The artificial way of being creative
Avantika Bhuyan | 05 Sep, 2018
AS ONE NAVIGATES Nature Morte, Delhi, each level reveals one unique work after another, with the eye coming to rest at The Anatomy Lesson of Dr Algorithm clearly inspired by Rembrandt’s masterpiece The Anatomy Lesson of Dr Nicolaes Tulp. A panel of 20 prints on archival paper, it features vivid abstracts—some bursting with angry crimson hues, others featuring more muted reds in delicate floral-like patterns. It elicits a gasp from nearly every viewer, not just because it is so striking, but also because it has been made in collaboration between man and machine, with the final image having been generated by Artificial Intelligence (AI). Rembrandt’s painting was created at a time of troubled fascination with medical technology. ‘Today, we are in a parallel time of troubled fascination with AI,’ writes Bengaluru-based Harshit Agrawal, a new-media artist and human-computer interaction researcher, in a note about the work.
It is this ‘creative dialogue’ between human and artificial intelligence that runs like a thread through Gradient Descent, the first group shows of AI art by a mainstream gallery in India, curated by 64/1. It leads us to a few larger questions: how much of a human should a machine be exposed to? If machines can create, then what is left for the human to do?
The show at Nature Morte also prods one to cast a deeper look at the world of AI art, which is made up of a small and closely-knit, albeit growing community. “There are possibly only a few hundred AI artists across the world, and very few in India,” says Agrawal, over the phone.
Though united in their engagement with technology, the interests that fuel each artist’s work are manifold. Mario Klingemann, one of the best known in the field, is a Munich-based artist, neurographer, coder, data collector and artist-in-residence at Google Art. A self- proclaimed sceptic with a curious mind, he uses neural networks and algorithms to explore deep learning, generative and evolutionary art, glitch art and data classification. His engagement with the digital archives at the British Library or the Internet Archives, has been especially significant. Then there is Memo Akten, a PhD candidate in AI, based in the UK, whose interests lie in the collisions between nature and science, technology and ethics, ritual and religion. Nao Tokui, a Japanese media artist and a DJ, is another interesting artist working in this field. Recently, he presented live performances, featuring an AI DJ, playing alongside a human one, often it was Tokui himself.
An exploration of this genre would be incomplete without the mention of Obvious, a Paris-based collective, consisting of Hugo Caselles-Dupré, Pierre Fautrel and Gauthier Vernier. One of its works, Portrait of Edmond Belamy, has become the first work of art created by an algorithm, GAN, or generative adversarial network, to be auctioned by Christie’s next October. ‘The portrait in its gilt frame depicts a portly gentleman, possibly French… . The work appears unfinished: the facial features are somewhat indistinct and there are blank areas of canvas. Oddly, the whole composition is displaced slightly to the northwest,’ writes Jonathan Bastable in an article on the Christie’s website. ‘[The auction] will signal the arrival of AI art on the world auction stage.’
Meanwhile, with works such as Portrait of Edmond Belamy, Obvious—a collective of friends, artists and researchers— seeks to ask societal as well as philosophical questions linked to the increasing advent of AI. “Even though we all studied different stuff, we converged for our interest in AI. The first thing that struck us was the creative potential of GAN models, and we began to find parallels between these and the human creative features,” says Vernier, over an email interview. Art seemed a logical way of informing people what can or can’t be done with this technology, and the impact that it can have on society.
With such significant efforts taking place, the potential that AI art holds is immense. According to Anders Petterson— of London-based art market analysis firm, ArtTactic—over the next five years, we are about to witness the increasing influence of AI, and machine learning on all aspects of our lives. Hence, it becomes extremely important that the current and new generation of artists tackle this issue head-on by creating a new model for creative expression.
MEANWHILE, CLOSER home, Agrawal is diligently at work on newer pieces. Contrary to expectation, his home-studio is sans high-tech gadgetry. “I don’t need a lot of physical tools or canvases. Instead, my studio consists of touchscreens and some hardware prototyping tools. I end up using online servers to do my trainings and algorithms,” he says. Agrawal also develops interactive art tools, such as Tandem, a software system, wherein a person’s drawing input is ‘imagined’ upon by a computer, which then suggests an outcome. With a bachelor’s degree in design from the Indian Institute of Technology, Guwahati, and an MS in media, arts and science, MIT Media Lab, Agrawal has always been interested in this subset of art and technology. “I began by making tools for creative human experience and expression. There is one called the Flying Pantograph, in which a work created by a human is translated by a drone carrying a pen. These markings are squiggly because of the friction. A lot of back and forth takes place between the human and the drone in the creative process,” he says.
“One needs to think like artists, programmers and data scientists. All these disciplines come together beautifully in this genre” – Karthik Kalyanaraman, artist
Treading the world of AI art is like connecting the dots, with one work leading to the other. A show like Gradient Descent doesn’t just introduce one to artists like Agrawal and Tokui, but also to artist-researchers like Raghava KK and Karthik Kalyanaraman, who together form the collective 64/1 and have curated the show. “Raghava is an artist who has always been fascinated with technology. And I am a statistician and have also studied art history. Together, we explore art for the post-human future,” says Kalyanaraman, a former professor of econometrics in London.
For both Agrawal and 64/1, the big push towards AI art came in 2015 when Google came up with the Deep Dream Project. Though the ‘computer vision program’ caused a stir, for Kalyanaraman, it wasn’t interesting enough as all the images looked the same. However, it did stoke their interest. “By 2017, pix2pix happened, which was an easily available realisation of the GAN technology. Initially, it resulted in pretty picture making, but not in mature concepts. But this year, we started seeing art that was both aesthetically diverse and conceptually rich, from across the world, which were matured and thus pitched a show to Nature Morte, one in which the final image was created by AI,” says Kalyanaraman.
In an interview with the Deccan Chronicle, both Raghava and Kalyanaraman have mentioned how this genre brings the world of science and art closer—prompting artists to think like scientists, and for scientists to be more creative in the way they look at the world. “One needs to think like artists, computer programmers and data scientists. All these disciplines come together beautifully in this genre,” says Kalyanaraman. The algorithm is like a baby that knows nothing about art. The artist here chooses not to teach the programme anything —be it form, rules or colours. Like a statistician chooses a database to base his model on, the artist needs to curate a set of images— also called training set—which would be shown as examples of art to the programme.
For instance, for his work at Nature Morte, Agrawal curated 60,000 images of human surgical dissection, and then let the machine create its own imagined representation of it. For the Portrait of Edmond Belamy, Obvious too created a data set of 15,000 portraits painted between the 14th and the 20th centuries. ‘The Generator makes a new image based on the set, then the Discriminator tries to spot the difference between a human-made image and one created by the Generator. The aim is to fool the Discriminator into thinking that the new images are real-life portraits. Then we have a result,’ writes Caselles- Dupré on the Christie’s website.
It is believed that portraiture is a tough genre, even for AI. But that was the challenge that Obvious wanted to take on. Also, because people can relate to portraits, having seen them in museums and history books. This allows for their message to reach as many people as possible, and not only be understood by art critics and historians. According to Vernier, the collective wanted the subject to be figurative to everyone in order to evaluate the capabilities of the AI as well.
In such a scenario, how can one, then, reimagine creativity and human purpose in thenew futuristic world? And the unanimous response is that the artist hasn’t gone anywhere. Right from the intent of the artwork to designing something that is aesthetically interesting, the human mind is at the centre of it. “It’s just that the generation of the final image is with the machine. So, creativity gets distributed between man and machine,” says Kalyanaraman. Agrawal concurs, and says: “You can make that call as to when, during the training, should the machine generate the aesthetic, or when to stop its training. Once it creates a sample, you can choose the ones that fit your aesthetic sense or change the dataset. So, through the process, your concept is evolving, with the artist exercising control over it.” His relationship with AI, for one, varies between it being a sophisticated high- dimensional tool and a collaborator.
This also brings one to the question of authorship—does it rest with the machine or the human? And how would one price these works? According to Petterson, even if the actual output is made by a machine, the instructions and architecture is still human. “I don’t see this as any different from when many artists come up with an idea, but the idea is executed by someone else—a studio, a foundry etcetera,” he says. Though, this should not affect the pricing of the works, he expects resistance in the market against AI-made art going mainstream. “But I think this is largely a generational issue, and will most likely fade as the next set of buyers and collectors enter the frame,” he says.
Artists urge people to think of AI along the lines of photography. When this invention came about, it drastically changed the way painters thought. “Painting earlier was trying to achieve realistic precision. But, photography could achieve that much better. So, painters began to think, what can we do that the camera can’t, and you had exploration of surface, planes and depth. I hope that AI too disturbs the complacency of traditional artists,” says Kalayanaraman. Even though societal concerns such as loss of labour are very much real, one at least needs to start thinking that if AI can do art, what else can it do? How do you want it to work for you, and not the other way around? 64/1 is now going to take the premise of Gradient Descent forward with an official programme at the upcoming Kochi Biennale. While the show at Nature Morte focused on the aesthetic, the next chapter in Kochi will highlight the socio-cultural consequences of AI, ranging from its use in surveillance to labour issues.
What’s interesting to note is that while much of this work was earlier spearheaded solely by individuals or tech- related institutions and programmes such as Google Art, The Museum of Tomorrow in Rio de Janeiro, and Ars Electronica in Austria, traditional art spaces are now beginning to take note of this genre. “For instance, The Grand Palais in Paris exhibited Artists and Robots, an immersive show around robots as artists. Victoria & Albert Museum, London, will be opening a show soon on AI. The field is still so new there are no established residencies within the traditional art framework yet. But, things are changing as we speak,” says Kalyanaraman.