How can art and AI interact? We discussed that with Anna Ridler, a London-based artist and researcher. Since 2017, she has been working with GANs, Generative Adversarial Networks, creating thought-provoking works, including Fall of the House of Usher I-II (2017), Mosaic Virus (2018- 2019), Bloemenveiling (2019) and Circadian Bloom (2021).
I would like to start with the most challenging question: what is art for you?
This is a challenging question! I think art for me is a way of exploring ideas of how the world works and articulating it in a visual or experiential way.
Why did you choose to work with AI and machine learning?
I’ve been using machine learning as part of my creative practise for about five years - not just a tool but as a process. I came to it as I’ve always used data and large datasets as part of the way that I work, and this particular technology seemed like a natural extension of this interest and gave me so many different ways of working. Because it is a complex system and there are lots of different parts to it, there are also many ways of exploring different ideas in each part – from the algorithm to the data to the final installation. By working in this way, it also allows me to understand and show error states and assumptions and reveal the labour of making whilst inverting the usual process of such things. Many conversations on AI and creativity centre around machines making art or replacing humans – which to me ignores what is fundamentally interesting about the technology: humans. It is often underestimated the large amount of humanness and human decision making that is essential to making AI work.
Could you tell us what GANs are and how do you work with them?
GANs are a type of machine learning – I work with and am particularly interested in those that produce visual imagery. It is a complex, iterative process with many interdependencies where one network learns to mimic imagery that is given to it (the training or dataset) to the extent that the “the counterfeits are indistinguishable from the genuine articles” over the course of many cycles of learning, or epochs. The images are not just parts of photographs that have been fed to it, stitched together but pictures that have been entirely generated or imagined of what the GAN thinks it should be produced given the data that it has received. These images are, to me, sometimes incredibly beautiful – having meandering, dreamlike quality to the results - results that are recognisable as being real but at the same time have small tells that show that they are not real. It is these images - either in still form, or more commonly as part of moving image work, that I tend to work with in my installations.
What fascinates you the most about the images generated by GANs? Can we define these images as “creative products” or does creativity only lie in the process of generating them?
The images that come out are loser and wilder than I could have made by myself. Entropy and decay are incredibly beautiful. Making it reminded me of a short story by Tlön, Uqbar, Orbis Tertius, a short story by Borges, is an account in part of a fictional country of Uqbar where “the duplication of lost objects is not infrequent” but these duplications - hronir - “were the accidental products of distraction and forgetfulness” (my missing chair, lazy way of drawing eyes) “no less real, but closer to expectations” but which over times and cycles become comes “a purity of line not found in the original”. It is my work but also not my work - recognisably me (especially the first one) but nothing I would have been able to do by myself. Watching it is a very odd sensation like catching a glimpse of yourself in a mirror before you realise it is you.
There are always questions about creativity when it comes to art and where the creativity lies in this piece - is it me creating the training set, is it the GAN producing the image? What is “real” art? a bricolage artist combing existing things or someone with memories and ideas trying to convey them to an audience through a new medium? In many ways, the conversation that is had around machine learning is not too dissimilar to conversations that were being had around conceptual art in the 1960s. For me, because art is so closely tied to the concept, to the idea, to the process I will always be the artist who is in control of the making.
Could you describe a work and/or a project you are particularly fond of?
Circadian Bloom is a relatively new ongoing project that I’m really excited about – it is the start of an exploration into ideas around other, non-human ways of keeping time. It uses digitally created imagery plants that have a particular type of chronobiological clock - one that will consistently open and close its flowers at fixed times of the day - so that the piece essentially works like a kind of clock. These plants behave this way regardless of external stimuli - for example, a night-blooming cactus will only bloom at night, even if it is exposed to darkness during the daytime and light at night; a morning glory moved into permanent darkness will still flower in the mornings. The clock is designed to start at dawn and end at dusk, and changes daily to reflect the precise longitude and latitude it is programmed for. Throughout the day the imagery of the different flowers evolves in real time in synchrony with their natural counterparts, blooming and closing at the correct time of day. Because the length of daylight changes throughout the year, looking at it at the same time will result in different flowers being shown (7 am will be in darkness in winter months, but will be daylight in summer). The project is inspired by Carl Linnaeus’ idea of a floral clock or horologium florae that he proposed in his Philosophia Botanica in 1751 after observing this phenomenon of certain flowers opening and closing at set times of the day and harks back to an earlier, medieval way of constructing time in temporal hours when the hours of available daylight divided into twelve so that an hour was dependent on when and where a person was.
Is there any potential of GANs that you have not fully explored yet?
There is always something new! So many academic papers come out each week with new advances in technology it is almost impossible to keep up with the latest advancements. I’m also extremely interested in experimenting with other generative machine learning models to produce works in the future, to experiment and play with some of these ideas before incorporating them into a project. For me, it’s always important to make sure that the concept mirrors the technology that I am using and that it is not just being for the sake of it but is somehow echoing what the piece is about.
What are the advantages and the risks of using GANs for an artist?
Like any medium, the advantages and disadvantages depend on what the piece is trying to do. It can be quite hard to ‘control’ a GAN which might make it tricky to do particular projects. There is also the issue of the data that goes into training the GAN. The datasets that are needed to make training sets are often extremely large - thousands and thousands of images and inputs. I think there are interesting issues that come from using someone else’s dataset when it is that large - it is virtually impossible to look through everything, so the amount of control as an artist that you have - what has been included or excluded, what biases and prejudices you are replicating and repeating - is very difficult to control. Unless you are absolutely certain it has a diverse and comprehensive vocabulary, the cultural, political or social biases of whatever groups of people that defined the training set will be exposed. I make my own dataset, which offers opportunities to do push back against this, but it is very expensive and time-consuming to do so. There are also issues to do with computation – many of the latest papers require large numbers of GPUs which is both expensive and extremely energy-intensive. However, the advantages are so many! I think it allows you to produce imagery that would nearly be impossible to do otherwise.
Is Artificial Intelligence significantly changing artistic practices?
For some artists, I suppose mainly digital, it absolutely is changing how they work in a very fundamental way. There are new tools and new ways of making – it is now possible to do things that were would have been in science fiction a few years ago. It is giving new tools and ways of exploring the world.