Have you ever seen a picture of Elon Musk dancing under the rainbow in a clown costume? Or a teddy bear kicking a football under water? Well, neither did I – until I tested image generators with artificial intelligence. Generators such as DALL-E, Imagen and MidJourney have managed to fascinate the whole world during their short lifetime. Simply explained, this is a type of tool that can generate images based on an optional text. All you have to do is enter a description of what you want to depict – and wait for the generator to do its job. The result you get is a completely unique image that has never existed before. Say hello, Elon: – Want to revolutionize the design field One of those who have allowed themselves to be excited is leader and researcher Klas Pettersen at NORA.ai, a national collaboration for research, education and innovation in artificial intelligence. – These tools have become so popular simply because they are insanely good, he says. He explains that the generators are based on highly advanced algorithms that have been trained by being fed billions of images and text descriptions. Eventually, it has enough data for it to understand the connection between different types of text and images, and can thus create completely unique images based on this information. Here are some examples of what DALL-E achieves: “Surrealistic glass sculpture in the style of Salvador Dalí” (Surrealistic glass sculpture in the style of Salvador Dalí) Photo: Xin Li/DALL-E “A metallic sculpture of a corgi inside a gallery” (A metallic sculpture of a corgi inside a gallery) Photo: Xin Li/DALL-E “Woven tapestry of two Pikachus” Photo: Xin Li/DALL-E “Robotic Prada heels” (Robotic Prada heels) Photo: Xin Li/DALL-E The researcher believes that we will see image technology being used in many different ways and areas in the time to come. – I am very impressed by what you can achieve as an amateur with these tools. One thing we see immediately is that this is going to revolutionize the design field, he says. Can these image generators be creative themselves? – Many people think no, but I definitely think yes. When an algorithm creates something based on everything it has learned that has not been shown before, I believe that you are close to what we call creativity. – We are in the early stages of this type of technology. It will be extremely exciting to see what happens over the next few years, says Klas Pettersen. “A greek sculpture inside a burning room” (A Greek sculpture inside a burning room) Photo: Xin L/MidJourney Who owns the photos? The new tools also bring with them a number of legal and ethical pitfalls. Among other things, the question of who has the rights to a work produced by an algorithm, which in turn has based its effort on existing images. Both Google and Elon Musk-owned OpenAI are among the companies developing the image generators. The latter develops DALL-E, which is so far the only tool that gives users full usage rights to the images they “create”. Another concern is the misuse of images to spread false information and deepfakes. The companies have therefore introduced guidelines that prohibit users from producing images that are harassing, violent or of a sexual nature. This person does not exist. Here was the phrase “realistic portrait of a woman” (realistic portrait of a woman) Photo: Xin Li/DALL-E For the time being, the most advanced image generators have not been opened up for sale and use by the large masses. Until then, you must sign up on a waiting list to gain access. There are simpler free tools available such as Craiyon, which has also made its mark on social media. This was the result when we asked it to give us a picture of an airplane painted by Edvard Munch: Image generator Craiyon’s interpretation of “Edvard Munch-painting of an airplane”. Photo: Xin Li/Craiyon news’s art critic creates AI art Those who were early adopters and got access to the image generators have already started sharing recipes for keywords that give the best possible results. The most impressive images are highly detailed and accurate. Many therefore ask themselves the question whether the new technology can change the way we create art. Already in 2018, a single artificial intelligence-based work was sold for over 400,000 dollars. Another question is how artistically good these generated images are. To get a little closer to the answer, we let our art critic Mona Pahle Bjerke try the image generator DALL-E for the first time. She asked for a cubist version of the Madonna and Child. Here is the result: Photo: Mona Pahle Bjerke/DALL-E – This was very interesting. – It is just as if the generator has not fully understood what cubism is. This looks a bit more like a helpless stained glass window from the early Middle Ages, so it fell a bit short here, says Pahle Bjerke. But adds: – It’s incredibly fun to use. I see the fascination. news’s art critic believes that today’s art world no longer uses craftsmanship as a yardstick for what constitutes good art. It is the idea behind the art that matters. – Then it doesn’t matter if you’ve got another person, or a generator with artificial intelligence, to produce and visualize the work for you. Do you think AI art is good art? Pahle Bjerke, on the other hand, does not believe that the image generators will replace more traditional ways of making art. – If you have talent as an idea creator and the only thing missing are the craft skills, then this can be a tool in art production. But I don’t think that people will be able to use this as a shortcut and “cheat” their way into the art scene, she says. “Impressionist oil painting of a herd of rabbits in a field of flowers” Photo: Xin Li/DALL-E Want to read more art? Check these out:
ttn-69