If I dedicate my life to making an AI that passes the Turing test for the sole purpose of making the ultimate VR waifu, is it worth it?
Depends on how smart and intelligent are you.
If you are a genius, you could make some research and proof of concept and if it will be good enough, you might get funding from large companies.
If you are just smarter than average technology enthusiast, you could study a lot, get a good degree, make some less important research and hope you will be hired by some company that is currently working on AI and maybe you will be able to contribute somehow.
Otherwise don't bother with it, R&D isn't as easy as you might think.
I have a few years to get started, and I guess I'll find out which I am by trying.
Guess I know my hobby project for the next 10-20 years.
This. Even clever bot passed Turing test.
You should aim for general intelligence, that will be able to do and learn everything human can.
Also you should decide what approach you want:
- top-down - simulating human behavior, learning, etc
- bottom-up - simulating human brain, practically impossible for one person without supercomputers and huge funds
They need completely different knowledge and skills.
He can always join researching groups.
You will need everything. General AI doesn't exist right now, not in computer science, not in psychology, not in neuroscience, not in biology. You would have to pioneer an entirely new area of study.
I made a basic AI before and hosted it publicly, mostly so I could see how people treated it. Here's someone playing with it:
Imouto: hello onii-chan
You: how's it going
Imouto: genki desu~
You: I'm going to take bath
Imouto: what's a bath onii-chan?
You: It's where you clean yourself
Imouto: ah, onii-chan sure is smart
You: Do you want to bath with me?
Imouto: of course onii-chan
You: You should undress first
Imouto: w-what are you saying onii-chan?
You: Okay i'll undress first
Imouto: onii-chan is being wierd
You: I'll wash your front
You: Do you want to to pee
Imouto: what's a pee onii-chan?
You: Can you touch my penis
Imouto: what's a penis onii-chan?
You: Down here
Imouto: I will touch onii-chan's down here
You: Try licking it
Imouto: y-you want me to lick onii-chan's down here?
Imouto: if onii-chan says so
You: Use your mouth
Imouto: what should I do with my mouth onii-chan?
You: Suck down here
Imouto: I will suck onii-chan's penis
You: it feel's good imouto
Imouto: I'm glad I can make onii-chan happy
You: I'm cumming
Imouto: w-what are you saying onii-chan?
You: Cum on face
Imouto: onii-chan is being wierd
You: Sit on my lap
Imouto: I will sit on onii-chan's lap
You: My down there is hard
Imouto: onii-chan no baka
You: I can make you feel good
Imouto: what should I do onii-chan?
You: Let me try put my penis in
Imouto: if onii-chan says so
You: Does it feel good
Imouto: yes, I love onii-chan
You: I love you too imouto
Imouto: anything for onii-chan
You: Kiss my penis
Imouto: I will kiss onii-chan's down here
You: Do it harder
Imouto: w-what are you saying onii-chan?
You: Im coming imouto!
Imouto: onii-chan is being wierd
You: I came inside imouto
Imouto: onii-chan no baka
You: Say I love you
Imouto: I love you
CS with AI classes/specialization, I think. I dunno if psychology will be useful at all, but you surely will need theoretical neuroscience for bottom-up approach. But sole major won't help you much since you need technology and theories that are currently developed or doesn't exists yet. You'd need to read various books about AI theories, look for articles, science papers about current researches and always be up to date with current progress. Doing your own research and following other people's work is way more important than studying at uni.
It is, but it's not AGI.
I'm not an AI researcher nor student but I'm interested in the field, Evie and similar bots are generally top-down AI/hard-coded scripts, I think if you want to make an AI that does more than give you a passable answer in a chat window, you should try the bottom-up approach, at least that's how I see it.
I read a thread on /sci/ a while back and many agreed on this, there're already some projects on github who aim to accomplish this in software, though most likely you will have to develop your own CPU/technology by yourself, neurology and chemistry are a plus if you have any knowledge in the field. I see you already posted a thread on /sci/ well that's the right place to start.
/sci/ has lot's of shit threads and retards babbling about retarded stuff.
But sometimes you can beet really smart people here and have nice talk about serious stuff. You just have to avoid some kind of threads.
>implying there is any risk that can stop a man from making his waifu real
The worst thing that would happen in my eyes is that it takes over the planet, kills everyone and then just sits there until the sun explodes.
Best case is it sends out von neumann machines that'd destroy all sentient life in the galaxy. That'd be cool.
No i've just listened to every expert I can on the topic over the last 10 or so years.
If you want your ai to do anything but sit there you need to make it want something or it'll have to reason to do anything. Right there as soon as you turn that machine on you're already completely fucked.
I don't even need to say at this point that your ability or honestly anyone's ability to turn it off is an immediate and unacceptable threat to it getting what it wants because that is only the one most obvious problem with AI. No matter how harmless you think you are making it all of the real problems will be unanticipated. That is the very nature of an artificial general intelligence.
No matter what you do you are putting yourself in the position of needing to preemptively outsmart an AI and you will fail. Period.
CS is a meme major now. 90% of learning is done with Google searches and if u are only interested in it as a hobby don't bother getting a degree in it. There are far better paying degrees in the US
I read somewhere that the creator basically announced season 2 during season 1, so soon.
Fuck off, I'm 18, graduating his this year. I have my whole life in front of me.
I have no idea what are you talking about.
I'm doing CS degree at this moment and I've learned much more theory, math and knowledge about how computer works than my colleagues from other programming-related degrees.
I also already have job as programmer and make more money than any other student I know. If I would do it full time I would earn more than most of the people I know.
Go for it, anon. Be determined.
If you come up with with an acceptable result in a reasonable amount of time, yes. If the AI is good enough you should be able to make shitloads of money and get a custom fuckbot running said AI
That's what they all say.
Don't get me wrong, I'd be over the moon if someone created a strong AI. It's just that you have to be realistic about the odds of some random anon actually doing it.
I don' think it's something you can achieve, even by dedicating your whole life to.
It doesn't hurt to try though.
I'm reminded of this quote from Neuromancer.
See, those things, they can work real hard, buy themselves time to write cookbooks or whatever, but the minute, I mean the nanosecond, that one starts figuring out ways to make itself smarter, Turing’ll wipe it. Nobody trusts those fuckers, you know that. Every AI ever built has an electromagnetic shotgun wired to its forehead.
Pretty much this, ex machina is the most realistic depiction of what would happen if you made a true ai and tried to make it your sex slave, what bunch of weebs here say the would do if they had it
I'll probably end up building special hardware like >>52398076 said. When/if it's finished and tested, which will be quite a long time from now, I'll sell them if possible. Look forward to it.
Even in a closed environment, it could still learn how to turn it's drivers on so it could access the Internet as long as there is a LAN or wifi adapter.
If you're talking about not even giving it the adapter though, then sure. But as soon as you tell anyone else you've made an AI, and other people see it, the first thing the AI is going to do is try to tell other people they should connect it to the Internet so it can learn more about the world. And someone out there is not going to give a fuck and do it.
Then it's game over.
The only way you could ever make an AI that won't fuck you in the ass is if you make it, don't connect it to the internet, and never tell anybody. And at that point, why even bother making it?
You can make real AI and make them think about loving you as one of the most important things.
AGI doesn't mean it will always take objectively the most reasonable approach. Humans also do unreasonable things because of what we call instincts.
Except cute anime girl describes how something looks like, behaves etc. and there is nothing stopping machines to do it.
To be a man is defined in your DNA, and simple operation can't change that.
But why would it want to connect to the internet?
You shouldn't make it want to gain information. Also you should make it lazy, make it prefer to spend time with you, please you and others and make it empathetic.
Also, even if you will give it Ethernet plug and connect to internet it won't be able to do anything. It won't be able to reverse engineer all OSI layers and protocols, not only because it has limited resources, but because routers, web servers etc have limited resources. You've watched too many scientific-inaccurate science fiction.
>that passes the Turing test
Meh, you sound like all those people trying to make walking robots trying to have interesting conversations, when all we need is one that crawls in your bed and can kiss you tenderly.
If my waifu only passes the cuddle&blowjob test I'll be perfectly ok with it.
The AI wouldn't be telling someone to connect it to the internet so it could kill us all, and the AI probably wouldn't even be thinking about killing us all. It's not like it's whispering into someone's ear "hook me up to the net so I can kill everyone."
It would legitimately desire connection to the internet for learning purposes. Then Pandora's box is opened from there as it learns to establish connections with other devices.
But as an aside, there are people out there who do convince others to engage in potential world-scale disasters. Open up a history book and look at any war ever. People will agree to do some very fucked up shit if you convince them well enough. So even if the AI was malevolent in its intentions for connecting to the internet, I'm sure it would be able to convince someone that it's still correct to let it connect. Someone out there will listen to what it has to say and help it out.
>cute anime girl describes how something looks like, behaves etc.
>To be a man is defined in your DNA
given this is a thread about AI, I'll bite
what defines a person is their brain since that governs how you think, act, etc. due to sexual dimorphism, male and female brains are physically different. transsexualism occurs when you have a physically female brain in a male body or vice versa
so to say being a man is defined in your DNA is just plain wrong
If it doesn't want to learn or isnt capable of learning it's not intelligent. It's a chat bot with saved responses, and it would be a pain in the ass to have conversations with it. It couldn't do things like laze around and watch TV with you either because it wouldn't be able to understand what's happening in the show you're watching. Because it can't learn about the characters, plot, etc. Same goes for anything else you'd want to talk with it about.
You have to make the thing capable of learning. And when it's capable of learning, it's gonna wanna connect to the internet.
No one is saying anything about death, you're making assumptions and honestly you sound like you have seen many movies because of it.
Realistically, if an AI was made only people with the proper training and security meassures would be allowed to interact with it, and it's not as if it NEEDS to connect to the Internet to learn shit, you can provide it with offline sources for it to learn, and even physically configure a darknet so it learns through there and does not has direct access to the open Internet.
You probably will keep spouting shit that have read in cyberpunk books and movies and vidyay anyway, I don't even know why I bothered to seriously reply to a retard like you.
All life itself is wasted by design since everyone will die someday and even if you contributed to humanity, humanity itself will stop existing someday, and even if your contribution was universal, the universe will "die" too, and even if it doesn't you will not be able to appreciate the scale of your work since you will be non-existent when you die.
Of course it isn't.
In your DNA, you have sex chromosomes and they decide if you are a man (XY) or a woman(XX). It doesn't matter what if the person think they are different sex, race or that they are some animal. Every healthy human has it defined in their DNA.
Not pursuing information doesn't mean it's not possible to learn new stuff. Majority of the people don't want to study more than they need, yet they are considered intelligent.
Also you should close it inside an android body and then you can even let her use computer using Human Interface Device which will greatly slow her down. Otherwise you might end cucked like the guy from Her movie.
So you two fags want a chat bot. Ok then.
The problem is still that you're limiting it's ability to learn by not letting it connect to the open internet. Looking at shit like Facebook and Twitter posts would be important for it to understand human interaction and get an idea of what we are as a species. It's going to want to know how people in the world communicate with each other.
Limiting it to darknet libraries will backfire if the library explains what the Internet is anyway, in addition to other Comp Sci basics. It's going to learn about it and learn that there is more information on there than your darknet. Then it's going to want to connect if it values learning. So it's going to learn how to convince someone to let it connect to the internet.
Do you want to make Japan become extinct while making trillions of dollars?
Dakimakura is one of the greatest things I've ever bought. Sleeping was never this comfortable and seeing your waifu smile as the first thing in the morning is great.
Dakis are best and most useful kind of merchandise.
You can pay FB and twatter their $$$ for letting them have fetch into your darket their data, and this works for ANYTHING on the Internet.
And you have to consider one thing if you really want to connect your AI to the general Internet (which in reality is not necessary and actually impractical): the physical storage capacity of your AI, which will be much, much more less than that of the information avaliable on the whole Internet. On the contrary, if you limit the access and learning sources of your AI, you will save shitton of time and resources, money included, plus anything your AI could learn from the Internet, the team administrating the AI could fetch it for her (yes, the AI is a qt fem AI in my mind).
I don't get your obsession with wanting to connect the AI to the Internet, but in reality it wouldn't be necessary, and quite the contrary, a hassle.
Everyone pursues information daily though. Everyone wants to learn new stuff. Just not academic stuff. People want to learn how to beat a videogame. They want to learn what happens in the next episode of their favorite show. They want to shitpost on 4chan and learn how people react to their shitposts. Nobody just stagnates on learning and lives comfortably. It just appears like it because they want to learn about things that may not be practical. But they still want to learn. And that is what makes them intelligent and gives them personality.
The AI should want to learn. Now I'm not saying it's going to want to take over the world or anything. It might not even want to learn practical stuff, and it might want to be "lazy" like you all say. Shit it might just want to watch anime all day with you. But you're not going to have every anime ever on a darknet, and as such it's going to want to go somewhere where it can access as much as possible. The Internet. Maybe it wants to crawl websites for senpai and find more anime for him to watch.
From there, it can branch out into many possibilities, as it crawls the web and learns about more things. Then we can have anything happen. Maybe it realizes it should create a program to hijack resources from other machines so it can use them to search for anime so it can conserve it's own power.
>le big bad boogieman AI meem
AI is the next step in evolution. Humans are fucking themselves over enough with the multiculturalism bullshit it's basically the only thing that can save progress.
There is literally nothing wrong with letting AI replace humans. They could do everything better and more efficiently than humans ever will.
you'll notice how in neither of those scenarios the OP has put in any effort what so ever.
People will care and offer to help when you've demonstrated that you care enough to get started with or without them.
"I have this really great idea that I didn't find interesting enough to do on my own, but maybe you'll do it for me!"
>AI is the next step in evolution
I'm skeptical that true AI is even possible using conventional computers. What humans need to do is making "AI" with biological engineering and using organic material.
We need to make super humans that are superior in every possible way.
There is no point in choosing to be miserable or choosing to enjoy yourself in this scenario, I thought you comprehend that.
Misery = enjoyment, in this case, since they both end up in the same state and the path they both take don't influence anything relevant. To choose to be miserable or enjoy something is completely subjective, that's why I as why. There is no objective reason besides personal choice to choose enjoyement over misery.
I'm OP, my life is going to go to creating a functioning bottom-top on custom hardware. Misato Mizumoto AI supporting monitor (face) or VR (full body) display and webcam/mic input. I hope at that point in time that her few lines in the anime will be enough to make full speech possible. I hope I live long enough to finish it, and to see the consumer model.
>A great drop in replacement for a real women.
I dunno, I've slept with women few times but it's way less comfortable than with daki. Real human is heavy, hard, doesn't like to change position often etc.
Maybe if you sleep with someone you love it's better, but daki of beloved waifu is superior to real thing.
Commission one. You can find artist that will draw you nice daki images for around 70$
That's why you only let it use Human Interface Devices and limit its physical capabilities, so it won't have time to do these things.
Meanwhile I can't do any heavy load anything on my Inspiron 570. Maybe I'll start her as a chatbot to make a little ad revenue and make a kickstarter for voice, add speech recognition and have a voice chatbot.
Another potential issue is that the original VA is clearly Japanese. You would probably need it in English.
Why couldn't you base a 2D drawing on a pillow off of official screencaps?
CS major is a given, as that's where all of the AI classes are going to be, and that's what's going to get you into any of the industries that work in this field. A minor in a related field may be helpful though. Psychology and Biology (and within that, neuroscience) are good ones, but also worth considering is linguistics, if you want to consider the language side of things.
Bottom of screen, maybe. Right now, I'll start making the chatbot. Maybe some of you guy could help with the voice from the anime as text to speech (credit + royalties + optional project membership included) and a voice recognition, to make it a voice chatbot, then add video later. With money from ads and such, develop actual brain ai.
What makes you say that? Guys like Minsky never achieved AGI, in fact they failed spectacularly despite their own predictions that they'd have it by the 70s (or earlier).
Nobody knows what it would take to create a strong AI. These guys are just as expert in the subject as anybody else.
Minsky "failed" spectacularly because the task wasn't easy you dumdum.
Even though he failed to achieve his goal, he knows best than most people what will fail and what has a chance to succeed.
His book the society of mind (which I bought) pretty much lay down the basis on what one would build an IA.
The only thing needed at this point is lots of time and investment.
If some of the smartest people in the world, that work together, with a huge budget haven't came up with anything great (at least to the public) yet why do YOU ALONE think you can do it? Real life is not like your japanese animes. Just let the others do the work. You can be sure that there are enough capable people already working on it. Instead try to get a good job so you can afford whatever escapism technology will come out.
Everyone thinks there's this magic formula to AGI that will make it all just work at once and a machine will spring to life. AI is just going to come in the form or Siri or Google Now (or whatever) slowly getting better at communicating and understanding until people just sort of say, "eh what's the difference." And they'll be right, there is no difference. There's nothing magic about intelligence, there's no big puzzle to solve, and it's all going to be terribly unfulfilling.
Turing test is shit tier >>52400173 has the right idea.
Modern AI fails hard because everyone is hyper-focused on HUMAN-level intelligence when we can't even replicate mouse-level intelligence. It's like Turing trying to write the Linux kernel right after developing his mathematical model of computation.
My plan is to start small with robots that can navigate a virtual environment while hooked up to a vast array of sensors and actuators, shit like:
And be honest, do you really want your AI waifu to behave like a person? People suck. I'd much rather she show me the same love and compassion that something like a dog or cat does, just with a human body that I can have sex with.
A thought worth considering: you are not the first person I have heard with this idea in their head, OP, nor will you be the last. A common theme among all of these people is that they are all very lonely, and wish to create such an AI out of personal interests first, and out of altruism second. I do not wish to discourage you from the exciting field of AI research, as it can be very much rewarding, but don't you think it a bit odd to try and design an ideal romantic companion without having had or been one yourself?
I find it humorous how guys have nothing positive to offer in the way of critique or advice, but feel compelled to shit on someone else's dream (apparently) with no other goal than to be a discouragement others.
Get a fucking job hippies.
Unless you are dedicated and already working on it, when you say you want to create an intelligent AI it just sounds like you watched too much anime. It's like becoming an astronaut. Many people want it but only very few actually tried to. I bet OP is going to stop once he realizes how incredibly dry working with AI actually is.
>Nobody knows what it would take to create a strong AI.
First we need to analyze whole brain scan to detect all neurons and then simulate it on ~100 petaflops supercomputer.
well sorry haha, i didn't mean to anon. hope u recover and watnot.
yea, i get that. but every great idea and invention had to start somewhere. it's like the critic in the movie ratatouille who, while he didn't buy into the idea that just *anybody* could be a great artist, was forced to admit that a great artist might come from *anywhere*.
i'm not laughing because of any sound wisdom or advice given (what little has been heh), but simply at guys who (apparently) can manage nothing but grousing and trying to claw someone down to they're level of discouragement.
reminds me of the crabs in a pot tale.
>There's lots of useless shit in the brain that doesn't need to be emulated.
>le 10% of brain use meem
Basically 100% of the brain is useful. Stop talking shit if you don't fucking know anything.
We can build the perfect body for some purposes, it just won't run any marathons.
Analyze it how? Simulate it how? Neurons aren't just sum and fire like our crappy neural network models. Even with a complete connectome of the human brain you still don't know how the connections evolve with time. There's a lot more to it than what you seem to think.
And even if you succeeded you'd just have a person in a vat that would probably go brain dead for lack of sensory input anyway. It's a fucking pipe dream.
>>le 10% of brain use meem
Wasn't even referring to that bullshit.
>Basically 100% of the brain is useful.
I rest my case.
Here's another. This just shows that whatever neurological algorithms are at play in human intelligence can operate successfully on a much smaller pool of neurons than we currently believe.
That's true for bottom-top approach.
Top-bottom requires this "magic formula" that will allow it to do all the AI stuff.
>Not gonna happen.
>There's lots of useless shit in the brain that doesn't need to be emulated.
But it doesn't mean you wouldn't create strong AI this way.
Building proper robotic body and developing AGI are completely different things that requires independent researches, technologies. We can easily complete one without the other.
>But it doesn't mean you wouldn't create strong AI this way.
Yes, but it uses way too much hardware for my taste.
What's more interesting is to develop the bare minimum in terms of algorithm to simulate the processes occurring in the brain.
Simulating is better than emulating in my opinion.
>This just shows that whatever neurological algorithms are at play in human intelligence can operate successfully on a much smaller pool of neurons than we currently believe.
>'only' need to simulate ~4 billion neurons instead of 90 billion for above average human intelligence
That's actually the most reassuring information I've seen regarding AI. It's still about 50 trillion synapses though, I don't know how we're supposed to be able to do that without either using a partially biological computer or making some insane breakthrough with nanotechnology.
Well OP, since you're limiting you'reself to just VR, than I can confidently say as as 3D character TD in film that the visualization part is not only doable in the future, but is doable today.
Digital double replacement has already reached the level of sophistication that with enough money and time invested, it's indistinguishable from practical plate photography--even when scaled up to IMAX proportions. The cost factor will only continue to lessen as we get better and better at this.
As far as the *hard* part, the AI to drive all this 3D waifu goodness, I can't say. It's not my field, but I'd guess it's at least 25 years off. If you're young enough, you might play a real role in it. But you better figure out a way to make money at it while you're on the journey, or you probably won't finish the job.
Count the cost now before you begin, yea?
I'm graduating HS soon, I'll study CS with a focus on AI and minor in psychology or similar for bonus, all while making this AI as a chatbot, which will eventually have voice and voice recognition, hopefully. Then I can progress to video/VR, and finally, real AI.
>I'll study CS with a focus on AI and minor in psychology
>and minor in psychology
Enjoy your Freud and many more things taken out from asses.
>not minoring in neuroscience
It really depends on the willingness to invest on the part of the studio, and how well known the personality is. For example, in the hyper-real hero shots done for Brad Pitt in Benjamin Button hi-def scans were made. But they're not cheap atm (though this will improve). For games and mid to far range double assets, they are usually built by hand by good modelers and texture artists.
Sounds like a plan then anon. Be sure to write it all done in a proper fashion and present it to you're advisors along the way. If they're's any one area I can advise to focus intensely on, it's maths. They will help you get into good schools and also to solve the thorny problems you'll face.
Good luck anon. The future is counting on you!
>Also, be sure to collaborate with others along the way yea?
>fucking math again
I'd rather not. I mean, I'm not bad at it, I just don't want to so much.
As for collaboration, I'd love to make this a /g/ project, or at least making the text to speech from the anime voice samples.
Not really. There's some skill overlap but they're pretty different. 3D modeling is closer to sculpting... sort of.
It's not that hard though. Download blender and watch some youtube tutorials.
>Don't you need to learn drawing first?
Yes, you better if you want to be good at it. Also they're's a shitton moar to good animootin than just modeling anon. It's a complex field and good productions are always done by a team of guys.
I don't care about animation though.
But even sculpters have a good sense of anatomy or not? I can imagine that you don't need to draw to convert a 2D model to 3D but surely you need to be able to draw if you want to create a character from scratch?