The Art of Artificial Intelligence: Make your own astronaut in Rembrandt style

in the novel 1984 By George Orwell, Winston hears a woman sing a song through an open window. It is a song made by a machine, which Orwell calls a “hair-cutting machine”. There was no human hand involved. Orwell says: ugly music. But the woman sings it so beautifully that it is almost bearable. he reluctantly admits. That song is a hope in Orwell’s dystopia. The woman-singer is not called the writer, she is old, fat and ugly. Is she also an artist? Orwell doesn’t think so. I’m still thinking about it.

Because this woman—let’s call her Eve, after the first woman in the Bible and a robot named after that first woman in Pixar’s Wall-E cartoon—came to mind when I first opened the computer program Dall-E (named after artist Dalí). Dall-E is a miracle in the form of a computer program that can convert text into images. For example, click on “astronaut on a horse” and a few seconds later an astronaut on a horse will appear. You can also specify the type: should it be a drawn, drawn, or depicted astronaut on a horse? And within that, countless variations are possible: painted in the style of Van Gogh or Rembrandt and much more. The possibilities are amazing.

Along with Midjourney and Stable Diffusion, Dall-E is an AI art generator that started making waves last year. They have a huge stock of images collected on the internet that the algorithm can collect in all kinds of ways. In this the chatbot is similar to ChatGPT, an equally amazing piece of software that can prepare readable text with the help of AI from a few words. “Write an article about Dall-E, artificial intelligence, Vermeer and lemons,” I asked the chatbot, because I was busy making life with lemons on Dall-E in the style of several artists. Dall-E made this suggestion: “With Dall-E it is possible to create Vermeer-style images by providing a text description of the desired image and selecting Vermeer as the desired style. For example, a description such as ‘a lemon on a table with light coming in from a window’ can be used to create A portrait in the Vermeer style, in which the lemon and table are rendered in very realistic detail and the lighting and shadows of the scene are captured with the same level of skill as in one of Vermeer’s own paintings.”

there he is. Boring but true. Possible.

Varying the command always results in slightly different text, though sometimes the bot took my prompts too literally. A request for a lyric essay resulted in a poem, which rhymes unsolicited. Dall-E sometimes has a hard time asking people questions that seem simple. The chatbot popped up when I asked for a creative essay featuring Vermeer and Lemon with the following proposition: “An example of DALL-E’s creativity is the ability to create a unique image of Lemon with a human face, complete with eyes, nose, and mouth, all made up of the textures and shape of the lemon itself.” This type of creative output became made possible by the advanced DALL-E engine learning algorithms, who are able to analyze and process massive amounts of data to generate new and imaginative ideas. In the case of the human-faced lemon, DALL-E was able to draw on his knowledge of human facial features and his understanding of the lemon’s characteristics to create an image that’s both visually stunning and highly imaginative. “

Naturally, I put it to the test. But with Dall-E, creating a lemon head like this was much more difficult than a chatbot made. The lemon with the human face was above all a lemon with the human face. It wasn’t until I added the word surrealism that it started to sound like something. And it wasn’t until I let Lemon take a Magritte-style selfie that Lemon really came up with a human face. Perhaps Charles Philippon did it faster and better in the 19th century, turning his cartoon of King Louis Philippe into a pear. But he can draw. not me.

Bianca Steiger, “without an address”. (AI art created by Bianca Stigter using Dall-E)
Bianca Steiger’s photo

lemon art

But is it art?

This is the main question for many who are faced with the art of artificial intelligence.

who cares?

If a musical instrument produces a beautiful song, does it matter that it was made by the instrument? Perhaps a new type of artist will appear, someone who is particularly good at writing prompts, AI commands. My lemon allergy indicates that it is not that easy. When American artist Karen X-Cheng B worldwide The first AI-powered magazine cover in June 2022, and it boasted that it only took 20 seconds to achieve this result. But in fact, artist Karen X-Cheng spent days writing Dall-E the right router. “Hours and hours of creating claims and refining them before you get the perfect shot,” Cheng says on LinkedIn. “Female Astronaut Walking on Mars, Digital Art” and “Female Astronaut Walking on Mars, Digital Art” did not produce the desired image. It just hit the mark with: “A wide-angle, bottom-up shot of an astronaut with an athletic female body walking swagger towards the camera on Mars in an infinite universe, synthetic wave, digital art.”

Surrealism of the third degree

A kind of cheap surrealism of the third degree is the highest result of ordinary people. The game gets bored quickly. How many seals can a girl with a pearl earring and a surfing rat take? The website encourages this sense of fun by offering things like an avocado-shaped armchair or a Van Gogh-style painting of an American football player. The algorithm found it difficult to deal with the surreal saying “as beautiful as the chance meeting of a sewing machine and an umbrella on a cutting table”. First of all, the cutting table didn’t make it to the Dall-E ballot. The word does not fit into their “safe system”.

Can’t wait for the first AI masterpieces

I was surprised by that, because racism and sexism seem to be ingrained in the algorithms. White men sitting in front of Dall-E. If you don’t explicitly ask for something else, even a lemon is a white man. Fat old women a la Eve Orwell out of the question. Lensa app, which is primarily used to create user avatars in styles known from popular culture – a no star Wars or game of thrones He made me unattractively younger and gave me bulging breasts that you could use as a weapon. American artist Cindy Sherman has been able to create outrageous images with the help of – presumably Lensa – ugly in all kinds. Eve Orwell would be glad of that.

American Jason Allen worked for several months with AI Midjourney, resulting in the artwork, among other things D’opera Spatial Theatre, which he printed on canvas and submitted to the Colorado State Fair Fine Art Competition in the Digitally Manipulated Photography category. Allen won first prize. It is not clear if all members of the jury are aware of the creative process. Regardless, many critics felt Allen cheated: “It’s like running a marathon in a Lamborghini.” Others objected: why ride in a carriage when there are cars? A compromise between all those racetrack comparisons: People still ride horses. They also continued to paint after the invention of photography and cinematography. And what great unreleased paintings.

Bianca Steiger, “without an address”. (AI art created by Bianca Stigter using Dall-E)
Bianca Steiger’s photo

Can’t wait for the first AI masterpieces. Until now, he often works in another style, not only by old masters like Van Gogh and Rembrandt, but also by modern masters of digital art like Greg Rutkowski, whose works are mentioned the most in the claims (about copyright and payments nothing is said. arranged after with artificial intelligence generators).

Pioneers

The art of artificial intelligence now has a long history of its own. Among the pioneers was the British mathematician Ada Lovelace, who already in 1842 conceived of a machine that could not only calculate all kinds of problems, but also solve them. I called it “the science of poetry.” The first true AI art was made in 1973 by American scientist and illustrator Harold Cohen, who had AARON software independently produce abstract artwork. In 2015, Alex Mordvintsev developed Deep Dream for Google, which in turn understood visual concepts and was able to create hybrid images based on a massive data set, like a dogfish and a camel bird. In 2018, the French art group Oburred fed an algorithm with 15,000 images obtained from WikiArt on which it made its painted image. Edmond Bellamy’s photo (which never existed before) was printed on canvas and sold at Christie’s for almost half a million dollars. The signature in the bottom right of the board is part of the algorithm used to create the board.

So not all artists use large corporate data sets like Google or OpenAI, the company behind Dall-E and ChatGPT, which still suggests software as a tool. New York’s MOMA is displaying a work by Turkish-American artist Rafik Anadol who fed images of 138,000 artworks from the museum’s collection into a computer program, which now produces its own images based on that collection. „What could the machine dream about after seeing the Museum of Modern Art’s collection?Anadol wondered. It is titled work Uncensored. So the machine there is more than just a tool. Is this machine much different from the 1984 copier? Eve then still needed to sing it. Soon we’ll just have to listen. or to watch. Or maybe not that.

Leave a Comment