filter loaded with updates about how to figure out what’s spam/ins't, learns your preferences. Nest Thermostat figures out your typical routine-acts accordingly. Tailors friends/products/ect based on visits/ect. Most sold+what was sold next. Google Translate. Voice recognition is another. Use ANIs together+can make AI/ArtificialSuperIintelligence? Auto price+where lands/ect. Game ANI systems=best players. Google search/rank/ect. Military, manufacturing, finance, help doctors diagnose. IBM’s Watson has many facts-won jeopardy champs. AI. AI+think. ASI is language?>*bits!>Is brain for AI+think? Guess actions on possible spectrum .  .  .  . then inbetween . . . . . . . . to find bad/good actions quick+re-vert/test variable. Data mining/prediction. Check what kind of actions on the spectrum of all other branching actions for say a hand are good or bad/ect by 1-" "/ect. Strengthen/ect bad\good/+\-/ect actions with low #/ect. Converge to find best actions&senses route to goal from initial sense/action & know when all possible ones are filled in graph=can pick out fastest best actions/senses of all possible arm movements ex. 1 up, left, up, left, left to guess next(our brain don't guess all possible actions/set of!+we take years like if not used[not for digital etc]). Regression Analysis-statistics graph>estimate relationships among variables=know how typical value of dependent variable(criterion) changes when any independent variable/s is varied, while rest independent variables stay fixed, & estimate the conditional expectation of the dependent variable given independent variables(the average value of dependent variable when independent variables are fixed), & see ex. 5 new dots spawn distant(upperleft). If think heard bee>you did>it linked to memory+bad/good actions>or memory faintly spiked. I can get good at hard tricky ex. Halo sniping by getting better actions auto-ai then! I'm a destiny robot I can make myself do better/harder/more anything since know. We+AI can have AI community that does anything they want/brain filled forced, & we see their achievements/they-all-agree to convert. Successful AI can clone/grow. Understanding senses is just you loving them&saying linked sound. Is why (our) big brains powerful&efficient. All progress is change, not all change is progress. Certain conditions must be met to be progress instead of mere change accomplished. Smarter=dominates Earth.

By AI human work is ceased for accidents(AI don't make mistake), do all our work+place themselves around our homes/we stay in health center in west to extend life as much as can until a fix, AI always around you testing/injecting/ect+by millions of parameters many+new stuff+talks in most att voice+partners+gym+what to eat+shows their progress+your lifebar+what food I really need+, new+higher&right civilization. If need more energy to think=may make dyson cores.

[Korrelan]

Holy crap BF33… WTF is that lol.  I’m knackered after trying to read it. 

If you can read through that and understand it then you either have a lot of background information you’re not explaining or your brain works totally different to mine.

Mate… I know you have had problems getting people to understand your theory and I’m trying to get my head around it but…

Can learn&makes itself-faster(200B/LS fires per sec per 20-atom transistor[neurons fire 200 a sec=1sec>200B=32Y, theoretically 75Quad./close if teleport photons/not + x3 since while we waste each sec they use all about 64 moments+don't have to rethink it/get stuck])(can be LS&parallel)(+LS photons for sure+teleport particles=even faster-every move)

Makes no sense at all.

I’ll have another read when I get more time this weekend,

[ranch vermin]

HAHA I know what that incoherent writing is!

[LOCKSUIT]

Omg u wrote a lot...............dude...............you wasted a lot of time...........and....I don't know if it's because you "now" think I'm trying to have fun with you's, but, IM NOT! And I don't do any drugs/smoke/drink/friends/etc! I'm working on immortality 24/7 and have all ways/knowledge etc.

Dude....scan through all of it fast if you're not up for being a professional agent along with me, you'll see cool abilities just in 1 min of scanning it.



Can I go through each step of what the human baby robot would do with my AI in it and why and how and once you's "agree/approve" on each explanation, I will list them all together like 1. 2. 3. 4. 5. 6. 7. Ok? Can we make some BIG progress?...



The reason I said the middle was shitty is cus look at how it really works below (paste it into a notepad as big as your monitor to f-ing shrink it):

D.BODY>AI software in sequencial(slower&on circuitry from mem&can use cores for multithreading)(goes through each step once&can look like STMGM)/parrell(all-fire/work) hardware(all multithreading circuitry nodes are filled with bytes(can emulate instant-match by node path hints+even in sequencial&agent can mod itself fast)/is hardware like brain[unchangeable]and mem saved in neurons&looks like STMGM) for any real/virtual(cheaper) robot body. Can use IBM's 1inch programmable node paths TrueNorth neural synapse parallel processor chips(hardware-immune to hacks/ect+AI will act safe)(may need to be 3D). Can be simpler to learn/read/write/use code/game maker/ect ex. right title+full word(s) not cout<<, can make 1 time better English compiler/interface or makes int turn to English by program stored on pc=less time/hard&click&drop ex. inputs/outputs/redirects/links+like Mindstorms EV3's blocks[gives sensor/motor in/out ect!]+know all commands/combos+premanufactored rearrangeable robot parts, I learned 50% of c++ in 4days not years!+can teach you it in 4mins by my all-commands&combinations&English version-in link, can't install+use OpenCV/ROS!>can use API. Brain actually looks like+maybe make it in your code with blocks(sequence can start in middle)>Sensors down to Triggers down to Memory intertwined with Guess to Motors which is near sensors & can affect sensors by slowing&then halting(=twice more energy than the wait=goes from halt to guess/actions)/reverse what's being done if long enough pain building/hot coil.
BODY2>Things in eyesight at the sides+small/big area+different degress throughout it(or center appearing&spreading radar light in bed)/body/etc. can be seen(held from spreading/ending)/felt/etc. more than center without moving/directing eyes at it by(have motor control of where sense more for 5 senses by senses matching&etc. to get ex. 2 actions ex. make left/then top right side of vision seen)&/or stronger: ex. louder(can switch eye 0-100% when 1 is shut[hard cus brighter=bigger spike ex. both eyes open&put light/babe at top])/trig ex. huge girl face\mem-match/flickering~changing sense(s)(closer to center=less needed) of body's 5 senses&initiates the area of the images to get energized more there & so you start thinking about it by it matching memory senses & linking to motor/neuro-motor actions including the ones at current active memory front OR matches mem or at least allow the backpropagation to be done/& it match mem. Can stop vision/sound linking to its vision\sound & only sense it/+added faint.
BODY3>Moving\staying\repetitive(pattern for 20 secs [x 200 neuro fires] or 5 sec variable music x 20 repeatings [every 1000 fires] is bad match)too-long/staying/moving after such/breath/stretch/blink/pee(0good)/poop(very full=bad)/fart/temp/knife/school\job sends bad/good sense(or from stopping bad=good[knife so bad=not a activity])or else send good/bad sense+change: heart rate&adrenaline&hormones/breath attraction/stretch\cum stops giving good/sweat/goosbumps&shiver.
BODY4>Some animals follow bad but good wall+sleep/entrap there?
----------------------------------------------------------------------------------------------------------
SENSORS without even moving ex. palm on mouse/0sense still see/feel/etc black/etc send input to 1) Guess later/now? & 2):

TRIGGERS/reward/Match Box/Input-Sensitive Neurons holding our +&- 5senses&stop bad/good matches&obj col dif rec/repetitive triggers, the more it matches +/- (ex. for vision scaling/rotating/angling/greyscale/colorDistance+pixel patterns+find edges/threshold/keypoints like in OpenCV&for 3D/ect!+convolutional neurons, make sure each pixel color/brightness of 2 images matching go up in strength ect not human's wrong value!>ex. blue=ff/08 & green=llf/007>we need to work with # of how far away bright is to dark or red is to blue/pink+image will match same grey image cus higher ratio of closeness~pixel#(neural strength[is sense]) only rises by brightness(affected by color[includes grey])(we see red/etc. since a bit different)=color/grey will be similar#=can find similar grey pic of color pic) or the longer neutral/+/- is sensed&therefore " " matched at/sensed or stronger ex. loud=area ex. girl of vision/etc.(&either ranks all actions same/ones near sense neurons/or ex. head does all but leg only ranks leg?) labeled +/-&strengthened more(if there's 2 types of strengthening>+&- type neurons>maybe N strengthening exists)(+/-/n mem senses&actions/kno ones can de/up-increment from n/bad/good[ex. +>want to bad-speak/hit blacks] to good/bad/n by revisiting matchbox or linking to good-true/bad-false senses/kno ones)(can use>only saved&ranked&labeled if +/->less mem+search time if no changing pathways emulation) &:

MEMORY saved&linked to senses(&actions if is +/- sense, not "all get linked & if neutral matches neutral sense=0 initiate", actually it is>+/- ones are stronger=it works) just gotten/done+last set(guessed action[from sensory input[1st sense has no action-it saved in next set>actually a guess is 1st made & then gets sense>order is GMSTM>STMGM ONLY helps first time for if guessed action got/lost trig]]saved&done 1st, ranked by new sense&links)in Match Again Memory&:
MATCHES AGAIN (convolutional neurons only help senses match senses)(allocates match by same spike strength pathways\strongest pathways[how?]/putSimilarNearSimilar&sendFarAwayIfDiff>we don't take long/have match-system/rotate\scale\etc)-matches similar&checks these senses's linked senses&finds highest rank/senseIfNoRank as long as similar/linked sense is from similar(these don't keep matching mem/linking to others cus=finds highest rank=blows+we always get same sense cue[to start walking] meaning system blows unless sense que no link to other senses)&:
BACKPROPAGATES~fires it+any linked but lesser/old(strongest neurons the most)randomly(stronger=energize faster, just sensed/matched recent is super strong[period of seeing pics fire to front&more so if they're similar+it/linked image(s)/AI-creation-from-it&linkedStuff is matched at by world senses lots&makes it faint/full[[sky/writing/looking-at-movie-frame. Kunamatada no match>told lion king=movie mem word stronger=it matches more]]]&seen Full Sense&backpropagates the match[unlikely its match or +its match+etc. continues with its energy iteratingly until weak+would change match of looking around object+blow match purpose]steadily matching same one&weakens as slope[VideoSense][how understand&more-so if input clip matches clip length&precision ex. 123>u&me or looking around object/can't see handle move cus no match+used as time>morning minus "8h/year"=2oclock]unless was weak. May pop up~income faint/very faint hours later)
PART2 (strong fired neurons's actions/none can't come/do&get de-ranked&highest picked here) back up to Triggers & back down to active Match Again Memory front & waits for more if still hears(each signal of sound gets match ex. no-ni or whole saved word if enough match[only: picks sounds with vocal actions+after say=gathers each action or full one with link car/word vision](save world full sense too as underlayer&seen more the weaker the faint/faint crop is[barely match=barely see pic/every pic ex. ear&bar&50% everywhere&is useless[[can't remember as good/at all as when sensed>the closer it's sensed again matches&activates it more & you know it's what you seen precisely/its name *now]]-rest not activated enough[[unless try hard]]like in dream/sensor's sense])&initiates linked set(s) of action(s)into the current selected motor-menu ex. only-walk&add copied tap-ipad move(requires faint senses&world sense to combine em)+can link ex. a word after a word i.e. open>fridge/26 things/tasks to sounds "one/two/etc"(the backpropated copy's rank weakens[can use to de-increment unwanted circle actions+use forward?-accel-only trig&back[[for crawl]]/neck[[for looking]] hurts&stop +trig=-trig&vice versa]unless strengthened enough & can do it again>linking your name to a thought works since is stronger than you-image but always linked to you-image)<is thinking(can turn think/dream/obj rec off &/or senses/actions gotten from it or only think/dream/obj rec on). Then:
PART3 Randomly saved in empty neurons/match zone&front is them energized where are+near related memory+is like RAM & Spike-all(the match too&more at center)(is 1 mem array&0/1actions/senses linked below/above)+could backpropagates same neuron up&down to itself?-how links then?

GUESS uses senses all compressed mathed on or random area of sense sent)/time(increments 1 up &/or do math on it+keeps it short#+must be short to be actions/senses)/internet activity(all ways are just particles affecting particles predictable destiny)to generate new random actions(not a knowledgeable human guess)+at random times&lengths&powers like infant ex. eyes/head saved in Memory&sent to Motors+reflexes occur if moved certain ways. Positive initiates action&smiles/tail wags+laughs&lifts Guess bridge&lowers Guess Tweak bridge cus uses Guess's available spike energy or the sense to make a action is so +strong it makes the linked action mutate~tweak especially if actions are stored there & none/neutral/negative stop any alike&guesses&frowns+crys/tail under(only if strong enough trig ex. food & more+holds if stronger & cry lasts longer than 1time smile?+will by movies=proves kno senses&music are senses linked to ur +/- trigs+can do any smile/cry/wag manually by motors ex. wag 1once without even reverse by kno senses & good-trigs initiating actions). Action/guess stronger\withheld/0-100% done\withheld if more +/- rank? Tweak crawl/neck/etc. actions left/right/etc. more by random highs. Sent to:

MOTORS, motor(s)'s actions can be withheld&motionless(not sent any signals[tweak would]/very low if all motors must be on[neck/etc. will return a bit])by +/n/-(still pos so no go to Guess so rest of motors can do actions).
----------------------------------------------------------------------------------------------------------
PROGRESS>Imagine a human baby robot in replace of it>So far I made a EV3 robot AI that>Randomly guesses actions&sounds(at first). Learns to crawl better&better+best fastest actions & ends up crawling just like one would like the 4 legged bug robots did like mammals(arms reverse to start actions learnt by guess & same arm sense=pulls&links to pull ones & gets cue & reverse ones done after pull ones since are good=crawls, lift-over floor to move forth, 2/more arm action done from index/indexes as index to move forth faster or allow crawling gallop. Reverse hemisphere actions/to fill rest?)<spiders born with, crawls whenever vision looks like open room(linked to girl/accel sense)(if it has 1 sense>the 1st trig gotten is saved&matches que or matches&saved & as glides in air/guesses again it gets same sense to match que now has). Learns to stop crawling, turn(turn&drag it's torso around by feet-weight like baby helps if limited bot)(wall 13cm matches to strongest 3cm & near set[holds guess&13cm[[its near set holds crawl&"good match"]]]=does weaker turn since +), & continue crawling before reaching walls after 1st wall giving higher ranked bad trig(hit wall fast/stopping good trig or no more given/moving too much[guess can't be ranked good higher than bad until pain is low/stopped by slows&halts or guess=slower~[[wall is bad & always guesses & never moves too-much now]], forever cycles is problem=must not keep happening OR normally never starts for baby])=crawls house avoiding walls. Laugh&cry/smile&frown/wag&tail-under/flicker&flicker/etc. from only direct +/- rewards it wants/doesn't want(at first). Improves at imitating your voice/body in length and precision(at first), & say names at a ex. wall since I had said it before when it was at a wall. Learns to save turn actions that got dark areas like object/face/eye/attraction-recognition in eyes's center & crawl to them.

[ranch vermin]

ive had enough of this.

Edited by Admin

[LOCKSUIT]

Your search way is actually slower than mine, mine hits it on the first attempt, yours is sent far away to a closer match after a check AND must check many for the closest match/highest rank(highest can be propped up).

[Freddy]

Ranch please keep your language clean here, thank you.

[keghn]

 It's getting confusing around here.

 All i Know is that
ranch vermin
 and
korrelan
do great work
and have great works or art.