Slashdot Mirror
Artificially Intelligent Painters Invent New Styles of Art (newscientist.com)
Dthief shares a report from New Scientist: Now and then, a painter like Claude Monet or Pablo Picasso comes along and turns the art world on its head. They invent new aesthetic styles, forging movements such as impressionism or abstract expressionism. But could the next big shake-up be the work of a machine? An artificial intelligence has been developed that produces images in unconventional styles -- and much of its output has already been given the thumbs up by members of the public. The team [of researchers] modified a type of algorithm known as a generative adversarial network (GAN), in which two neural nets play off against each other to get better and better results. One creates a solution, the other judges it -- and the algorithm loops back and forth until the desired result is reached. In the art AI, one of these roles is played by a generator network, which creates images. The other is played by a discriminator network, which was trained on 81,500 paintings to tell the difference between images we would class as artworks and those we wouldn't -- such as a photo or diagram, say. The discriminator was also trained to distinguish different styles of art, such as rococo or cubism. The clever twist is that the generator is primed to produce an image that the discriminator recognizes as art, but which does not fall into any of the existing styles.
Easier way - just give a kindergarten class a load of paint.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
To be complete, it still needs to explain how it decides to paint what it painted using emphatic words...
Now the robo workers that will replace us in 20 years have something to spend their money on.
sudo rm -r -f --no-preserve-root /
This will produce new art because humans and AI will get new ideas from each other, then produce results and those results are fed back in to the creative loops.
Somewhat similar effect can be seen between tabletop games and computer games. Both 'platforms' benefit from the design developments done in either camp.
And there is no better noisy environment than "high art" where paint by numbers picture might win the first prize much to the embarrassment of the officials, and museums mount art works upside down unbeknownst to the patrons as well as the artist!
AI pitted against humans in seeing patterns in noise, is probably a high point, acme, zenith of intelligence. What next? Illusions of grandeur?
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
It needs to create "art" based on personal emotional experience AND induce such emotional experience in human audience.
Otherwise... it's just a drawing, photo, sculpture, video... but not art.
Just like those "paintings" by monkeys and elephants are not art but paint slapped on canvas.
Or like how birdsong is not art, an anthill is not architecture and dogs urine on the wall is not graffiti.
Mit der Dummheit kämpfen Götter selbst vergebens
The clever twist is that the generator is primed to produce an image that the discriminator recognizes as art, but which does not fall into any of the existing styles.
Yet every example I saw on that page was abstract.
This will produce new art because humans and AI will get new ideas from each other, then produce results and those results are fed back in to the creative loops.
That's not necessarily a good thing. Flood the market and exclusivity doesn't mean much anymore. The market gets saturated and the users satiated.
Somewhat similar effect can be seen between tabletop games and computer games. Both 'platforms' benefit from the design developments done in either camp.
Which is a prime example of over-saturation. People spend less time on both tabletop games and computer games now, and classics like Monopoly, (A)D&D, Doom and Half-Life just don't happen anymore. Or rather, they happen, but they are a dime a dozen and don't stand out.
What art does is convey feelings. So far, machines have none.
It's like a mountain - it may be beautiful, but in itself, it is not art.
A painting of the same mountain, or dance for it, is an attempt at letting an audience catch a glimpse of the feelings the artist had for the mountain.
The difference is that with human-created art you think "what was the fuckhead smoking?", but with computer-created art it's more like "what was the fuckhead who programmed it smoking?"
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
Machines judging art.
What art does is convey feelings. So far, machines have none.
If you lay out a hundred abstract paintings, half made by humans and half by this GAN, do you think you could tell which were made with "feelings"? I doubt if you could do any better than chance. It is silly to say there is a difference if the difference is undetectable.
"One creates a solution, the other judges it "
Wouldn't work for natural intelligence artists, they think every other artist's work sucks.
"When I'm a fan of something (art, music, or any sort of creativity) part of what makes the experience pleasant is showing your appreciation to the human artist that created it."
Good luck trying that with Homer, Velazquez, Michelangelo, Bach...
If *really* not being able to show your appreciation to the author ruins your experience, I have to say you are already losing a bit too much of what people generally considers as art.
Thankfully I now have an article to link when people tell me how development can easily be automated but "creative" jobs can't be.
The AI did not actually create the art. The programmer gave the AI the ability and method for which to create the art, therefore the programmer is the artist and the AI is simply his brush. AI does not exist without the programmer. If it ever does, its name will no longer contain the word artificial.
What art does is convey feelings.
No, what art does is generate feelings. An artist may be able to use this mechanism to convey them, but I think in lot of cases -- especially in modern art -- there's very little conveying going on. The artist just creates something that generates reactions in viewers, reactions that may have nothing to do with the artist's intentions, and may vary widely among viewers.
So far, machines have none.
Which doesn't in any way mean they can't generate images which provoke emotional reactions in human viewers, or even that we couldn't apply machine learning algorithms to make them better at doing exactly that... or even generating particular kinds of emotions. It's not necessary to have feelings to create them in others.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
Sure, let's just twist "intelligence" and "art" to mean nothing.
The machine has no will, no soul, no innate purpose; therefor no intelligence.
It has nothing to express, so it can't make art.
This kind of thing just promotes soullessness. It's a literal, physical opposite of art. Instead of inspiring you toward living life, it 'inspires' you to be complacent with death.
Sure, the idiotic researchers making this poison are probably 'innocent' and just think they are 'progressing science', but the people funding them know what this is and are wholly malicious. Our culture has been hijacked since the end of the World War, and so few people seem to have noticed.
Post-modernism is a weapon to destroy us all, as foretold by many. If only the middle class used their wealth to improve themselves instead of pissing it away because 'life is so hard I worked so hard I deserve it". I guess that's what happens when you have an anti-meritocracy to suppress competition against the plutocrats and give a bunch of morons money.
What art does is convey feelings. So far, machines have none.
If you lay out a hundred abstract paintings, half made by humans and half by this GAN, do you think you could tell which were made with "feelings"? I doubt if you could do any better than chance. It is silly to say there is a difference if the difference is undetectable.
So we've demonstrated that abstract art is BS ... not sure what that has to do with AI though ...
No, what art does is generate feelings.
Please read more than the first sentence before hitting reply. I give an example of why this, specifically, is wrong: A beautiful mountain generates feelings, but the mountain is not art. A painting of the mountain, or an abstract painting attempting to capture some of the feelings the mountain gave the artist is art.
Way to forget the fundamental theorem of computation on the equivalence of data and program.
Program: A few grad students over a few months, on top of program libraries composed by a few hundred grad students over a few decades, on top of general-purpose computational abstractions as devised by a few thousand notable wonks over a century, on top of a hundred primary patron saints of abstraction inscribing circles in the sand over roughly three millennia.
Data: Millions of artist years, as winnowed down by billions of critic years, to thousands of artist years, on top of human visual perception over the whole 60 million years of human evolution, and this not from a standing start.
I'm not 100% certain I trust this long historical wall to be less than entirely porous, but if I did, I'd declare: advantage, data.
This whole business of glorifying the last touch is a weird human institution to begin with. That's how you know the real future has finally arrived: in the sudden discarding of the forever unjustifiable, if only you've got the wits to see it.
This credit-to-the-last-touch business has always been tainted by the aggregate shoulders of giants.
Now that we've reduced stirring the aggregate down to a semester project in graduate school, we need to park this ridiculous cultural legacy, pronto.
Remember "information wants to be free"?
By the late 1970s—if you had the wits to see it—certain premises of our long standing copyright system were coming into fatal contact with a new reality, a future long past the original paroxysm that continues to arrive in sluggish fits and start:
Wall Street Journal To Cut Back Print Outside the US
As they say in cryptography, attacks only ever improve. This, too, of our generative-art adversarial networks. Unlike human talent, this isn't a flash in the Picasso pan. This is remix culture on steroids passing through the inception stage of a one-way function, up escalator.
I think I've posted this link before. It was one of my favourites, and good background for the present discussion:
Talking Machines interviews Doug Eck on Generative Art and Hamiltonian Monte Carlo — August 2016
Please read *my* whole post before responding. Whether or not the artist intends to convey something, or successfully conveys it, is not really important to the perception of the work as art by the viewers.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
The question is, does a donkey's tail count as an early form of AI? http://aworldelsewhere-finn.bl...
Van Gogh set the art world on its ear.
Contribute to civilization: ari.aynrand.org/donate
If your point is just that "abstract art is usually bs anyway so it's easier to imitate", then I agree with you.
No. My point is that an intelligence (AI or BI) should be judged by what it produces, not on some subjective qualification like "feelings" or "has a soul".