I found it interesting to note how viscerally uncomfortable I felt while looking at some (or, most) of these pieces.
Of course not all art has to be pleasing to the viewer, but confronting/challenging art is made deliberately that way by the artist, and the process of experiencing the art involves understanding the feeling and learning something from it.
It's a different feeling altogether to be discomforted by art, but for no purpose.
Honestly, most of them feel very bland to me. Often in the same way, with the blank white square as the extreme case, but there are a number of other series typical of its style.
The algorithm could use some additional imbalance-creating modules.
I guess in reality, the physicality of the art plays a big part of how you experience and appreciate these types of pieces.
The textures and subtle tones, the method of craft, the interplay between mixed media, the energy (or lack of), the emotion (or lack of) etc and so on.
The Apple Maps screenshots book and 9-eyes.com are two projects that really stood out for me. Neither was intentional and both involved curation, which is perhaps what is missing from this (and similar) projects.
These are vastly more artistically satisfying than the OP, and most other GAN images that are presented as art. That's a great statement about process mattering as much as result (or is that just my POMO talking?)
Do you like any type of art, or is it only painting that leaves you unmoved?
If you appreciate the technicality of it that's still appreciation. I'm like that with photorealism, awed at the skill of it, but otherwise unconvinced by it as a form of painting.
Pretty much couldn't really care any less about most art. It's just some paint on a canvas or a statue or whatever.
Some people say they are emotionally moved or mesmerised by the beauty etc. Sure I can appreciate it is a skill etc etc, but I don't understand the emotional reaction people say they have - I just see a "thing" like I see advertising boards on the street as a "thing". I wonder sometimes if people just say this stuff because they think they are supposed to.
100% this. In some I think I can see recognizable images (a car? A bird? A human face?) bizarrely decomposed and distorted. Combined with the manically meticulous, not-quite-repetitive patterns and discolorations, the pieces have a distinctly tortured quality to them.
>It's a different feeling altogether to be discomforted by art, but for no purpose.
There's something alienating in the emotional response evoked by an AI designed to simulate human creativity, which ironically transcends itself in the relationship between a "real" and "artificial" intelligence, creating a legitimate work of art in the contemplation of manufactured emotion. You begin to question whether your response is any less the product of a blind, stochastic machine than the stimulus.
Also, no sharp lines or objects. There is a lot of geometric abstract art in a continuum from organic (Pollock) to sterile(Rothko) and everything in the middle.
Hi all, this was trained by Michael Friesen (https://twitter.com/MichaelFriese10). I liked the latent space so much that I decided to share it with everyone. Have fun! I'll add some html to credit Michael with the training once I find the time.
The initial page just shows the art, but after a couple seconds it pops up with an explicit credit to Michael Friesen and a link to his twitter:
> Imagined by a GAN (generative adversarial network)
StyleGAN2 (Dec 2019) - Karras et al. and Nvidia
Trained by Michael Friesen on images of Modern Art
Funnily I get nothing from these. I can't see any tension in the compositions. They're aimless. I guess it's what I'd expect from mixing attributes from other paintings.
What did the painter want to do here? I don't know, I can't detect a painter to begin with.
For me it's about 1/3 hit 2/3 miss, and those that hit I usually want to tweak the composition or alter it somehow. Rarely am I completely satisfied with output from it, but that does happen.
I would take many of them and scale them up to a resolution that would work for LARGE prints, with a tool that does its own upscaling/stylish embellishing (like Dynamic Auto Painter), then work them up further in digital.
I think it's a fantastic tool for brainstorming ideas.
Anyone know the license terms for images generated by this web site?
At an abstract level, this stuff is still derived from works done by humans.
I can't process this (at the meta level), ha ha.
Basically, the universal question, but where does art begin, and where does it end. This is in essence, taking something already abstract (I'm assuming) created at the hands of expressionists, and turning it back into something—random—
So what's next? Combine this with thispersondoesnotexist? Generate fake descriptions about the art? Make a real-fake exhibition? Make a fake Wikipedia entry?
I guess at one point, all this makes it real art, where the artwork isn't the object, but the craftsmanship is. Might be interesting to put the repo on display at a gallery.
It is arguably already "real" art. Art does not have to be physical to be "real," it can be virtual; art does not have to be made by humans to be "real;" art made by an AI is real, etc.
make a virtual world game where everything (faces, bodies, dialogue, animations, bedrooms) are generated by GANs. Put people inside, plug a needle in them, connect them to the matrix, and profit.
Funny that of all the "This X does not exist" sites, this one only gives relatively small, low-res pictures. Because I'd totally print out some of that stuff and hang it up!
This generator of all of them seems to raise the more existential questions. What does it mean to exist? Surely art is art regardless of whether it's on a physical canvas. Just the act of generating it and serving pixels changes the state from "doesn't exist" to "does exist"...
Not exponentially. More like linearly in pixel count, you just tack on another convolution block. (StyleGAN doesn't use self-attention, but even naive self-attention is just quadratic.)
The problem is more that you hit diminishing returns fast from training at higher resolutions. There's not too much difference between training at 1024px or training at 512px and using an off-the-shelf superresolution NN upscaler, but the latter is like 4x faster. So why bother? You don't always even have input data which is 1024px+.
As an abstract digital artist I find most of these not interesting enough for me but the real issue is a GAN cannot make images that are big enough to print without looking like crap or taking forever. A decent sized 30x30in looks good at 300 dpi which is 81M pixels. I need a decent computer and video card to work at this size. A GAN would need to be massively bigger to function at this scale.
GANs can be made fully convolutional so that you can generate images of abritrary size. But the training of the latent space will still be limited by memory and speed, so you'd end up with images that are perhaps quite repetetive in nature.
I've always wanted to do something like this, but wired up to your webcam to detect when you blink. When you do, update the picture to a new one, to make extremely ephemeral art.
I totally misunderstood your comment. I thought you meant it would track you and take a picture of you and create art out of those pictures. Reading your comment again I'm not really sure how I came to that conclusion at the time. I thought that was a really dumb idea, but turns out it was mine the whole time.
Your actual idea is much cooler.
To expand on your idea, it would be cool if every time you blinked or looked away, it wasn't a completely new picture, but the same one that is subtly changed, such that it slowly morphs over time, but only as you aren't looking. I suppose you could pre-generate quite a few steps in advance and then mete them out gradually.
In 1929, Rene Magritte made a painting of a smoking pipe and, on the bottom of the picture, he wrote “this is not a pipe” in French [0]. So, this website seems to be using that same surrealist theme.
I think the more shocking version would be ThisArtworkDoesExist.com - just fill it with some of the worst 'real' modern art that has somehow got displayed in a gallery.
Reminds me rather of that not-technically NSFW image generator based on Yahoo's NSFW-detecting neural network, particularly the art gallery section. (Probably NSFW, by the way): https://open_nsfw.gitlab.io/
StyleGAN2 takes 512 dimensional vector as an input. Would be neat to attach some physical input device, like a matrix of dials... And encode your own painting with it :)
Out of sheer self-interest: Do you intentionally obscure the source of the file? I'd be interested to save some of these (for non-commercial, strictly private usage).
curl -v https://thisartworkdoesnotexist.com/
GET / HTTP/2
content-type: image/jpeg
...
Warning: Binary output can mess up your terminal. Use "--output -" to tell
I'm pretty sure the server is just serving an image rather than an html file hence the lack of a source.
I tested this using Safari and Google Chrome and when inspecting the image element, both browsers produced different <img/> tags which I assume is because of the browser itself.
No one. All the 'X Does Not Exist' websites use random generation, without any human curation; the US Copyright Office is clear, non-humans (and machines in particular) cannot create a copyright, and since there is no human author expressing any creativity in random generation via NNs, there is no copyright, and they are automatically public domain. I got this question a lot on my anime faces as well and provide a few references on why they're public domain: https://www.gwern.net/Faces#faq
What is that interpretation based on? I'm not aware of any case law on this.
The input of these networks are all copyrighted. The closest analogy in terms of copyright is probably "collage".
In any case, the idea that you can just take some artist's life work, train a CNN on it, and then do what you please with the results doesn't seem "right".
> What is that interpretation based on? I'm not aware of any case law on this.
There's plenty of case law that non-humans are not creators and cannot get a copyright. How could you have not heard of the Naruto case, for example? It was famous! (And funny.)
You're ignoring the fact that the inputs are copyrighted and that therefore the "non-human" is not the (sole) creator and the "work" is derived. The CNN isn't creating anything new except for random numbers.
Taken to the extreme, your position would be that any transformation performed by a computer program and involving random numbers would effectively remove copyright. That can't possibly be true, so there would have to be some "threshold of originality". Again, no case law on this, as far as I'm aware.
In effect, I believe you are spreading legal misinformation based on a flawed understanding of copyright law, which may put others in harms way.
> How could you have not heard of the Naruto case, for example?
Not really a good comparison. The "inputs" here are the monkey and the jungle, neither of which are themselves protected by copyright.
A better comparison would be a monkey creating a "director's cut" of Disney's Jungle Book, interspersed with footage from National Geographic, both of which are copyrighted. I doubt the "a monkey did it" kind of defense would stand here.
The 'inputs' are the random numbers. Not the training data. They don't produce a copy of an existing work, any more than a human trained on looking at faces is producing a copy and 'stealing other people's lifework' (without even paying them a cent! how dare they), unless it's so similar to serve as a substitute (among other considerations).
> A better comparison would be a monkey creating a "director's cut" of Disney's Jungle Book, interspersed with footage from National Geographic, both of which are copyrighted.
That wouldn't be a good comparison at all, because a human could very easily identify exactly which frames are literally copied from which copyrighted work. That's not 'transformative' at all. In contrast, if you look at the faces, most of them don't have even a remotely similar exemplar in the dataset (you can do nearest-neighbor lookups using features or pixels, and GAN papers often do to demonstrate that they are not just 'memorizing' - I discuss this in my FAQ section).
> The 'inputs' are the random numbers. Not the training data.
You're dodging. The inputs to the neural network are clearly the training images, they don't just disappear. The outputs are derived works, at least.
> They don't produce a copy of an existing work, any more than a human trained on looking at faces is producing a copy and 'stealing other people's lifework' (without even paying them a cent! how dare they), unless it's so similar to serve as a substitute (among other considerations).
I'm not talking about the faces, I'm talking about CNNs in general. Surely these faces all look like generic anime faces, because that's what you trained on. On the other hand, if you trained on the works of Van Gogh with his very distinctive style, that would easily be discernible.
Also, the very fact that it is not a human doing this task is very relevant for copyright, as you point out yourself.
Note that I'm not saying this is illegal or that this violates copyright, I'm saying that there is no case law on this, so you can't just claim that there is no issue here. It's undecided. It comes down to the intention behind copyright. Furthermore, even if you can dodge copyright law, you may not be able to dodge trademark law.
> In contrast, if you look at the faces, most of them don't have even a remotely similar exemplar in the dataset (you can do nearest-neighbor lookups using features or pixels, and GAN papers often do to demonstrate that they are not just 'memorizing' - I discuss this in my FAQ section).
Again, your argument fails the general case. Overfitting and memorizing is a problem when working with CNNs. Some CNNs may just literally copy - and visibly so. At which point are you not just literally copying? What's the "legal threshold" here? It would have to be decided on a case-by-case basis, just like a "fair use" defense.
No, it doesn't. Again, the law is very clear about this. The US Copyright Office guidelines are extremely blunt: machines are not authors. Period. It doesn't matter if you twiddled with the hyperparameters or how much you paid for the electricity. (Similarly, you can spend billions of dollars measuring a physical constant, but as a fact, it is not copyrightable.)
I was wondering this too. Thank you for sharing this. I've had a similar idea with https://art42.net and wondering what type of copyright to add. Is this Creative Commons?
has anyone ever sued for using their dataset? Most publishers will not care enough, but museums are very protective of their art, and i can see them suing if someone uses a large collection.
Is it really a work of art if it doesn't include the obligatory ridiculous description? Something like "Opus #4, in mixed media. Through a re-imagining of the interaction between form and light, the artist challenges our preconceived notions regarding the interplay of history, science, and religion, transgressing the boundaries imposed by our own engagement in a patriarchal society, and forcing us to reconsider the role of art as a medium for change."
You nailed it. I mean, my first thought on looking at the output was it's shit.
But I'm pretty sure that with some meaningless verbiage generated from a corpus of art criticism (thisartcommentarydoesnotexist?) as a caption, it wouldn't look out of place in some modern art galleries and some collector would lap it up.
I have friends who are (modestly) successful fine artists, and they uniformly hate having to write that dreck. But you wanna pay the bills, you pick up a pen and write.
> it wouldn't look out of place in some modern art galleries and some collector would lap it up
Sure, if you were dishonest about its origin, and that says a lot about the pretentiousness of the modern "fine art" industry.
If you told them it took 0.00001 cents worth of electricity/compute power to create both the artwork and the description from an algorithm and some random numbers, I doubt most collectors would be prepared to lap it up at any price.
Charitably, you could say they are paying for the artist's ability to synthesize art from their experiences but then you could argue that the NN is doing the same thing but is just more efficient at painting. If that's the case then why should there be a difference in value? Are collectors just paying for the blood, sweat, and tears of a human artist regardless of what they produce?
If you're honest about it, and the customer is prepared to pay your asking price for the computer generated art, then it's no issue.
All collectors are individuals - you'd have to ask them whether/how much they are prepared to pay if they know the art is synthesized. All I'm saying is that pretending it's generated directly by a human instead of a human-driven algorithm is unethical. Whether they can tell the difference or not on their own is beside the point.
I know it's popular to hate on modern art, and your post will be the most upvoted (and I upvoted you too, so that my response would be more visible), but I want to offer an alternative view.
Art is about the artist attempting to generate an emotional response in an audience. Books, painting, music, are all there to create joy, sadness, melancholy, laughter, suspense, excitement.
Human emotions are universal, which is the beauty of art.
But humans themselves are not. What one person finds funny, another finds disgusting. What one finds sad, another finds pathetic. Which is why we need so many genres of art.
This is why we need both Classical Music, and Heavy Metal, and Country. Dramas and Comedies.
But what happens when art ceases to generate an emotional response? Does it even have value?
This is what happened to the classical art world at the turn of the impressionism. Classic "Renaissance"-style painting styles perfected the human form to the point that there was nothing exciting any more about the form. Nothing stimulating the viewer. And besides, photo-realistic representations of scenery was about to be improved and replaced with...photography.
So the artists went back to the drawing board, started challenging the status quo, and came up with Impressionism. Cubism. And a million other *isms that I, an uneducated rube, can't even begin to name or speak about with competency.
But each one isn't meant to be universal, to generate an emotional response in the audience. You can walk through a museum, and see a 100 paintings, and feel nothing. And that's okay! What's worth it is the one painting you see that causes you to feel something. Then it worked.
"Modern art" (which itself can be exploded into a million different categories falls into a bit of that. And it simply cannot be appreciated with a simple low-resolution gif on a monitor.
For one, it sits on the context of what came before. It responds to art that preceded it. You can get mad about that and think it should stand on it's own, but consider how much of mass media depends on an implicit understanding of the culture and what preceded it.
For another, it's frequently a 3 dimensional medium. Seeing the brushstrokes of an impressionist painting must be seen in person to truly marvel at the incredible ability to make something appear out of nothing ("This brush stroke is a dog. It's clearly a dog. but when you look up close, it's just a single simple brush stroke.")
Finally, and this brings me full circle to your point, sometimes you need to actively place yourself into a position to feel something.
Consider the difference between listening to a new song on earbuds on a loud subway train, vs on high quality headphones on your beanbag chair at home alone, vs with an audience in a magnificent concert hall. Your emotional response is going to be entirely different.
So when you see something like "Opus #4, in mixed media. Through a re-imagining of the interaction between form and light, the artist challenges our preconceived notions regarding the interplay of history, science, and religion, transgressing the boundaries imposed by our own engagement in a patriarchal society, and forcing us to reconsider the role of art as a medium for change.", you could default to "Wow, whoever painted this must be a giant gasbag".
You could scoff and say "My 4 year old niece could have painted that, and I wouldn't put it on my fridge"
You could say "What does this millennial know about the role of art as a medium for change."
Or you could stop, clear your head, and actually think about the words. How do you feel about society? How does the art before you make you feel? How does it feel to stand on the shoulders of giants?"
Maybe you'll feel nothing and move along, and that's okay.
But maybe you'll actually get some insights or an emotional repsonse from the piece, which is what the author hoped you would.
I agree with everything you've said. However, many people I know who produce art, are not thinking in these terms, and resent being forced to do so, and they are forced to do so, by the commercial reality of the fine art industry. Furthermore, if someone is trying to find meaning in the word salad I produced in about 90 seconds, I wish them luck, because whatever meaning is there exists entirely in their own head. I suspect the same is true for a great many (not all) descriptions of modern art.
What I'm asking of you is to be less cynical about what actual art and artists and be genuine in the search for meaning.
If you don't find it - no harm.
Also, the meaning in the author's own head is irrelevant. Death of the author. You can find your own. Maybe their guidance and perspective helps there, maybe not.
"What I'm asking of you is to be less cynical about what actual art and artists and be genuine in the search for meaning."
There's a typo here, and I honestly am not sure what you're trying to say. I will say this: I am not a modern art hater. Far from it, I love modern art, and I go well out of my way to enjoy it. My complaint is entirely with the ham handed descriptions of pieces that require no descriptions at all.
You're right, I definitely accidentally a word there, may be several.
Ironically, I DO hate most modern art, and the treatise I wrote out is a description of my own journey, and what I frequently have to consciously remind myself to allow myself to experience and feel things that I don't have an immediate connection to.
Totally get what you said IF the artists of this day and age weren't shilling for money. If it was pure art that was an expression of their self, sure, I agree it's art. But peddling mediocre crap for large sums of money should be equated to trash.
Most of the discussions around modern art is around who sells for the most amount of money. It's a numbers game, gaming galleries which exist for no other reason than to peddle crap possibly for tax reasons, and con artists posing as real artists...
I agree with everything here except the suggestion of trying to find meaning in International Art English, which is abstruse, opaque, hifalutin, and divides The Anointed Ones of Art from the (they would say--but they wouldn't say) unwashed masses _by design_.
International Art English is, from a virtue ethics standpoint, a pure waste of time and pure crap.
> Classic "Renaissance"-style painting styles perfected the human form to the point that there was nothing exciting any more about the form. Nothing stimulating the viewer.
I don't think that follows at all. My favourite cafe has almost perfected my morning Americano, but that's exactly why I enjoy going there. The TV series that provoked the strongest emotional response in me wasn't doing something nonrepresentational, it was doing exactly the list of tropes that you'd expect to provoke that emotion, just executed in a really polished way.
And even - especially - if we talk about pictures, there have been plenty of representational works that left me awestruck by their beauty. Look at, say, Sparth's spaceships (which couldn't possibly be replaced by photographs, even though they're "photorealistic" in some sense). Heck, look at Luis Royo, or the dragons on my college friend's deviantart. Even if you just want to talk about photorealistic oil paintings, take a look at the aeroplanes and board meetings in the China Art Museum and tell me those don't make you feel something. The things that people found beautiful and inspiring and provocative for centuries can still be all those things, and the idea that everything that can be done with representational, conventionally beautiful artwork has been is as absurd as the idea that everything worth doing in fine dining or orchestral music or classical ballet has been done.
Humans are good at pulling meaning out of noise, even when there's no actual meaning there - consider rain dances, or the face of Jesus on a shower curtain. Of course ambiguity has always been a part of art - the famous Mona Lisa smile, or a lot of poetry rests on things that can be taken in different ways by the reader. I'm sure it's possible to look at a random pattern of brush strokes and have a meaningful emotional response - heck, I'm pretty sure I've done it myself. But surely there should be some equivalent of a double-blind test; art that performs no better than a placebo should not be considered as art, just like in any other field. And I'm convinced a lot of modern art would fail that test - e.g. Damien Hirst tells a story of how he tried to deliberately do one of his spin paintings badly, but ended up with one that was indistinguishable from the others. All of the art forms I talked about in the previous paragraph have a better hit rate at provoking interesting emotions and insights than modern art (and I don't think it's just a matter of accessibility - classical ballet takes a fair bit of effort to be able to appreciate it fully). So it seems crazy that the art world is so exclusively, overwhelmingly focused on non-representational, looks-like-random-noise works.
If modern art really was just a scam, what would you expect to be different? What would it take to convince you that there's no there there?
Yes. So often I've seen art that is completely devoid of meaning without the label, and it isn't obvious if the label is supposed to apply to a nearby pedestal or the wall it is affixed to. The label is the actual art work.
It is right and good to poke fun at artists describing nonrepresentational work as being about something. It's endemic in contemporary art music as well.
One can still dig the work itself, though. Forget the words around it. Talking about art is like dancing about architecture.
Now that's a hard problem to solve. AI generated text is usually more readable. I don't know of an algorithm that can dumb it down similar to your example.
It's clearly about depression. Look at the drab blue and the computer's lack of motivation to continue after one stroke. It knows it needs to get back into painting to snap out of this feedback loop of sadness. But even one brush stroke in and it feels like a pointless endeavor.
Of course not all art has to be pleasing to the viewer, but confronting/challenging art is made deliberately that way by the artist, and the process of experiencing the art involves understanding the feeling and learning something from it.
It's a different feeling altogether to be discomforted by art, but for no purpose.