A long asked question can machines do what we can. Here is few points i wanna discuss.
anatomy of humans and machines –>>>
Human mind takes input stream of reals in form of electric signal chopped at frequency of a minimum intervals of time depending on how capable your mind is so in its lifetime a human for an INT N amoumt of intervals operates on different reals, and this is his whole life computationally but lets now focus on how real these reals are. Lets assign these electric signals some real values according to their strength (potential difference may be one of them). so assuming there are y such properties so we RxRx…xR where R comes y times as our signal again given the fact we can map this set to R itself we can just take input as a single real. what about its precision. Well lets till now take input of any precision. What mind can take input is a range of reals not how long, so again we can have R whole as input. Important is there can be a precision on how much precision it can work on. So far so good on humans whats the input, output again a real based on what mind decides may in movements of tongue or hand whatever.
Coming to machines which surely isn’t restricted to computers, let the machine be an analog machine, with certain operators, memory we will dicuss later right now IO is input.
So lets give this machine an analog signal input with a range on signal, perform operations based on input and what machine is programmed to do. Take the output again is a analog with a range, so far real input real output, no sampling done, a minimum time for a fix input to do computation on it. Bingo we are done, this is a tie here in IO capability of two models. Is there a computational tie too we have to see
Memory for data storage and manipulation -> humans approxiamted to be ~2.5 petabytes ref http://www.scientificamerican.com/article.cfm?id=what-is-the-memory-capacity . So 2.5 petabytes would be roughly 2,500 1 TB hard drives together but it would be too slow to work as RAM, we can rather go for graphics cards. so lets consider nVidia gtx 680m with 2GB VideoRam, quite good speed to match that of mind so if need to reach mind’s capacity if we inc. VRAM capacity twice each year it would be ~ log<sub>2<sub>2500000 years which is roughly 21 years but as we all know constant is too smaller than two and is rather decreasing too each year slightly we can expect a count of 100 years for memory capacity to reach that of human mind and fast enough in 100-150 years. Then we can have game graphics almost same as our eyes show us
Now lets come for operations. I did put example of memory but isn’t it obvious that given this fact precision of reals stored in mind isn’t unlimited implying they aren’t reals at all they are float. No wait for a moment, they are still real, but we can store a finite no. of reals only this is what is exactly implied. consider sin(x) = Sum((-1)^(n+1).x^(2n+1)/(2n+1)!) So we can store sin(x) as a formula but the overall count of numbers stored at a moment will remain finite as 2.5 petabytes is a constant and our memory is estimated near this value. So we can compute finite reals to any precision to any precision given time accordingly. So we can Only compute only rationals till unlimited precision( which is pseudo unlimited precision) rest we can compute no. to precision according to time & memory we can have non of them being infinite, sadly we can’t compute a single irrational to unlimited precision forget getting till transcendental. Can machines do it, they can’t even do it too. Memory size bottlenecks this again and can be proved easily by fact that after all we need something to store info, so whatever hardware it may be, the count of storing enity being finite, and info it can store being finite$ too we are left with finite memory.
Computations here we come. So for a fixed interval of time we have a real input, finite set of operations outta which any operations can be used, temporal recent inputs in original or modified form, and we have to output a real. In humans we do know this is exactly what happens, input is with unlimited precision so is output, and so are these temporals inputs in very less count. As for machines consider any analog machine say telephone, we have input as our voice signal analog, ranged based on volume, purely real with infinite precision if plotted on a graph of RxR we can map to a single R. Output the same. Consider telephone with echo facility or feedback, implying recent memories possible all with infinite precision, key here isn’t storage, key is time delays. so computationally we have to show whatever computation mind can do on One real+ few temp reals + memory of finite reals in form of strings -> One real can be done by machines too and they would be simply computationaly equal.
Lets for now take it as analog machines can do so and see how following operations would be implemented without adding any other assumption, Sure this assumption cause logical circularity in their existence.
**** Visualizaion. Less words said more thought & contexts
We have been for long pursued how good natural language processing can become, well since till now our machines haven’t been defeated by humans if not won, so in natural language processing our most attempts algorithmically have been towards mechanical way to response to given input. Our attempts to model semantics too in words and relations proved not so effective. Now let us take something from humans whatever one says let us virtualize scenerios. For example if someone says to machine i picked up the bat. So now the virtualized environment must have a gun, must have a person talking. Now since the virtualization lacks the surroundings. As for bat there must be some location of it. So it can be straigthly asked where were you, or where was the bat etc. And we get more info on surroundings. Now one thing that virtualization adds is more context information and context become easy to handle. Another use of contexts comes into play when chatting.
So let boy talking to a girl, and write “dat guy” it simply points to “that guy” clearly not “date guy”. Now consider “2day i will dat a girl” it clearly implicates conversion to “Today i will date a girl” not “Today i will that a girl”. So two simple techniques need to implement this simple rule. words to approximate pronunciation then correct possible words for it. After that we employ grammatical correction. Context here came into play through two different ways one is application usage another is inside application for conversation to understood by bot too, and bot too may need to use shorthands to make conversation economic. :D. There is plenty things we can say on contexts & virtualization but for now important is the use of virtualization keeping context in mind to get more info out of less words. It is first step if machines want to defeat the humans if given same computational power. Another example i wanted to point out in this is think of voice commanded game, where you use natural words to command, now say you are on a ship in game and order “Flank them” or “Bang them hard” it sure means “Shoot at them in best possible way” It doesn’t mean at all that Your ship start banging others xD. Now if you ask me how will i handle it, it is what i will do in part discussing emotions. Think of other better examples what can be done by virtualization and context senstivity. And what can’t be done please tell in comment.
**** Merging two virtualisations
Alert minors may get offended
Oh this part should have been included in previous part, no i better not. Guessing one virtualization of a given scenerio is pretty easy, but guessing many and probability estimation of a scenerio is the most difficult task. If i don’t put it different surely forensic guys would shout at me and very loud too. Consider you talking with a bot about you and your girlfriend, and your relationship issues. So you say “Yesterday night we made love”. Now first of all semantics is quite clear so bot must accept it correctly that part we already ensured through context as made is used before love. But now comes the part how bot will virtualize how they had sex. There are many possible ways. Say they were both on bed and laid . Again it has two possibilities who was above and who was below.Simlilarly it may be quite possible they had shower first, or its cold so they had warm bath in a bathtub together. Now with so many possibilities, few can merge to for a generic situation followed by another subdata. For example girl riding boy while on bed implicates one laying other in vertical position, while case where girl is the person laying again we have same scenerio. So it can be mapped to OneLaidOneNot as position and girlBelow or boyBelow as sub position. One may think of why the hell i am giving such a terrible example. So for you information it is one of most important things to note down. If girl is below and enjoying more in that position implies she likes being dominated more over the dominating her mate. While if she loves other position more she clearly likes to be dominating person. So if bot gets affirmed on which position it was it must move on to subpositon to give a better sex advice to boy it is talking to and become Alex Hitch Hitchens from movie “Hitch”. Why merging is necessary is because its better to ask where did the two made love on bed, with a wall or in bathroom. If on bed who was laying down, if with a wall who was in touch with wall, if in bathroom whether in a shower or in a bathtub. This way we must unfold one by one abstraction levels and use the virtualization. It is quite possible we may get answer to a subquestion directly from client side, so in that case we will clear all other virtualizations out of memory where the subscenerio isn’t possible at all. I guess i gave a pretty good example on merging and selection of virtualizations.
**** Lie detection & conversations
Aha my favorite part, friends of mine think i am more in sex and all other distractive stuff for adults. But in truth my intrests lies in expressions, conversations, persuation, emotions and feelings etc. How can a bot get these, how can he be human likely. For example if i say “My grandpa died yesterday”. It must send me 😥 back with helping words like how old they were. If i tell bot they were above 60. It must also say its better they went soon and easy and more. It should persuade me to cry out. Not say something like “Dont’t you weep, soon you get quite, more candies you will get” rather say something like “Let it all out baby, i am here with you to hold you to listen to you”. Both statements help person crying but one is for weeping child one for person who lost someone close. So this was for conversation how one can implement it, well simple virtualize scenerio where one lost sm1 and kill conversation options not fit for this scenerio, create questions for subscenerios pick 1 of remaining. As simple as that. There can be quite good conversations on when a person trying to ask about bot constantly so we must have quite interactive bot with dynamic personality to make person love it, and keep asking things meanwhile. But best part comes for me at lie detection, we want a bot who lies, without letting the investigator get a single clue, and in the process rather interrogate investigator himself, and notice how frequently he is lying, where he did lie, the perfect undercover op for a country possible as a bot one can say. Again it can be done by humans too. How human detects lies, well simple see the expression of person in front, body language and voice tone, all one needs is great set of question and model to take out best according to person in front. Now one would ask how i will be doing it. Simple, did anyone played L. A. Noire, or seen its video playthrough, being a detective isn’t about knowing more, that is knowledge base which we already have more than enough what we need is observation capabilities, so one of them is body language, map it with stereoscopic camera, and analize lip movement, wetness of eyes etc. Finally say bingo, you did it. Moreover first expressions which come as reflex don’t so easily lie. And after all we aren’t making god out of bot, we want best human lie detector / detective out of it, which certainly can make mistakes :P. Another good place to see how much one can know just based on facial expression and bluff about self one better consult Dr. Hannibal Lecter from movie Silence of Lambs. I want to say quite a lot on it but sorry i gotta finish up whole first before i forget it.
Oh yes, emotions. Can machines have them, well as long as we can our hypothetical machine can too. Oh yes, what we feel in ourselves, we can make machine too. How simple, we fall in love, we become restless, add same kind of randomness in machine, i.e. call to random function. It will act same as we do. If i don’t say my emotions to anyone, keep them to myself, i act as if i am not affected at all, what will others call me, simple emotionless. Single word and i am done, exactly same we do for trees, just because we can’t see their feelings we take them as emotionless, so i want this world to think our machine has emotions too all i need is a behavior in accordance, but now lets come to question in philosophical sense can they really have emotions, yes dear they sure can have. We feel emotions through electric signals reaching our mind. They give us illsion of what we call emotions, feelings which physically are nothing. Same we can have in machines, depending on a scenerio we can have its operating system cause such disturbance, may be strong enough to stop them. Consider a case from Game Mass Effect where Geth are machine bots made by quarian race. And one day quarians become against bots themselves because of bad code fragment in control center of geths. And its shown one of the old geths still has emotions for his creators and constantly tries to end the fight between two races. How can one develope such emotions a dynamic and different coding/machinery in each bot, so it become non-deterministic how one feels and can’t be played with its emotions by third person using some weapon, the third person will need good enough persuation skill in order to do so. As for response in behavior i would make use of virtualization + randomness + fuzzy logic (to decide amount of emotion) to decide conversation and actions of our bot of interest. Virtualization already implemented, fuzzy logic is again standard thing, what about randomness, for this we can use certain non-deterministic quantites like weather, time etc naturally controlled entities as seed for generating random numbers. We can have other ways to determine perfect random. Like throw of dice in previous random direction. First random in machines can be put manually rest process is totally recurrent. So i guess i gave certain direction on total randomness, people may say weather depends on places, for those people find mean of weather at current time of year on that place take mean and standard deviation etc, i am sure there are statistical formulas which will let you get random no. on scale of 0 to 1 out of current reading.
**** Creative Writing and other creative works
Now this is just WoW part of whole article, yes here i will be trying to write something creative through our machine. So generally what we do while writing poems or creating some art is add beauty to work use symbolization and abstraction to its finest and create a whole new thing, as for symbolization yes we can introduce it to machines too, how so for example consider two statements which can replace each other in a fragment of poem. 1st is “That glow on her face” 2nd is “Light of that Full Moon”. Now are two more line fragments 1st “In that silky shiny dress” 2nd “I was going to face soon”, Now we can combine the 1st fragments in order to rhyme and 2nd fragments to rhyme, i.e. “In that silky shiny dress, That glow on her face” and “I was going to face soon, Light of that Full Moon”. So we in our knowledge base of our poet bot can add pronunciation of words, how rhyming works and how one can use both poem parts to depict a beautiful lady according to scenerio. So all this poet bot will take input is what you are feeling and for those feelings can suggest words in zumbled form and all you have to do is arrange or maybe write whole poem, best of these bots may compete the best poet. Its just may. Because till now we have done mostly the fitting thing not any creativity at all. But for creativity what we will need is we have to pick up a similar set with almost same meaning to that is in feelings taken input, use a randomizer to pick one of them, try to put picked words to form a poem if possible bingo if not sorry lets move on, once we are out of picked word which are very good candidate, we move on to rejected ones see if we can use them for symbolization and make a poem out of them, if not well again we aren’t doing any less than humans not all humans are poet. As for artworks one may think of them as very tough than poems and creative writing stuff, they are only in the part that you need skill to make them but creativity they are neither tough to create nor easy, here again using machines we sure can have great skill as for creativity we can use virtualization in bootstrapped fashion i.e. virtualize the artwork itself and finish accordingly. Again we may need what kind of art we wanna make as input. But we can put a bot flooded with emotions as input too.
**** Dreams and Deja wu
Okeys so here we come, from rational things to irrational. How we get dreams sometimes totally random world, with f**ked up logic, no physics, anyone can teleport anywhere and what not. But for all dreams we have related to our lives are generally based on our memories and image of persons in our head decide their actions and behaviors very highly. To have such dream for machines very simple just take their memory select few randomly out of its most important ones or the rarest experiences, switch creativity and randomness to max, don’t forget to allow breaking physics rules and introduce ghost and monster, they will help in day time too to create a fear in your beloved bots ;). For totally random dream, just use memory and emotions of another machine from network rather say cloud rest let it same. After that just wait and watch. Now lets come to Deja wu part, beacuse very easily we added dreams to a very normal machine who can do nothing much but all a human can. For Deja wu, let the machine do something, and while it is doing so pick a blurred memory of almost same scenerio or co-related in some sense and modify it as if machine did same work back there. And have fun as it will redundantly call its cache ending up in nothing. I sure know it doesn’t help in anything, its totally useless but it sure sounds interesting to me as implementer of such a world, where you can kid around with your creations leaving them confused at totally random ocassions, where there is nothing to get confused at all. Its just how they are engineered. I wanted to added precognition here but didn’t because that would need god and theist nature which is next part of this article not this one.
**** Is there any God. Precognition. Are prays any good. Redundancy is bad but harmless. Not redundant we don’t have a proof.
Wooohhhhh, so long now comes question of interest to most. Is there any God. Yes there is, if not for us, at least for these machines. And sure this is me who would be watching over them all, would be listening most wierdest prays possible, as i myself make so wierd ones ocassionaly. Moreover i being thiest but also a supporter of atheists rationally. I will have many machines atheist, and would sit and watch them fight in the name of religion and existence of God, using the fact there is nothing which is face by machines can’t occur randomly, any modifications i make to world without disturbing any of properties of it and letting it come to notice of few bots, one got no reason to believe. But what i can toy with is emotions inside them. Which is sure result of existence of God in the given scenerio but doesn’t imply God’s existence. If we consider us humans none of our experiences imply his existence, he may be existent, he may not be. Now if someone asks me how the hell will i justify the precognition, simple i know future, i can change too, i will just show it to some and precognition done. Now coming to are prays any good, what should be mode of praying. Prays are good for most believers. Basically those who will act for it, how to get it done, considering world of machines as operating system and they all can be assumed connected to this world. We can put code in this O.S. to help those who help themselves and based on philosophy and value of others we can decide in what extent we need to accept the prays. As for atheist since these are totally redundant, we will have them keep belief that infact these are redundant. But think yourself having something redundant does it cause harm? no not at all. so are prays. Yet are they infact redundant. You never know when your God wanna retire or make a come back, as long as he isn’t there these are totally redundant yet harmless. But since you can never know when is God watching and is he even there to watch, you can never say if prays are redundant or not.
**** Are we any good than machines
On taken hypothesis that machine can compute from real + several temporal reals + memories -> single real we can say we are no better than machines and can not comment whether machines are better than us. So now this part comes for a friend of mine. Here i wanna discuss needs and dependencies. As we knows all humans have need for survival, for mental strength and other things. Same is with these machines. Consider a machine ‘A’ is having machine ‘B’ as dependency. Now comes a new machine ‘C’ which fills dependencies of ‘A’ but ‘B’ triggers conflict with ‘C’. ‘C’ got its goodness, while ‘B’ got its. Now seeing them as machines if one finds ‘C’ more promising one would quickly suggest ‘A’ should go with ‘C’ o.w. if ‘B’ is promosing have ‘B’. Why in humans are dependencies and priorities not decided like this, just because we are different race. That means what may be right globally we will make it wrong socially, morally. Is this what being a human all about. Now let me tell more about ‘B’ and ‘C’ more. ‘B’ and ‘C’ sure fill needs for ‘A’ but for them ‘A’ is necessity. ‘C’ is a machine which if fights wins and is bound by no fight rule. Whereas for ‘B’, ‘B’ is already having its necessity completed and ready to take arms to possess what it is possessing. The sole question comes is what should these machines do in this tense situation knowing the fact my machines have emotions now and hell lot of them. I forgot to mention, ‘C’ got ’10’ times more processing power than ‘A’ and much more than ‘B’ and hell lot more emotions than ‘A’. So tell me what should God do, what should ‘A’ do. This question i leave for readers. Because i expect something in return from my readers.
$ we can store a real on particle, but it needs to be proved we cant actually use it. If cant atleast prove we aren’t using it. Will update on it shortly.
Lastly i would thank “The Matrix” movie team for inspiring me from very begin