AI in Art
Andy Warhol, the great painter, once said art is everything you can get away with. Now, art has taken a step forward towards breaking the boundaries yet again. Once upon a time, using a camera to perceive more details in the painting was considered the cutting edge of technology. Today, artists are using artificial intelligence and machine learning to either change their work of art or to create totally original paintings.
But the question is this: can these artists get away with it? Is a work of art created using artificial intelligence still considered art? Can critics and the audience accept the paintings made in this manner as creative? Can they even tell machine-generated works and human-created ones apart?
Yes, Yes, and No. recently, several exhibitions and galleries have shown paintings made with the help of AI, and the public reaction has been quite positive about it. These paintings are even sold at exorbitant prices in auctions and private events all over the world. For instance, in an art exhibit composed of both works generated by people and AI, visitors could not distinguish which one is which. At the end of the show, one of the software-developed paintings was sold for 16,000 dollars.
How exactly does artificial intelligence create a unique work of art?
Artists work with the raw material they’ve got. This raw material is made of their daily experiences, the other works of art they’ve seen, and their feeling about the world. The distinguishing feature of an artist is how they try to make sense of the world around them.
This is what artificial intelligence does too. One way AI works to paint something original is by using the pictures the artist/programmer has fed it. For instance, an artist fed his program 500 years’ worth of portraits and asked it to make one that’s not the same as any real face.
The other way of using AI in art is to create the work and then give it to the AI machine. The AI software creates images with many different perspectives, angles, and lights. The possibilities are endless. The final work would be a combination of the original work created by the artist and the variations created by the software.
Sometimes the data fed to the software is a combination of images, sounds, and even newsletters. The machine is supposed to create an original work using all the information it has. In this method, the final product is also a combination. For instance, one can mention Rafik Anadol’s projection on the Walt Disney Concert Hall. This projection was made using 52K images and projectors with remnants of the concerts performed in the hall previously.
What are some of the AI systems that create art?
Generative Adversarial Networks (GANs)
GAN or Generative Adversarial Networks is a system in which the AI is designed to find patterns and regularities in the data it is given. Then create new examples that could come from the same data by applying the regularities. The generative model in this system is made of two sub-models: one that’s supposed to create the new example and the other one that judges the example as real or fake.
This system is extensively used in AI painting because the system uses a set of real-life images as its model to create something original and creative. On the other hand, the work of art created in this manner should not be too outlandish that the audience cannot make sense of it or emotionally connect to it.
Ian Goodfellow is the person who came up with the idea on a sleepless night. He has worked for Google as a research scientist and a senior research scientist. Currently, he is working for Apple’s Special Projects Group, a company that produces hardware solutions for various industries.
Artificial Intelligence Creative Adversarial Network (AICAN)
AICAN is a patented AI system that has managed to pass the Turing test. In 2006, the artwork created by AICAN was mixed with some abstract works by human artists. The visitors of the exhibition could not tell the difference between the human-made works and the AI-made ones.
This system is designed by Ahmed Elgammal, professor of computer science and director of the Art and Artificial Intelligence Laboratory at Rutgers University. He fed the system 80,000 previously made artworks by some of the greatest artists in history.
However, the goal was to make AICAN genuinely creative. Therefore, AICAN is designed to break the styles and create something new, not just to follow the current techniques in art. This is how AICAN differs from GAN, which simply follows what it is fed. At the same time, just like GAN, the created work is not supposed to be too shocking by going too much out of style.
Aaron is a fifty-year-old robot created by Harold Cohen. This robot was the very first example of artwork created by artificial intelligence. At the beginning of the process, the robot was able to draw childish forms. As the robot’s drawing progressed, the pictures started having biometrical shapes, which Cohen painted himself.
After twenty years of tireless work, the robot learned how to create color painting autonomously. The only information Cohen gave AARON was an abstract set of descriptions of objects, body paints, and animals.
Over the years, AARON’s painting style grew, and the size of the paintings it created immensely varied. Some of the pictures were small and letter-sized, and the others were huge. AARON even had a mural painting several meters long and wide. It even had a turtle robot that painted the floor in the exhibit it had.
How large is the market size for AI-generated art?
The market for AI-generated art started to show itself to the world as a potential for profit when it sold for 45 times the expected price at a Christie’s auction. The painting, generated by GAN and it’s called The Portrait of Edmond Belamy. This seemingly unfinished portrait sold for $432,500 in 2018.
This was a significant moment in this history of AI-generated painting because it proved to the world that this art style is as authentic as any other, and it’s here to stay. What’s for sure is that this technology has just started on its track and has enormous potential for development. With its growth, its market will grow too.