Christie’s will auction an artificial intelligence (AI-created) artwork this October. This comes right after a landmark art exhibition at the Nature Morte gallery, New Delhi. These types of artworks raise new questions about who owns and author of these works, when it becomes obsolete and what are the areas where algorithms can be used.
A lot of creators of AI art use adversarial generation networks (GANs), which allow computers to analyze a library of images and sounds, create their content, then try again. You will learn new things and improve your skills by trial and error.
These artworks, which result from the connection between artificial neural networks, include paper prints, videos and multimedia installations, are often hauntingly realistic.
Mario Klingemann, a Munich-based algorithmist, trained it on Old Masters portraits before exposing it for images of him. This results in a video that has many melted eyes, often being compared to Francis Bacon’s works.
Memo Akten, a London-based Turkish artist, was the first to sell an artwork using artificial intelligence. It was sold at $8,000 at a Google charity auction held in San Francisco in 2016.
Christie’s will auction its first AI artwork two years later: Portrait of Edmond Belamy (2018) by Parisian collective Obvious. It is expected to fetch between $ 8,000-$ 11,500.
The art of AI can be described as a nascent art form that is trying to get into the art market. Aparajita Jain is co-director at Nature Morte. She says that the prices for works in the recent “Gradient Descent” exhibition were “quite aggressive,” between $500 and $ 40,000 to establish AI art as an actual genre. This is a significantly lower entry price than the usual range of $ 10,000-$ 100,000 at the gallery.
Tom White, a Wellington artist who sells Nature Morte pieces, created Kandinsky-style summaries using AI’s understandings of everyday items such as electric fans and binoculars.
Jain claims that the show attracted new audiences, which suggests that AI art might help grow the market. White says, “I have seen many unusual art collectors purchase my work, including scientists and video game creators.”
Nature Morte stated in the press materials of “Gradient Descent” that the works were created “entirely using AI in collaboration with artists.” They even signed the work with the mathematical equation they used instead of the collective name.
While artists and gallery owners may like to claim authorship to AI and stress that they can’t predict what an AI algorithm will produce legally, there is no question who it is.
According to Jessica Fjeld (deputy director of Harvard Law School’s Cyberlaw Clinic), AI is a tool artists use in the same way that a photographer uses Adobe Photoshop or a camera. It reads:
“Humans are involved in every aspect of creating and training AI technology today and tomorrow. It is not the software that could be considered property, but who among these people will acquire the rights to the works. This question for me is much more intriguing.
Mason Kortz, Fjeld’s research partner, identified four essential elements in the art and science of AI. Each of these involves copyright in different ways. These are:
All artworks in this article can be sold as products, including prints, videos, and installations. Any copyist attempting to resell these products would be violating the copyright of the human artist, just like if he had copied an oil painting or a photograph without permission. The art of AI presents new challenges.
Fjeld asserts that AI artists can create their own algorithms (items 2, and 3, much like White). It reads:
“The code could be sold as a work of art by the artist, but I’m not aware of this happening.”
It is an intriguing idea, however, that collectors might find appealing. They could use an AI artist for their productions.
It could be difficult to preserve the code’s original intent, particularly when working with proprietary software or hardware.
Harshit Agrawal (participant artist in “Gradient Diescent”), who lives in Bangalore, says that “one of the major maintenance problems is the software structures that quickly update, making the trained neural networks models redundant over the time.”
Akten is particularly interested in jobs that incorporate web technologies.
Things like Google Translate or sending a query at Microsoft’s facial recognition cloud API or using Amazon Cloud services or jobs that live on now-defunct Vine. I know of quite a few jobs that have been ‘died’ because an API in cloud has changed or disappeared.
AI is like a series of shows. This solution can be seen as a way to look at it. They work as long as technology permits them to, then they vanish. We are left with the documentation, the memories.
AI artists rely heavily on public domain images and audio libraries to train their algorithms. ImageNet, SoundNet and Google Art are some of the most popular examples.
One reason is that copyrighted images used in training courses (item 1) can lead to results that are very similar to a particular image. Fjeld says:
“I don’t know of any copyright lawsuits, but I believe these will eventually be filed.”
Karthik Kalyanaraman, curator of “Gradient Descent”, claims that AI artists do not copy images or sound. This means they should be able to learn from copied images in the same way that art students can learn from books and visits to the MoMA.
Fair use could be used by legal defense artists if they have training sets that contain copyrighted material. But, “just to keep things clean, pragmatically, and I have insisted that images in the training set that are for the works in the exhibition aren’t protected by copyright,” he states.
Anna Ridler, another “Gradient Descent”, artist is even stricter on copyright. She uses her own sketches, photographs, and training sets.
It is the construction of a database, deciding what to include and what not, that makes it a creative act. This is a significant part of the work. These databases are in a sense works of art by themselves (I made them), so it will be nearly impossible for anyone to duplicate my work.
Artists who wanted to use their own algorithm or training materials in their art would need to discuss this with copyright holders.
AI art is not a threat to human artists’ livelihoods. AI artists are able to create their own work as long as they use open-source algorithms and training sets or ones that they created. The rise of AI art will have far greater long-term consequences for the art market.
Kalyanaraman believes it can change traditional art in the same way that photography revolutionized painting. This led to Impressionism and Expressionism and other schools who are more interested expressions of emotion and human perception. Insofar that this art is an expression of a description, it could be created by human artists using AI to create new forms of painting. Collectors may lose interest in works that are not new or cannot be described, such as Piet Mondrian’s.
Kalyanaraman is a collaboration between Mark Rothko (paintings) and Paul Klee (paintings). The first engulfs the viewer in a tsunami, while the second tickles the feet – this is the kind of artist that will last. This is what it says:
“All of our perceptions are linked to our emotions. This kind of thing will be more difficult to represent using an algorithm.
Hi, my name's Craig Malloy. I'm a tec blogger. Well, actually, I'm a computer analyst. Sounds boring, but the background behind it isn't. I work for a firm in South Carolina and in my spare time I like to write about technology. Actually, I like to write posts and publish them on my blog all about technology.
I studied computing at University and have progressed in my technological career ever since.
I hope you enjoy reading my tec and science news stories. Check them out!
