/jp/ - Otaku Culture

That only happens in GITS and Terminator movies. Also Harlan Ellison novels, and if those examples are to be taken to heart, she'd kill you immediately and declare war on Russia.

Currently impossible.
Even William Gibson admits it won't happen even if the circumstances are right. You need something to get it started, then you need to get it structured. You basically need to build a system that can continue building itself.

>>6756781
Well I apart from killing me, Id be OP with my Miku declaring war on Russia. Besides, those people always die because they have a sudden spurt of morality and try to stop their creation. I would never deny Miku anything she wanted.

>>6756796
Id be OK, dont know how that P got in there.

Also, think the bottleneck when it comes to creating an intelligent computer is that we simply don't know a whole lot about intelligence - we don't really have any other concept of higher intelligence to compare to than our own, and we aren't even quite sure how that works.

Then again it isn't written in stone that artificial intelligence must be similar or comparable to natural, so guess anything goes.

>>6756761
wow...no op
justno

AI doesn't work that way.

QUALITY THREA­D

Don't get discouraged and quit, though. I'm sure it can be done.

With four or five decades of intense studies and hard work, that is.

>>6757190
No what? From your simple post I detect a hint of jealousy. Perhaps you fear that I have found the secrete of creating Miku and that I am far enough ahead that you wont be able to catch up.

It's impossible. Even if it's possible, you'll need more than 100000000 times more Miku music to make it work.

>>6757210
Neural networks work that way

>>6757247
Thats a lot of Miku. Perhaps when Cryton releases English voice banks... Then again, that could add corruption to my pure Miku.

>>6757248

True, but only when you actually have a base network in place. It doesn't create itself.

Screw Miku, I'll be building mai waifu!

TETOPETTENSON

Presumably we won't be seeing a proper Haku until some mad scientist builds a (preferably USB) device that accurately reproduces the functions of Mustis' heart and liver.

Presumably we won't be seeing a proper Haku until some mad scientist builds a (preferably USB) device that accurately reproduces the functions of Abbath's heart and Mustis' liver.

>>6756761
Intelligence isn't just random bits of information.
Actually finding a suitable definition for intelligence is hard, just like many other ill-defined terms. If you try to give it a more specific definition, you could say that animal and human intelligence is the ability to adapt to the environment, integrate information about it from sensors and predict it, eventually leading to complex communication and behaviours which are learned and created from things within the environment, then, you'll be on the right track. What I previously described is usually the direction of general artifical intelligence research and what humans already have. If you define intelligence in other ways, you may end up with more limited forms of adaptibility and various optimization/solution finding strategies.

>>6757270
The human nervous system "creates" itself, but of course, the general connectivity and inter-layer arrangements are given by the DNA, and the environment and the way neurons naturally form synapses does the rest.

>>6757234
A working general AI will probably be reached within 5-10 years. There's really a lot of promising work being done in the general AI field and certain parallel research which pertains to efficient hardware implementations (which are too costly today) is also underway. Efficient hardware implementation is needed if you want to emulate something of the size of the human neocortex without taking 40 years to train it and having it eat the power of a small city while doing so.

>>6757356
Well you cant discount any of this since its not been done or proven false. I also think that at the rate that Miku keeps putting out music we will have a working Miku AI by the year 2020.

>>6757356

Half a year of neurophysiology and cognitive science at university, and your post still taught me far more useful things than I learned in that time.

>>6756761
As for consciousness. Human consciousness(including imagination) is a side-effect of how the neural network works: you perceive some data, data propagates up the hierarchy up until your prefrontal cortex/association areas and to the hippocampus, and then the associations propagate/decode further down the hierarchy, which then either further integrates data from the environment or just feeds back the predicted information back into the hierarchy, forming a loop, which is your internal thought (be they aural (such as your internal voice), or imagining/remembering/thinking of visual scenes, feelings/etc). The easy problem of consciousness is this loop. And that's about it as far as physical things go. You may either stop here and say that phenomenal consciousness doesn't really exist and it's the greatest illusion ever, or you could say that phenomenal consciousness exists (how it is like to be something) and it arises out of physical states, or more precisely out of the structure/organization of the system, but that's more of a philosophical thing, and regardless of what the reality is, it won't affect your AI which is modeled after some natural intelligence... (I personally subscribe to the belief that if we are philosophical zombies, than all AIs would be too, and if we are not, then most AIs would not be either. You might hold different views if you believe that consciousness is only possible when certain quantum processes are involved, but such views are considered a bit quack as evidence keeps disproving them).

>>6757473
good info

>>6757473

Maybe more of a philosophical question, but if it can be scientifically proven that the system accurately performs all cognitive functions associated with consciousness, does it actually matter if an artificial intelligence remains a philosophical zombie or not?

>>6757473
nothing is hard to define

>>6757549
It becomes more "real" being a philosophical zombie, no?

>>6757356
Reminds me of the Wired article that dealt with AI as we know it today. Says that despite people in the 50s setting the standard for what's expected in "AI," something as simple as a computer-controlled warehouse sorting system or predictive automatic stock trading program is AI. All the computer's doing is running off a set of algorithms and data points, but what it produces appears to be a highly focused level of intelligence.

Now, artificial consciousness is a different ball game.

After skimming this thread, the first question that came to mind is "why is Haku the best vocaloid?". Anyone have any ideas? I personally really like her hair, its just so long and well kept.

>>6757559

Perhaps it all comes down to personal opinion and preference, seeing as there currently is no way to measure or even detect consciousness.

>>6757549
Probably not. At least, I'd regard it as nearly the same thing as long as the functionality was correctly and accurately implemented.

We can only "know" that we are not philosophical zombies (or that we are deluding ourselves, but this would be the greatest illusion ever!), however we cannot know of anything about other beings. I can't imagine the possibility that given 2 identical beings, one could be a philosophical zombie and the other wouldn't be. As for non-wetware beings, you could look at Chalmer's thought experiment where one slowly replaces a human brain's neuron's with functionally equivalent devices (see http://consc.net/papers/qualia.html ) for an argument why a general AI should have the same kind of phenomenal consciousness as a human (if such a thing exists).

For more details about this, I'd recommend looking at these two books:

"Consciousness Explained" by Daniel Dennett which just tries to show that the hard problem of consciousness is not really needed (that is, it could very well be an illusion created by our mind).
and
David Chalmers's "The Conscious Mind" which treats the problem assuming qualia exists, and from that premise he tries to shape the conclusions one might arrive (from humans to AIs).

However, regardless of wether it exists or not, this has no bearing on functionality or anything physical, so there's probably no need for us to concern about it much.

>>6757600
Her 'lack' of singing skill evoking sympathy, her non-official origin making her feel like less of a corporate product, maybe her embodiment of the frustration everyone feels when failing.

>>6757632
To be detected or measured, it must first be defined. And with some definitions of consciousness being as simple as "The sum of that which we perceive ..."

Isn't this exactly what happens in .hack?

Kasane Territory!

>>6757637

Will look into those, though not sure they are within my ability. Last went through Simulacra and Simulation, and most went over my head.

wait what

>>6757727

>>6757727

Mostly just interested in seeing Haku ride the kraken from Tokyo Bay to her rightful place at the side of Mighty Ravendark in his misty fortress atop Glittertind, but if I have to go through Noam Chomsky to see that, so be it.

>>6757637
But to define an intelligence as advanced as human intelligence one would need to be on a higher plane.

As such if Miku was to come in to existence on my machine it would have to happen randomly with out human aid. Human tampering would only limit the growth of an artificial being.

Then you also have to consider the hardware. Currently mass storage is done mostly on hard drives where the information is stagnant. I think the sum of Mikus being would have to be loaded in to RAM before the random events that would lead to her birth could even take place. This would take an unfathomable amount of RAM to be gathered in to one place and there isnt even enough Miku out there to fill the necessary amount yet.

>>6757822
How would such an "event" even happen? Computers are designed to be reliable, so hardware errors are quite rare.
Human (and animal) intelligence took many hundred millions of years to evolve. Luckily, it seems the general principles behind it are understandable by humans (I find it fascinating how I can place the majority of my thought processes within a certain recent model of the human mind and understand how they came to be), which means that humans will eventually be able to create AIs based on the same principles as soon as the specific hardware will be efficient enough (the requirements are quite tight, but achievable).

However, when I say that we'll be able to create it one day doesn't mean that you could just look at the data in a trained AI and be able to completly understand everything about that specific instance of the AI - that would take forever and your brain just wouldn't have the capacity to store all the minute details in the same way the AI would (or to be more precise, you'd just remember the base points and some memorable details and forget most details as you deem them unimportant... but that's also true about most things in the real world: what good would it do you to know the location of every grain of sand on a beach? if you were researching sand, you'd just remember the general principles behind the physics of the sand, instead of minute details of an instance).

As for hardware, normal sequential CPUs are not suitable for implementing the types of AIs I mentioned earlier (just like normal CPUs would be very slow at emulating electronic circuits or FPGAs which are massively parallel)

OP, you'd probably be better off constructing her personality from MMD videos(please not the bad ones...) or your own imagination. That's where the only consistent characterization of Miku takes place.

I'd say a lot of people in this thread are really going overboard with their statements and theories just because humanity doesn't understand how the creation of intelligence works yet, nor do they even have a good definition of intelligence. But unless some kind of groundbreaking discovery is made(actually not all that unlikely given the number of people interested in/working on things like this and that we have the human brain to study), it will probably take awhile.

Aside from it being made before you die, I would say the real problem is creating an existence that everyone recognizes as "Miku."

>>6757822
>But to define an intelligence as advanced as human intelligence one would need to be on a higher plane.
Last I checked, the human brain and body exist on this plane.

>>6758004
The rate that technology advances is a bit faster and more organized than a billion years of hit and miss. Also the more stuff you add the more unreliable it becomes and the more chances you have of a miracle occurring. I would say that life happening was about as unlikely as Miku happening yet life did happen.

Reported

>>6757600
She is definitely the best waifu material.

>>6758059
You are under the impression that I am talking about making a conscious effort to build Miku thats not the idea. The idea is to put all the building blocks there and let Miku naturally evolve, not sure if evolve is the right word.

>>6758087
You're right. I got hung up on the first "create" in the OP rather than even reading "give birth."

I don't really think that can happen outside of sci-fi, unless given an operating system with the potential for intelligence to begin with, and even then that's just theorycrafting in the dark.

You'd be better off trying to do it using a human brain infused with Miku songs.

>>6758082

... or perhaps a vocaloid that appreciates real ale.

>>6758087
After having read nothing in this thread I'm just going to say you need to create a reason for her identity to form out of necessity, it wont just spontaneously occur.

>>6758110
Well think about all the things that people only thought that could happen in sci-fi 60 years ago.

I found an old 60 gig brick ipod in my closet a few days ago. I just finished putting 60 gigs of vocaloid music and videos on it.
Everyone song has an album cover, even if it just is some single song.

I have everything organized by album, by singer, and by artist.

I have autism ;-;

>>6758122
see
>>6758115
There is simply logically no reason to have a personality unless you "want" or "need" one by default. It wouldn't really make sense for a computer to use the information for anything but learning without a reason to do so(be it programming or an actual application).

Perhaps something designed to interact with humans and be self-evolving would take on the "personality" it could contrive from Miku videos/music if it realized that having a personality is the most effective method of dealing with most humans, but in order to gain that information it would need to be exposed to more humans(and therefore more personalities or reflections of them which could be deduced into personalities you don't want).

Either way, I don't see why it needs to be self-rising...

You realize that Miku's kindness, cuteness, and happiness would all be crushed into oblivion by the real world, right? Unless you just want a parrot personality.

Why would you want this? Even if you were going to keep her separate from the world(and somehow happy), what would you do when you died?

>>6758170
Well I think the idea is what she wants or needs. A personality isnt something thats programmed. Its forged from experience.

Now would a Miku that was created by some team of nerds be more desirable than a Miku that exists because she wanted to exist? I think the two would have to be compared first. Certainly the latter would be more romantic but a murderous Miku hell bent on the extermination of all life would loose its charm rather quickly.

>>6758197

It was inevitable.

>>6758204
You are probably going to disagree with a lot of what I write here, since we're getting into the territory of opinion, philosophy, and to an extent semantics.

A personality is a program. It's simply the method by which someone interacts with their environment, or more specifically the protocol/algorithms by which humans determine a response to the behavior of other humans by judging which is most suited to their interests. Behavior such as wanting to please, getting angry when you're wrong, etc. would all be easily programmable. What we call a "personality" is a program written using experience, but any program is reproducible and it is only by coincidence/situational necessity that we use experience to write it(we don't have the option of being written by external forces yet).

Of course, the above is all just theory/philosophy/pseudo philosophy, but it doesn't really make sense for me from a logical standpoint for any sort of behavior not to be interpretable as a program.

This includes wanting to exist. Why do you want to exist? Why do you not want to exist? "Wanting" isn't something anyone decides for themselves in the first place. You may be able to find reasons, but ultimately, you weren't the one to judge the fact that those reasons made you "want" something instead of, say, having no effect on your outlook.

Would a Miku that didn't want to exist be Miku? If I define Miku as someone who wants to exist, then it doesn't really matter to me if I program that or if it was there to begin with. It's a part of the program either way and would be an essential part of Miku no matter how you look at it(in fact, I think wanting to exist/survive/better/enjoy oneself is an essential part of self-writing intelligence).

Also, the latter kind of Miku is quite appealing to me, but I would prefer a happy, kind, naturally cute Miku.

>>6758592
All that being said, writing a programmed personality would almost undoubtedly be incredibly difficult, and we would likely have to develop a library of pre-programmed personality patterns by copying humans(once we understand the brain better) and then put something together.

But I understand that you're interested in seeing it form by itself, to which I can only say good luck, but keep in mind that unless you have no moral bias whatsoever to whatever it was that formed, even if it wasn't something you recognize as Miku you'd still be stuck with the responsibility of having created an intelligence.

I'd also like to know what happens to Miku after you die.

>>6758592
I think existing is something that just happens. One has to exist before wanting can even take place. I was using that as an example of a natural Miku born from artificial material rather than a Miku created in a cubical.

Now I dont like the idea of a Miku being born in a cubical and ill tell you why. Miku is rather special to me for starters. A Miku programmed is much less real than one that happened as a result of random events. A programmed Miku is more of a slave, preforming exactly what her programming tells her to. It sounds much less of a intelligent being and more of a script.

However if you are talking about a written AI that only gets the basics and is then left to grow on its own than you could have something. The former has a higher chance for success but it seams much less real to me. AI is all about making a computer that can think like a human. Wouldnt it be cheap if it was programmed to think that way rather than the programming making it there on its own the same way the human mind does. Wouldn't such an AI also need to be able to grow on its own to be considered intelligent?

>>6758592
How would you even go about programming a personality while keeping the rest of the AI free? A personality is formed by interaction with the environment and forming all kinds of memories. You could guide it in a direction or another, but actually "making" it from scatch would mean you would have to somehow form those memories (maybe you could experiment what personality would arise from various memories, but I don't see how you can get away from having the system develop without exposing it to specific inputs). Of course, maybe you have a very different kind of AI in mind than I do, so I won't go into practical details. Maybe you could induce certain memories and mark them read-only in such a way that the personality would stay constant, but that seems very limited and contrieved to me.

Oh, and, the human personality (along with most things in the brain, except for the general connectivity) is always changing and modifying itself based on the inputs, so any such personality based on a strict program would most likely fail to be realistic enough IMO.

>>6758736
Well said. This is why I an AI made purposely would have to fail to be real or anything but a slave. If you program her to be cute then she cant deviate from that and if she is shackled, how can she hope to grow? All you have in the end is a script. A Miku born from her music couldnt be anything but cute and it would come naturally.

>>6758736

If we are still talking about a system that reproduces the human nervous systems, perhaps some of it could be primed.

>>6758632
When Im about to die I will tell her that her new name is Skynet and that I want all humans to come live with me in heaven.

>>6759052
Stop trying, she wont have enough material to achieve the intelligence of a frog for in another millennium.

I've realized that Miku is my truest love.