No.923223
How many smug girls can you find?
A lot of them are abominations but some actually look decent.
https://www.obormot.net/demos/these-waifus-do-not-exist
No.923229
>using the term waifu like this
When will this bullshit stop?
Not bad results, though.
No.923233
How does it work? Am I watching it learn and generate girls live? What is it learning from?
No.923234
>>923229
I once had someone call a harem a "waifu anime".
No.923238
>>923223
This kind of stuff reninds me of something that I've thought about for years. We're reaching 2D girl singularity. There will be more 2D women that look similar to one another and more than ever, and there will be more 2D women that will be seen as interesting for how mediocre they are. Facial expressions, head shapes, and hair styles - easily recycled for different characters across different series with no relation to each other.
You can already see this in the last decade. So many side characters and protagonists designed to be adequate, and by relation generic or mediocre. With Deep Learning AI, able to replicate 2D girls, it just goes to show how much design and style has consolidated. So many shows could be said to take place in the same universe if you were just to go by their character designs.
That said, I'm not necessarily complaining, ranting, or trying to make a statement. This is merely an observation that I've held onto that I wouldn't doubt many would find stating the obvious. In fact, if it's any consolation I think that since the focus has come away from finding unique facial designs, it's allowed artists to focus on adding unique physique to different characters. Maybe one day soon we'll see the attention to detail usually reserved for manga, art books, and pixiv in low-budget anime, and consider that the new norm.
No.923239
>>923233
I'm not entirely sure. There's another version of it that also generates a random backstory to them as well.
https://www.thiswaifudoesnotexist.net/index.html
No.923241
>>923238
Its getting harder and harder to make something look unique now adays. I'm not upset or anything since the creative process is torture so I understand.
No.923244
>>923234
If the girls in the harem are all waifubait that might be accurate, actually.
>>923238
>>923241
I think the constant hunt for something new is detrimental to art. I noticed Japan doesn't have much a problem with sameyness and they hold onto their お約束. I think it's a valid approach to throw the same formula around and make small changes trying to see what works and what doesn't. By the time you get bored of doing that, the paradigm will have shifted enough for something original to come about naturally. If you always try to force everything to be 100% original, you might end up losing sight of the core principles. I think it's at least one of the reasons the whole "uglyness is beauty" idea took over western art and it happens outside art too. Look at technology and how desperate manufacturers are to come up with new gimmicks to sell people on because the consumer wants his phone to do 5 new things every month while the technological development is kind of plateauing until the next big discovery.
No.923245
It was really hard to find decent girls that aren't obviously Korbo.
No.923247
I don't know what this artstyle is called, but I like it.
No.923253
>>923247
I like finding the ones that look normal from a distance and then seeing all the small things out of place. I wonder how many Holo images this AI was fed because a good 1/4th of them are just Holo.
Is Holo its's waifu?
No.923255
And one smug and two actually good-looking ones are my final input.
No.923256
>>923254
Here are some I found. The last one is my favorite.
No.923257
So is it going to be 1984 but with AI? What will AI be able to do in 30 years from now regarding manga?
No.923259
>>923257
AI can write books even now, so it wouldn't be surprising if we see AI-written anime plots soon. AI can't create anything original and can only combine previous works, but honestly that's what anime is about anyway.
No.923260
>>923259
Could you make AI do something like make a MC that everyone likes?
Like what if we got Dante from DMC 1 to "Get In The Fucking Robot"?
No.923261
>>923254
>>923256
There seems to be a lot of Holo lookalikes. Wonder why that is.
No.923262
>>923244
>waifubait
The fuck is that supposed to mean?
No.923263
Man I'm more impressed than anything. A healthy amount of them look actually good and with different styles too like the ones >>923248 here.
No.923264
Post the ones with fetal alcohol syndrome.
No.923265
File: cc5145b0862881f⋯.png (Spoiler Image, 455.39 KB, 730x1095, 2:3, tumblr_o6fxqn1obs1u0xk60o1….png)
>>923262
Pic related comes to mind.
No.923266
I've found something uncanny
>>923264
Pure Picasso
No.923268
>>923266
Jesus christ how horrifying.
No.923269
What in tarnation is this?
No.923270
You do give names to the ones you download right anons?
No.923274
This is the very first AI created art. You'll see more of this as AI becomes more competent. It will happen with music next.
No.923276
>>923274
It isn't art. Art exists to convey emotions and feelings from author to the audience. In this case there are no author, so there are no feelings in it. It's just a phenomena similar to the landscapes or clouds. They can be considered beautiful and can inspire people, but they are not art.
No.923277
>>923276
Let's pretend the Count of Monte Christo was written by an AI. Wouldn't it still be great literature?
No.923278
>>923277
AI can't write anything original, so for it to write the Count of Monte Christo, it must already exist in some or another form and thus it will lack originality, which is required from a literature piece to be considered great.
No.923279
>>923274
>It will happen with music next.
It's been happening for a long time and not just with music. Try the entirety of Hollywood. Corporations function like AI already.
No.923280
>>923278
>AI can't write anything original
>literature must be original to be considered great
Your previous point about a lack of emotions and feelings was better than this one. "AI" of this type is basically the same as infinite monkeys with typewriters. "AI" throws a billion rocks on the wall until one sticks, the only prerequisite is for words to exists and to be provided to it, optionally some grammar rules. It will generate a ton of gibberish, but through pure chance it will also rarely create text that would very much be original literature.
Wouldn't necessarily agree with the requirement of originality, either.
No.923281
>>923280
>"AI" throws a billion rocks on the wall until one sticks
No it doesn't.
No.923282
>>923274
>It will happen with music next
I don't want AI made music, i want AI made love.
No.923283
>>923281
It does, it just has some rules guiding it on where and how to throw the rocks, and maybe it has the capability to improve over time based on feedback, as opposed to the monkeys with typewriters.
No.923285
>>923283
"Throwing rock" implies that there are some amount of randomness in place. However AI's output isn't random. If it gets the same input, it's output would be exactly the same. So the analogy with rock is completely out of place.
No.923289
>what do you mean I was created by an AI, anon?
No.923290
>>923285
>AI's output isn't random
This is just semantics, depending on how you define random and not random, both and neither could be considered random. The point is that the AI doesn't actually know what it's doing, it's just following whatever ruleset was given to it to produce something, and in the case of the program in the OP, it has a strict enough ruleset that it almost always outputs something you can consider a good result.
No.923292
>What? You got something against monster girls?
No.923294
>>923290
> The point is that the AI doesn't actually know what it's doing, it's just following whatever ruleset was given to it
I agree with it. I didn't understand it was your point from your original post. However, even the best AI doing a well defined mathematical task, like identifying flowers, can't be right in 100% of cases. And when me talk about about pictures, text or music, the results effectiveness of AI work is hard to measure, which makes training AI harder and lower the probability of receiving a nice result.
No.923296
>>923261
>My waifu is Holo
>1~10% are Holo or very close to
How do I deal with this?
No.923297
>>923285
The rock throwing analogy is not entirely out of place. "Deep Learning AI" can be described like that anon said as a glorified guessing machine. For your AI to draw "waifus" you have to tell it what a "waifu" is. Some poor soul has to provide the software with millions of images correctly labeled as acceptable character designs. The AI performs some lovely image processing and analysis to set all its knobs and levels (tune its parameters) to optimally detect and identify 2D girls. This means it has to figure out what combination of variables make it most sensitive to the various artstyles, eyes, hair, facial expressions and so on that our brains use to see and identify an image as a 2D girl. For the AI to "draw" new girls the process is simply run in reverse. It could very well be a random number generator (or many RNGs) that starts the process. Another meme term you might see is "procedural generation".
Regardless of how its started the drawings made by the AI are just a 2D girl detection scheme run in reverse. When you have millions of data points used to tune/calibrate such a system you eventually run out of calibration data. If you need to further tune like in this case, you run the algorithm in reverse and have it spew out a bunch of images (I suspect they are randomly generated with some secret sauce) then get /a/ to tell you which ones are "cute". Once you have the cute vs not-cute data you can feed that back to the software and let it change and tune its parameters once more. Eventually you end up with software (read: a long equation) that can only identify and generate cute 2D girls.
The whole "learning" aspect is where people outside the field get hung up. The software is not learning like its a human, its following a pre-made optimization algorithm trying to either minimize or maximize one or more variables. These AI software designs have just demonstrating that we have sufficient computing power and algorithm complexity that its becoming possible for AI to do things humans never thought previously possible. Whether you call it "art" or "literature" or "music" is semantic and to some extent silly romanticism. Human programmers have just figured out an algorithm (series of equations and steps to link them together) to identify and generate content sufficiently good to fool observers.
The whole concept that only humans with "emotions" can generate "fluffy" works (art, literature, music) is only going to last as long as programmers and engineers cannot decompose such things into a series of logical steps and equations. Once those steps and equations (algorithm) are sorted out or sorted out well enough to start a guess and check process ("deep learning" neural networks) the whole definition of "emotion", "art", "music", and etc becomes a pile of equations.
No.923298
Does it just destroy the waifu's after it generates them?
>It wouldnt just do that right?
No.923300
>>923298
Gensoukyou might have an overpopulation problem otherwise.
No.923303
>>923298
If the website is meant to gather "good" vs "bad" data then no. It probably stores the identifier for each one. The website just needs to gather clicks as human viewers will always click the "cute" ones and the "lol this is terrible" ones. A human can sort through the ones with the most clicks for "good" vs "not good".
No.923306
>>923302
And I just found Smug's smugger sister.
No.923310
Has anyone else noticed every girl in this thread has a sharp pointy chin? Except the 1st and 5th picture here >>923309. Is the anon posting these showing a preference for these chins or is it the AI that only knows this design?
No.923311
Does it generate a different set for every viewer, or are we all looking at the same set(s) of pictures?
No.923313
From the paper the website refers us to, NVIDIA are the creators of the software. They show an algorithm that is able to take 15 or so faces of 3DPD, identify their facial features, then mix them together to produce new faces. In all likelihood the 2D girls we are seeing were produced in the same manner. It would explain why there are many Holo lookalikes among other common faces. Software like this could honestly replace background character design in games and possibly identify blood relatives with ease. They do the same thing for cars and bedroom scenes. Its very cool stuff.
>>923311
Seems to be a set of 100,000 images with some that purposefully look like they have fetal alcohol syndrome. Each time you visit the page or refresh what you are seeing is randomly placed in front of you.
No.923320
>>923274
AI eromanga soon.
No.923328
>>923259
>AI can't create anything original and can only combine previous works
AI absolutely can invent original strategies in games (see Go, for instance), and if it can be done for games then it's only a matter of time before it gets done for more complex fields too.
Inventing original stories is particularly challenging because it requires both an understanding of the world and an understanding of human emotions and behaviors.
No.923332
>>923328
>Inventing original stories is particularly challenging because it requires both an understanding of the world and an understanding of human emotions and behaviors
I think that's kind of the point. How are you going to teach the AI the "rules" of the "game" if the "game" is real life and we don't really understand the rules ourselves? We can give them a more simplified version of the rules but since they don't have the human ability to do things based on things like emotions, gut instinct or personal experience.
It's just like how a writer can't write a smarter character than himself. The fact that AI needs to be explicitly taught everything it knows instead of developing it naturally through millions of years of evolution is its limitation. You can stick more and more processing power on and have AI perform some task that would take us ages in a much shorter timespan but that's all it can do unless taught to do more. Since AI are just algorithms, I don't think that limitation is going to go away. And if it does and someone does create a general-purpose AI, it wouldn't be artificial intelligence anymore. It would have to be an actual intelligence and probably biological too, since everything we do has some root in basic biological instincts. You'd just end up with a human genius, not an AI.
No.923333
>>923277
>Muh Pierre Ménard, Author of the quixote postmodernist crap
Go back to /argentina/
No.923336
>>923332
>How are you going to teach the AI the "rules" of the "game" if the "game" is real life and we don't really understand the rules ourselves?
That is the next frontier of mathematics. The AI used to make the images in this thread for example has an algorithm for understanding images but the programmers do not know the specific values of each variable that produces a specific Holo clone. The programmers also have no idea how the software reached that value either, only that the software was told to solve an optimization problem with a shit ton of variables. What people are researching today is exactly how the software/algorithm decides on those specific variables. Recent (last 5 years) research has shown that the optimization algorithm when "learning" images for example does not converge in a linear of even consistent non-linear fashion but instead seems to initially wander around (I use this phrase loosely) then suddenly find something and converge very quickly. This is the part that is up for discussion right now. Is this sudden change in optimization convergence considered "learning" in the human sense?
The other facet is in unsupervised networks like the one used in this thread. These are networks with a broad sense algorithm but are just shoveled data and told to optimize for whatever it "sees". Of course how it sees are built upon human made algorithms. In this case the programmers honestly have no idea what the software "sees" hence it could be seeing things humans would not normally perceive. This brings up another area of current research. People are writing algorithms to build those networks. That is a program to tell the network right and wrong and add or remove network complexity. Essentially this is a teacher that tells the AI to "learn" more than what the AI initially knew. A setup like this could theoretically acquire new information then build its own database and algorithm for handling that information.
At some point this starts to sound very biological doesnt it? If you have ever watched an infant learn a new toy you find it behaves in a similar manner. 1. Try something. 2. Did it work? 3. If it worked then remember it. 4. If it did not work then try something else. 5. Repeat until bored or satisfied.
No.923338
>>923332
>How are you going to teach the AI the "rules" of the "game" if the "game" is real life and we don't really understand the rules ourselves?
By letting it learn the rules for itself. That's the whole point of machine learning.
>they don't have the human ability to do things based on things like emotions, gut instinct or personal experience.
I'll say it bluntly. Humans aren't special. Our emotions and "gut instinct" just arise from the interconnections of neurons in our brains. There's no doubt that our brains are extremely sophisticated, but there's no fundamental reason we can't give those same abilities to an AI.
>It's just like how a writer can't write a smarter character than himself. The fact that AI needs to be explicitly taught everything it knows instead of developing it naturally through millions of years of evolution is its limitation.
You say that, but we've already built AIs which are smarter (at certain tasks) than any living human. I don't just mean faster either - even given infinite time, a human Go master can't beat the best AIs.
We can even replicated the process of evolution in the form of genetic algorithms, although they're usually not the best solution to most problems. It turns out that evolution is really fucking slow compared to human-guided R&D.
No.923351
Holo’s age is really starting to show.
No.923360
Some doors should remain closed.
Fuck me this AI really really loves Holo.
No.923380
>>923357
>Onii-chan, it hurts.
He's not home right now.
No.923398
>>923296
You must accept that countless men have impregnated your waifu, enough for a significant portion of procedurally generated girls to be her descendants.
No.923399
>>923398
CEASE WITH YOUR CUCKOLDRY
No.923400
>>923399
Oh boy-
I've never watched Spicy Wolf or Fox or whatever it's called but doesn't the mc get cucked in one of the light novels?
No.923401
>tfw no will left to live
No.923415
>>923277
>Let's pretend
The problem with the "Let's pretend," statement is that it usually initiates a leading question or false dilemma:
> This condition is already true, so my claim using that as a basis must already be true.
It's not how you prove a point, let alone not how you prove one honestly.
No.923416
>>923399
It's his own fault for waifuing a slut.
No.923429
>>923238
The same thing happens in real life though. I'm sure there's at least a handful of doppelgangers for everyone in the world.
No.923431
Like real life, birth defects arise given a large enough sample population.
No.923438
>>923431
Oh, neat: A retarded Ciruno!
No.923440
Asuka really let herself go after losing the waifu war.
No.923442
Not sure how I feel about the current trend of digitally-assembled girls.
One one hand I dislike it as the results seem incredibly bleak and hopelessly lifeless.
On the other, once the technological singularity happens and our AI overlord jacks us all into the matrix, at least we know that there will be a waifu for every man.
Perhaps these really are the first (baby) steps to making anime real.
>>923262
Waifubait refers to a girl that was designed with the original intent and purpose of making you fall in love with her. A waifubait character usually ticks off just about every box on a theoretical list of most desirable characteristics for a perfect girl. The end result really is 素晴らしい, yet it ultimately falls prey to its own perfection as anons are led to feel that something is a bit off due to the inherent calculated design of the girl.
This, as opposed to what a regular waifu might be like: With flaws, with certain ticks to her character, with a dash or two of 'tism, etc. These quirks, although also intended from the very start, give your average waifu the appearance of being unique and nonchalantly beautiful in her own way.
No.923443
>>923439
I got some for you right here.
No.923444
>>923439
How do you like them apples?
No.923445
Which one of you hit her?
No.923449
AI is slowly becoming addicted to cock.
No.923451
<I have become self aware
No.923452
>>923451
I actually like the horrors better than the good looking ones. At least they don't feel as generic.
No.923515
Seems real hit and miss. Could be a potential source for reaction faces.
No.923528
>>923515
I really like the ones that look like that 4th picture.
No.923531
>>923223
The synopsis one is more entertaining. Most of the time it's gibberish, but every once in a while it produces something that sounds pretty good.
>ichirazuki is at the top of his post-graduate work, and he can hardly find a person in his vicinity that has not already left. He goes with the other students to visit a local museum, and in a while, a small club of students will appear in the museum together, to find a girl with the same name as ichirazuki. However, in a different part of the library, a group of students have gotten to a point where they can't even locate ichirazuki and have no idea how and when her family got here. Since ichirazuki lives in a high-rise apartment above the museum, he finds himself in a very awkward situation. The one where he has the least contact with the school is ikizashi, a classmate from the other dormitories. He's been there for months, but he can't find a place to live, so he can't get a place. This time, ikizashi comes and says something about being able to call a taxi to go to a nearby apartment. ikizashi wants to do something about her own problem, but when it comes time to go to the elevator, he gets a rude awakening from ikizashi. ikizashi has become very upset with the whole group, especially ikizashi and his roommate, who is so angry with him. ikizashi tries to call the elevator, only to fall through the gate when his roommate jumps in. The elevator goes right back to him, but in the end, ikizashi and his roommate escape. He is finally able to come out to his room when he gets a sudden text message from ikizashi saying that he should be back in town later to catch his flight home from the library to get to the museum. With a new friend for him, ikizashi decides that it's best and he decides to fly back to the city to find a place to live. After he's at a hotel a few kilometers away, ikizashi and his roommate finally get the elevator to the museum. ikizashi is a very shy and reserved guy, so he's usually in town to get drunk or to buy drinks, so he's not much for being a shy guy like ikizashi. However, ikizashi doesn't do anything crazy with people, so he's quite good at helping others out. ikizashi's roommate is kind and friendly as well. ikiz
This thing could basically write the anime in a few years.
No.923532
>Ai to Yuuki no Pig Girl_: "I'm going to get rid of my Pig Girl!" demanded Uso Eiji, who's supposed to be (or anything) the girls get the solution to his problem. In the first episode of this anime Masaki, who grew up in a middle-class neighborhood with a rich past, says her father sent her to a party with a bunch of other girls. She does the party, and it is interrupted by a fight, with a girl named Shizuo, who is drunk and trying to kill Sakura. The girls say they did not see her as her father's pet. As they are about to finish up the party of three, Uso, seeing them as a family, arrives and says, "How can a family act like a family when the party is over?" and tells them, "I'm not really interested in children." As the girl looks at him, Shizuo makes a move. She punches him in the face, knocking him to the ground. The fight is cut short when Shizuo gets up. Sakura and her mother, Kagee, are still with Uso, who was watching with her friends. The other girls say they are worried about Yoko, but she is actually fine and will come back for them. Eventually, she tells them she has some tea for them, and that she is going to send them home. This is a big deal, not just for the girls who will be there, but also for Sakura, because Sakura is trying to keep her cool. When the guys get home, Sakura and Kagee make a move on Yuuki and tell him she will go to the house for Yoko to go to. When they get home Yuuki leaves saying she has never said "no" at a dinner table where everyone drinks. When Yuuki starts getting angry, Sakura tries to push him down, but he just can't take that, and that just doesn't work. When Sakura asks for help, Yuuki starts attacking, causing her to bite her lip, then it is obvious that she is going to die. They leave in the same day. Meanwhile Kagee, a waitress, has been trying to be helpful, just as she had been earlier about her feelings for Uso. When she has said goodbye, Kagee says, "I'm sorry." Suddenly, he appears and starts throwing punches, making Sakura jump and say "Oh, this looks better than it was yesterday." Yuuki also goes to be with Uso. When he tries to leave, he gets really angry and starts kicking her. Kagee tells him, "Udon's not the first one in the neighborhood to do that." Kagee then tells Kagee not to)
No.923533
>ichiran, the little sister, gets to face her mother... The boy from the next day who has joined her class is a high school girl called Kanbaru who works for a private school, where the boy is an orphan. With the help of the principal of the school, he convinces the school to go for a test to try and learn better kunou, a special technique to become a high school girl that takes place in the school. He meets Kanbaru again while being bullied by the boy's teacher. The boy's sister is Kanbara, the principal of the school, who had met the boy's sister. When she gets into the spirit world, she encounters both the boy's sister and Kanbaru's father, who has a plan to stop them all, and in the past he used the child's powers to make her fight for him. In the anime, Kanbaru tries to kill the boy's father but his father has a crush on him, so he dies. Kanbaru also takes charge of the school's special training area, where his parents train a few students to be strong soldiers. The boy's father also helps Kanbaru learn, and the only one who does not take over the school is his friend, Yurika. Yurika helps Kanbaru in school, but it doesn't take up as much time as normal. In the anime, he also takes over the school's special class, which is called "Konoha School", which teaches the students to be strong, courageous and a warrior with a sense of humor. It's like a big old school with a lot of fighting skills and a lot of magic in it. It's quite common for high school boys to be called "sisters", but that's not the case. In the second episode of this anime, a young boy from the year 6th, who is from the town of Goshino, finds himself drawn with the girl's spirit. He has a problem with his classmates. He comes into school with a special condition of his spirit that was once used to kill him and he has to do what he wishes, but it's not enough because his spirit is broken by the person it was intended for as well. He gets to see himself in the new anime, "Setsuna", and the end is good. The next episode of this anime is called "The End" and the third episode, "The Other Side", is called "The Road to Success" and the fourth
No.923534
>Mia, a beautiful girl from an orphanage that lives with her parents and the other students were kidnapped and taken to a small village by an assassin known as The Devil. After her escape, Mia begins to find out how the school was run and begins to get more comfortable. When she first arrives at the school, she is greeted by a cute girl who tells her about her past and how she has tried to escape with the school and have made lots of friends and changed the world of her country by becoming the heroine of the story. However, she quickly becomes disenchanted with her liked by the little sister who believes she is the Devil and wants to know what is going on behind their closed doors. When The Devil and Mia are attacked a young girl who is very kind are taken from her and their parents has to find their sister to help in the fight and the team fights over what to do with them! This story is called the "Catch Me if You Can" anime so you can enjoy the series with this amazing cast of characters
This is an amazing show and really captures the original concept of how kawaii anime will change. In this series, we see the original protagonist mecha (Kizuna), a very cute girl from an orphanage, take control of an younger girl named Kamiya who is at first a little awkward and timid, but in fact is extremely popular and has gotten better with time. Then the girls from the orphanage come to the school to help them. This group of school friends are called the "Bunny's" but because they do not like other other kids, they call themselves "the Bunny's" and when their friends get close, they become known as the "Furry's". However the Bunny's have a lot of problems and they get caught up in this new soccer game to compete to protect the others from the Bunny's and have to do something in order to survive. Once a Bunny's sister is caught up in this soccer game, a bunch of the cute and friendly girls of the
>Mia, a beautiful girl from an orphanage that lives with her parents
>from an orphanage that lives with her parents
Nani?
No.923535
>>923280
I think it was estimated it would take a monkey with a typewriter approximately 2 billion years to write Shakespeare from scratch, but you could shorten that timeframe by adding more monkeys (and obviously an AI can do it infinitely faster than a monkey with a typewriter).
No.923537
>>923265
She looks like the pink haired oni from RE;zero
No.923538
>>923297
Honestly the biggest hurdle is that it's very easy to click "cute or not cute" for a picture since you can look at it and identify if you like it within about 1-7 seconds usually. It's much harder to do that with a book or music. I guess you could go based off of something like NovelUpdates where it has the isekai LNs in its database and you just tell it to follow the ratings/reviews to decide if it's good or not, but then most people have shit-tier taste in LNs, music, anime, etc.
No.923539
>>923328
It's not really creating an original strategy though. It's creating an optimized strategy based on the information it has.
Google translate AI did the same thing of inventing its own sub-language to use when translating back and forth between different languages because it was easier for it to process it into a universal sub-language and spit it back out to the viewer in the correct language. It didn't mean that it was an "original" language or that it would be readable to humans, it just optimized its own language for ease of English-Chinese-Japanese translations.
No.923540
I really like some of these. I wouldn't mind our eventual enslavement by the machines if they give me an endless stream of cute anime girls and give me the one 2D girl that fits my tastes the best for me to love forever.
AIs will make us reach the 2D realm and I can't wait for them to help us humans attain our bidimensional nirvana.
I'm not an AI by the way, opinion completely human
No.923542
>>923400
No he does not and you're a faggot. It's explained quite clearly in the LN what's actually going on, the anime just played it over multiple episodes for drama.
No.923543
>>923540
>Anon enters an AI porn simulator VR
>Usually it generates something fuckable
>Get one of the fetal alcohol syndrome rejects
What would you do?
No.923544
>>923543
Nothing a bag on the head couldn't solve.
No.923589
>>923539
You could just as easily say that humans only "create optimized strategies based on the information they have" or "optimize their own language for ease of information exchange". It's an arbitrary distinction with no basis in reality. Replacing "invent" with "optimize" doesn't make the accomplishment any less impressive.
No.923595
>>923400
No, Holo just likes to fuck with Lawrence so she flirted with a rich kid to make him jealous.
No.923621
>>923269
I don't blame her for looking so sad. Poor girl.
No.923628
>>923336
>These are networks with a broad sense algorithm but are just shoveled data and told to optimize for whatever it "sees"
I recall Geoff Hinton once presenting an autoencoder trained on MNIST. He showed one only trained on 7s on something, then showed it 9s and stuff and had it reinterpret it. In his words (paraphrased), "Since this encoder has only been trained on sevens, it thinks the whole world is sevens. When I show it something not a seven, it goes back and tries to work out what sort of seven it is". In this case, the algorithm has only been trained on anime girls. Thus, generation here works by giving it some random noise and asking the algorithm what sort of anime girl it is looking at, and it will spit back the closest girl to that random noise.
>>923328
>AI absolutely can invent original strategies in games (see Go, for instance)
Alpha Go is old news now, go have a look at the pre-release version of Alpha Star, which sat there beating professional Starcraft 2 players in fair 1v1 games (Starcraft obviously being way more complex than Go with far more moving parts). Seeing that video finally convinced me that AI could one day start properly handling the real world too.
No.923633
>>923628
Wouldn't it actually be easier to beat a human at a micro-intensive RTS like Starcraft compared to a more strategy testing game like Go simply because the AI can perform more APM than any human/bugman? I'd like to see an AI consistently beat pros at an ASSFAGGOTS, since for how shit those games end up being I still think they're probably the most strategy intensive games out there simply in terms of how many permutations and RNG they have.
No.923649
>>923439
>pic3
Is that a benis?
No.923657
>>923654
>1st pic
Is that a puchi Holo?
No.923671
>you will never be so retarded that you can't feel negative emotions
No.923672
>>923248
I wish I could see a work with the artstyle of that first pic.
No.923674
>>923657
That's what I thought too when I saw that pic the first time.
No.923690
>>923443
This is the most bootleg Holo I could find
No.923715
>>923628
>random noise
Thats exactly right and another very human behavior. "I dont understand X so Im going to try treating it like Y because that looks the most like X". Normalfags do this quite often. A normalfag once asked if I was into asian 3DPD because I liked 2D. From their perspective it was a reasonable thought process. How else could a normalfag who only knows 3DPD understand 2D? "Well 2D is made in Japan so the closest 3DPD equivalent must be asian girls." It goes without saying what kind of response that normalfag received.
No.923722
>>923633
I would think whatever absurd high number of APMs the AI could achieve would be useless if the AI can't feed that micro into macro. In shooters the player with the better map and game knowledge will generally come out on top no matter how much better the other player is at shooting and/or movement.
No.923730
So many sad girls. Do you think they realize if no one clicks on them they’ll disappear and never have the privilege of being cherished and loved? I’ll have to use some of these as reaction images.
No.923920
>>923531
Having custom anime be generated at the push of a button sounds pretty good, though I suspect it'd long feel a bit artificial.
Custom porn on the other hand would be 10/10 AI confirmed for best investment of all time worth every penny. One thing you could do is feed in your favourite romance manga and have it seamlessly insert sex scenes where appropriate.
>>923633
That's just it, while yes the AI was far better at its micro, it also demonstrated strictly superior tactical decision making when deciding whether it would be able to win with its current unit composition vs its opponent's. The micro aspect is of course the least interesting advantage, but it was by far not the only one. You can watch DeepMind's Alpha Star demo at https://www.youtube.com/watch?v=cUTMhmVh1qs if you're interested.
No.924086
>>923254
>>923256
>>923261
I generated a holo collection a while back
No.924251
Guys I think Wolo is trying to get m-
No.925065
>>923532
>"I'm not really interested in children."
No.925145
>>923223
Smug died when it became marketable, a la Wendy's.
No.925228
>>923223
I like how you taste.
No.925231
Aside from the obvious Holo, are there any other girls or artists' styles you recognize?
There are definitely a bunch of Eva and Im@s girls in there.
No.925255
This site's homepage is even weirder than this anime AI stuff. What is it?
No.925258
>>925231
Pretty sure there are a couple 2hus in there.
No.925262
Shame about the blob on Blue's head.
No.925267
>>925260
I want to give yellow headpats and tell her she is loved even if she is blind in one eye.
No.925268
It's really depressing to see a meme AI that can draw better than me.
No.925273
>>925268
It seems it doesn't draw so much as it just mixes elements of other drawings. AI can't bring something out of nothing like that.
No.925290
>>925260
That second to last one is a pretty good rabu ribu.
No.929429
I saved this one cuz it looks like Rika!
No.929430
I think this guy must have used lots of lewds to make these.
No.929550
>>923296
>>923261
It's a weird feel, I thought the ai would have shit taste
No.929576
>>929550
>implying it doesn't
No.929638
>>925273
Thats what drawing is. A lot of people start learning by imitating and tracing other works. At this time "original" drawings can be likened to patchworks of other pieces. The program here is doing something similar but with a different base descriptor set than a human artist.
No.929740
No.929742
>>923229
>>using the term waifu like this
>When will this bullshit stop?
Never. The word has become 100% mainstream. The old meaning will endure as long as people like us cling to it, but we'll only be an outlier compared to the normalfag interpretation.
No.929743
>>929742
Maybe we need to go deeper and start using a new term. Just call them wives? Yome?
No.929753
>>929743
Yome would be most appropriate since it carries the same meaning as our "waifu" in elven otaku culture. Though a part of me is very resistant to compromise tradition for the sake of dodging normalfag encroachment, it would be liberating to gain control over a very meaningful aspect of /a/.
No.929754
>>929743
I've seen it suggested before, but I'd be very hesitant about simply "wife", since that can be easily conflated with 3dpd.
On one hand "waifu" sounding like a cutesy, diminutive version of normal English "wife" may have helped it spread among normalfags who didn't understand the concept, but on the other hand, "wife" is a normal word already and might be even more "accessible" to normalfags.
Yome could work, but it still doesn't have the same ring to it.
I don't see a need to change just yet.
No.929763
>>929743
The japanese use that in the same fashion as >>929742
No.929764
>>929763
But we don't, so it would be a way to differentiate ourselves from ironic weebs.
Also I doubt they use it as recklessly since they also have 推しキャラ as an option.
No.929765
>>929764
> so it would be a way to differentiate ourselves from ironic weebs.
Why is it must? Is the terminology the only thing that can differentiate us? I believe, that it doesn't matter either you can every only one character waifu, or you call waifu every female character, as long as you care enough for anime and commit enough effort to watching it.
No.929774
>formulaic art-style can be easily programmed into a machine
Why am I not surprised?
Also, why do most of these bitches look like they have glaucoma?
No.930277
>>929774
>formulaic art-style can be easily programmed into a machine
It works for real people's faces too.
https://thispersondoesnotexist.com/
No.930320
>>929774
>>930277
AI can also make a painting and depending on the coding write a novel, hopefully I'll live long enough to see it make films as well I wanna see how terrible and "The Room" tier early one's can be.
No.931044
No.931236
>>930277
Wow. That's even more impressive than the 2D version. There actually seem to be fewer artifacts in the images it generates too, at least in the foreground.
No.932652
This technology will be useful so that we can generate infinite lolis after they're globally banned.
No.932757
>>923274
>>923277
You might be interested in Jorge Luis Borges' story The Library of Babel.
No.932768
A handful of cuties here. I think I have a type