IT'S OVER ART IS FINISHED
https://github.com/alexjc/neural-doodle
>digital art
>>2437924
looks like dog shit
Why is there so much chromatic aberration on the right?
>>2437924
since last year, neural algorithms are going full throttle against artistic tasks traditionally done by humans. as far as i remember we've got:
>deepdream which finds architecture and creatures
>deepdream which simplifies images
>waifu2x which oversample anime style images
>deepstyle which applies a given style to another picture
and now this shit. by the end of the year we'll have probably a dozen more new applications. a nice time to be a tech savy digital artist. with a powerful GPU.
does this only work for landscapes?
>>2437994
pretty much. Zero point perspective natural landscapes. Meaning it has no practical application anywhere.
>>2438000
>being this miopic
is a step in the right direction, you must look at the big picture, and be ahead of today's implementation.
neural networks and AI will be nothing short of a revolution for arts in the next 50 years.
>>2438024
Yeah and we'll also have flying cars and hoverboards and full-dive virtual reality and can live up to 200 years old in 50 years, too.
>>2438024
you are wrong
these are photoshop oil filter 2.0
untill there's no helm that records your thoughs there is nothing to be afraid of for artists
>>2438024
ayy
>>2438050
>untill there's no helm that records your thoughs there is nothing to be afraid of for artists
Even then there'd be nothing to be afraid of because our mind doesn't work like a projector. What's in the mind of most people is completely worthless, even if they had the technology to translate it 100% accurately into a visual medium. The reason why professional artists exist is not just because they can paint pretty pictures, but because of their design sense, visual library and creativity.
>>2437948
I have the strangest deja vu right now.
>>2438027
And then we all will merge with the computer and be united as one being.
>>2438802
Imagine a being with collective d/ic/k consciousness
Our loooooooooooomis wail will shatter the walls like the trumpets of Jericho
Something like this will probably be usable in a part of your work, like for example texturing, but won't ever replace artists as a whole.
Just another tool to call someone a cheater over.
>>2438812
I feel like these would be very useful in generating concept art. Not stuff like character design probably, but definitely for environments. Like photobashing with the human element further removed.
People thought Chess was too hard for AI to beat, IBM did it.
People thought Go was too hard for AI, Google beat the best person in Go.
/ic/ thinks computer won't be doing all the art in the future.
Lets face it /ic/, humans are complete garbage and AI is the future.
http://www.ufunk.net/en/tech/ostagram/
For a second there I thought this thread was about this
>>2439043
Weren't people saying the same about photography ? Also, computers are good at following a set of guidelines. Nothing here shows any creativity nor is it inventive. But then neither are most artists today anyway, seeing all the animu girls and dragons on DA...
>>2439043
I think most people don't actually realize how dumb AI is. All this fucking serious talk even from educated people about robots taking over and AI becoming conscious and shitty philosophy problems when the AI is nowhere near there. I'm not a researcher but I'm in university with top notch AI research and I talk to a lot of the professors that do research and I have had classes on that subject. I can tell you AI is dumb as shit, the field is very young and progress of research in it is really slow. Most of what you see as "smart AI" is really just a bunch of complicated tricks that make computers able to do one VERY specialized task but is useless outside of that one specialized task.
>>2439183
Photography in the end is an incredibly limited form of expression. You can depict only real things, in one style.
It did pretty effectively kill the style of realism in art to the point it's almost not taught in schools anymore.
Once computers get to the level where you can tell them to produce an image of anything you want in any style you want that will eliminate the need for human illustrators.
It may not finish art, but that's mainly due to "art" being impossible to define - basically a meaningless word. You can use to mean anything you like. If you don't like computer-generated art then you can just claim it's not art.
>>2439043
And yet there doesn't exist even the most primitive prototype of an AI who can actually draw and design a fucking stick figure on its own. It's all glorified photo filters.
>>2439191
>It did pretty effectively kill the style of realism in art to the point it's almost not taught in schools anymore.
Not really. Realism was never about copying exactly what you see, that's photorealism you are thinking of. The decline of realism has as much to do with the invention of the photo as the decline of impressionism or mannerism. Literally zero. Retards like you really should read up on at least the very basics of art history before coming to this board to spout such unbelievably stupid nonsense.
>Once computers get to the level where you can tell them to produce an image of anything you want in any style you want that will eliminate the need for human illustrators
Well that's a walking contradiction for one.
Might as well have said
>I can't wait until we invent computers that require you to tell it every single variable of what it's supposed to do, because then human artists will be out of the equation.
wot
Neural networks are certainly going to be a valuable tool to have, but replacing artists entirely is a pipe-dream.
"Ok computer, draw a picture of Goofy in two point perspective with a warm color palette in the style of Rubens with chromatic aberration and the linework has to be rough like a drybrush."
"uh, no computer, I didn't want you to draw him looking to the right, I wanted him looking up".
"Holy fuck computer, the linework algorithm messed up on the left shoe, can you rework that?"
"you muted the colors too much on the grass, bump it up please".
With the amount of time spent tweaking the algorithm --
You might as well just draw the picture yourself.
Here's a question the IBM super computer Watson was once asked:
>"What do grasshoppers eat?"
>"Kosher".
WATCH OUT GUYS, THEY'RE GONNA REPLACE ARTISTS ANY DAY NOW.
>>2439224
>photorealism you are thinking of
Speaking of the basics of art history, photorealism didn't even exist before decades after photography came along.
>>2439224
And whatever realism is "about" doesn't change the fact that in most average people's minds it serves the same purpose as photographs but photographs are more accurate and cheaper, therefore there is less demand for realist art, as any ads or flyers or science books that would have previously relied on meticulously accurate drawings made by a draftsman can now do with a photo snapped by a photographer in three seconds.
>>2437980
waifu2x is also great for smoothing out pixel art
>>2439307
Where did I say that this is easy, desirable, or even possible?
I said once this happens, then this other thing will happen. A simple If statement. What is hard to understand about it?
>>2437924
Can this do more than landscapes yet?
i'm really excited to see what developments are made within the next 10 years with this shit.
>>2440040
is that you anon? looking good
>>2438078
If you think about it though, design sense can be broken down into some pretty simple rules. Surely at some point those rules could be integrated into a program.
>>2440040
Better than 90% of /ic/.
>>2438078
Can you even define "creativity"? Then how do you know that it can't ever be replaced by a simple algorithm making pseudorandom combinations out of a database?
Computer scientists don't know what goes on in the mind of a 9-dan professional Go player, but they can still make a program that beats him. We don't need to understand thought processes in order to automate them.
>>2440109
That's funny now.
>>2440109
>Computer scientists don't know what goes on in the mind of a 9-dan professional Go player, but they can still make a program that beats him.
Becaue go and chess is a very straight forward process. It is just math. The player makes a move and the program knows which move counters it because they are very simple "if x then do y" choices.
Art, not so much. If you told an AI to, say, design a bakery for a heavily stylized fantasy RPG. They'd have a bunch of real life bakeries in their database, they'd choose a photo, then search for a drawing or painting of some artist whose superficial technique it tries to emulate and mash it together. The end result will be at best a pretty painting of a real life bakery, maybe slightly distorted to emulate a fake stylization attempt. As you said basically, a pseudorandom combination out of a database.
And pic related is what an actual artist would come up with. What good art requires is not just technical skill but individual taste and design sensibility, which both is impossible for an AI to have because it requires consciousness. The moment we develop an AI that is self-conscious, we will have bigger things to worry about than said AI stealing our jobs. Visual art will literally be the last job on earth that will be automatized.
>>2440211
>individual taste and design sensibility
>consciousness
Can you even define XYZ? Then how do you know that it can't ever be replaced by a simple algorithm making pseudorandom combinations out of a database?
What you described is what AI's do now, but there's nothing to say someone couldn't come up with an algorithm that replaces creativity at some point in the future.
They used to say the exact same thing about Go only a little while ago: "Chess is just math wahh wahh wah you need a REAL HUMAN with CREATIVITY to make moves in Go!"
>>2440211
>Visual art will literally be the last job on earth that will be automatized.
agreed
there is also very little incentive to automatize such a small industry, compared to say manufacturing
>>2440222
>What you described is what AI's do now
No, there is no AI doing any of that. Right now there is just a glorified photoshop filter doing what a human inputs it to do. What I described is what a simple algorithm making pseudorandom combinations out of a database would do. It requires a future AI that is able to make choices on its own, which still doesn't exist, at least not for art.
I'm really not sure why you are so obsessed with this sci-fi dream that some day art will be done by robots, but don't hold your breath. I highly doubt your children's children will live long enough to see it, but I can guarantee you that no one currently alive will.
>>2440222
Also, do you believe soon there will be novels written by AI's as well? Because that seems far easier for an algorithm to do, yet still very much impossible on a practical level. Anyone could immediately tell that it isn't a story told by another human, with emotion and intent, but a random algorithm stringing together sentences and clichés from its database without any substance.
>>2440237
>No, there is no AI doing any of that.
>Becaue go and chess is a very straight forward process. It is just math. The player makes a move and the program knows which move counters it because they are very simple "if x then do y" choices.
This is what I was referring to, retard. And it's an AI making choices on its own.
Another example would be the AI Watson who beat humans in Jeopardy. People thought you would need some mysterious undescribable quality only humans can possess to play Jeopardy. Whoops, you don't after all.
It's 100% certain all human jobs will be replaced by the end of this century. likely much earlier than that.
>>2440240
"Emotion" and "substance" are something you project into the text as a reader, not something that exists as an inherent quality of the text. Again they are something you almost certainly can't even define yourself despite using the words.
Find out what buttons need to be pressed to cause that effect in the majority of readers and you probably can make an algorithm that makes stories people will think are profound and meaningful.
>>2440211
>intividual taste and design sensibility
Both can be defined, thus both can be incorporated as a choice into a program.
Im definitely interested in how this developes.
Artists talk about AI and computers. I'm in fucking tears, this is comedy gold. Especially when people talk about simple stuff like chess AI and show complete lack of knowledge about it.
>The player makes a move and the program knows which move counters it
hahaha fucking gold
Now don't get me wrong, the program in OP is really interesting but it's more like an improved photo filter. You are not replacing artists with it any time soon. These threads pop up like every month yet there isn't any significant change to the art industry.
>>2440258
>the program in OP is really interesting but it's more like an improved photo filter.
This.
>>2440258
>The player makes a move and the program knows which move counters it
Is exactly what happens. You're thinking the process behind it matters somehow, but if the algorithm arrives at the correct solution - knowledge is knowledge.
>>2440270
It's not what happens. Not even in principle. Hahaha this fucking thread.
>>2440277
If the move the computer makes is not knowledge, then it's not the correct move.
You're not even making sense.
>>2440258
>C-c-computers wont take our jobs
>R-right g-guys?
Even if the AI's won't making art all by themselves, by speeding up the workflow to the point where one talentless fucker can do the jobs of a hundred artists, you must realize the demand for artists is going to plummet.
I mean, yeah, AI will take over art, maybe in 50 years, maybe more.
So what? It's not like you can't just go into another career.
What are you going to do now? Give up art because AI will get your job in the future? cmon
Also, if art gets taken over by AI, that means every single other job has also been taken over, which is a good thing because people will no longer need to work, and we'll live in communist utopia forever (which also means you will be able to make art without worrying about anything).
Either that, or skynet apocalypse which is also great because who doesn't love battling against terminators.
http://gizmodo.com/americans-think-the-robot-revolution-is-coming-but-not-1764029435
>everyone thinks their own job is safe, or among the last ones to be replaced.
Hmm
>>2440291
I'm not an artist by trade tho, I'm a software engineer.
>>2440284
>If the move the computer makes is not knowledge, then it's not the correct move.
Ok m8 I actually don't understand this sentence. I'll refer to the oxford dictionary: "Knowledge is the information, understanding and skills that you gain through education or experience". How can a chess move be knowledge?
>>2440320
>through education or experience
That's awfully specific and if it's what it says then the dictionary is wrong.
It shouldn't matter by which method you arrive at the knowledge, the only thing matters is whether it's correct. Otherwise it's just data (which doesn't need to be correct)
If an AI algorithm searches through a move tree and comes up with the winning move to a game, then by definition it "knows" the move to counter another move. It being the correct counter is what makes it knowledge.
>>2440330
>the only thing matters is whether it's correct.
First of all, that's wrong. You can have incorrect knowledge. There is nothing that specifically defines knowledge as true information. Lots of knowledge is based around things that aren't cleanly defined fact anyway.
Second of all, I'm willing to bet you everything you know you've learned either through:
Being told it by a person, book, etc.
Figuring it out by doing it.
What are the alternatives?
Programming a computer is basically educating it.
>>2440344
I'll have to admit I don't know what the english equivalent is to the word that is basically the "incorrect" counterpart of knowledge. Like "thinking" versus "knowing" but as a noun.
...Thinkledge? Doesn't sound quite right.
Maybe english is just a pleb language.
>What are the alternatives?
Experimentation, calculation and logic. Basically the scientific method. It's probably the closest equivalent to what the chess program does that humans regularly do as well. The computer simulates a whole lot of alternative scenarios and picks the one with the most promising outcome. A human could do the same just much slower.
>>2440291
>C-c-computers will soon make all jobs obsolete, so I don't have to be ashamed to be a NEET failure
>R-right g-guys?
>>2440344
Who cares? The only problem is that plebs will one day not care about human artists because the differences would be too subtle, you'll lose nearly your entire audience but in turn you'll have so much more free time and all sorts of new tools.
>>2440344
What you learn is the technical foundation of art. How you apply that is your skill. What you do with that skill depends on your visual library and your personal taste. You CHOOSE what to do with your knowledge. A computer cannot do that. It can have the technical skill and it can have the visual library, but it has no consciousness to choose what to do with it. Art is not a game with win conditions.
>>2440350
>Experimentation, calculation and logic.
I think these, for the purpose of the definition, fall within education and experience. But still, there's no need for us to start arguing semantics when we both probably know what we're getting at.
The way I see it, the person 'knowing' something thinks that knowledge is correct (that's why they've retained it), but it can still be an opinion. "Well I know that this book is the best one ever written". "I know Trump's gonna win because he's the best!" "I know Hitler did nothing wrong" etc. It's data we've applied value to.
As for artificial intelligence, it's different to human intelligence but it also makes use of logic. Yes, their intelligence is defined by its programming, but you can make the same arguments for human brains (and we can get disgustingly philosophical about this and still get nowhere so again it's probably best not to debate it too hotly).
There was one thing that sticks out to me that made me went "damn that's pretty clever" when it came to AI and the sort of 'churn through the options quickly and get a result' thing that actually had creative results. Sadly I can only find a shittier video rather than the really good one but the premise is similar.
https://www.youtube.com/watch?v=z9ptOeByLA4
The video I saw basically the AI did a big natural selection thing where it'd combine elements, make them fight (each had a cube or something the other one was trying to get, the one that stole the cube survived and moved on) and the AI would independently design more and more advanced creatures to protect their cube and take the other. It was, as far as anybody is concerned, a very creative process based on experimentation. It was just done leagues quicker than a human would manage so it felt like the computer didn't need to 'think' about it.
>>2440330
but under what definition of "knowledge" can chess move be knowledge? That doesn't make fucking sense.
>>2440330
You do realize that just because you store an information in computer it doesn't become smart right?
And chess AI algorithms never come to a "correct" answer, by correct I assume you mean a best possible move. With the exception of the end game of course.
At the end of the day these are just tools. I think the really question is not how do we compete against these machine, but rather what are the ways we can use them in our own practice.
A good analogy i've heard is the relationship between a shepherd and sheepdog. This article was talking about the rise of autonomous street cleaning machines and the human workers they might replace. It was suggested that pairing them together as team might be the most effective way of using them. They both have different abilities, strengths and weaknesses which complement each other. A shepherd can't chase sheep like a dog, and a dog can't make the strategic decisions a human can.
This applies a lot to design, given the rise of complex computational tools, but is relevant to art as well. Its about synthesizing a method of working that takes advantage of both human and machine.
>>2439307
lolol, the reaching here is hilarious. So rarely do I get to see such a base idiot furrowing his brow and marvels of technology he can't even begin to comprehend the genius of.
No one cares about how computers aren't going to be able to draw exactly what you want. Guess what, artist's can't either. But if you're able to tell a computer
>I want a 100 images of a girls being fucked by shrimp
And it spits them out, and they're fappable, guess what, artists just lost.
