Editor’s note: As Nature Morte Gallery in Delhi plays host to India’s first exhibition of artworks created by Artificial Intelligence, Gradient/Descent, it raises a number of questions, both philosophical and practical. One of the co-curators for the show, Karthik Kalyanaraman, spoke to Firstpost about the implications of the exhibition, the disruptive powers of AI, the position its art must hold in the world, and what questions such work will ask of human conscience.
When we talk of AI, we talk only in terms of application (safety, accuracy etc) or disruption (how sex-bots could change relationships etc). What are the various considerations we must make when AI is taken as a driver of aesthetics, art and culture as a whole? Can you cite examples where this is already happening, noticed or otherwise?
When AI is considered in terms of its impact on aesthetics, one of the notions we have to reconsider is the notion of creativity. We think of creativity as an exclusively human phenomenon. But AI art forces us to question that belief.
Devising new chess moves, which for hundreds of years was considered a very human and very creative act, is no longer considered truly creative anymore, well, because of AI! What is the future of human creativity if most traditional areas of labour are taken over by Artificial Intelligence? Well, this is a hugely vital question I think because you can argue that it is precisely her creativity and imagination that enabled the human to out-compete other species, and even ecological limitations, and to be what she is now. So this gets at the heart of our self-definition: what will it be to be human in the AI future and how can we make it worth our while?
As for examples of how AI is already driving culture, it plays a significant role in online social network behaviour. In so far as the “news” and the information that we see on social media is curated by AI based on its inference of our preferences. Therefore, we have to accept AI as a driver of our online behaviour and culture.
What was the genesis of this exhibition, and the immediate challenges of executing something like this? How does one deal with the complexity of the idea itself?
Raghava (co-curator) and Aparajita Jain (co-owner of Nature Morte gallery) had been talking since around July 2017 of creating a trading platform for digital art and also for algorithmic art, because they felt the lack of such a platform was holding back young artists who were migrating towards those media. Raghava brought me in (we had already been working together, as a passion project for both, on some basic ideas for why we thought the notion of transcendence deserved to be resurrected in art theory) to think about ideas for how to create a pioneering show to pitch to AJ when she was in Bangalore to discuss the platform in January 2018.
We had intense brainstorming sessions about what would be exciting and pioneering. Piece by piece, an entirely new framework to think about art history and the future of art emerged, on that blackboard in Bangalore. Based on the thesis, when AJ came down, we proposed two shows, again on that same blackboard: one a broader show about emerging trends in art and the other a more radical one about art actually made by AI, which we felt was an important genre that the art world should face. AJ, who felt that the other options we gave her had already had pioneering exhibitions done about them, immediately chose the more radical option and said let’s do it. She also decided to commission the thesis and got us a research assistant for it. And we were on! The next couple of months were spent interviewing artists and making the thesis a solidly argued piece of innovative art theory to back our larger endeavour.
The immediate challenges to executing it were, 1) we had to understand not just the aesthetic but also the technical side of the work in depth to assess the originality of the artists and the depth of their conceptual innovation, 2) it’s a field in its infancy so we had and have communication challenges. People either fear AI is going to take over all of art or AI art is not art at all. We believe neither of these positions is nuanced enough. We also have challenges in explaining the process of the artists to people who might know about art in the 20th century but know nothing about how AI works and vice versa. But on the whole, these have been good challenges.
A keyword for me in this exhibition is the word ‘trained’. AI, like humans, works with a level of intelligence, I presume, acquired through experience and learning. Considering the context here, how do we make the differentiation that our intelligence isn’t ‘artificial’? Nao Tokui’s piece, for example, speaks about that ‘training’. Does art then become the final frontier that AI must conquer?
The first question is a deeply philosophical one. With the AI art based on neural networks, all the artists are doing is showing general purpose neural network. The artist, in general, has no sense of what the final work will be like, the final image is produced by the algorithm alone. Do we call that final work original imagination? Some wouldn’t because they’d say this is computation too. Well, I would. Because it depends on how one makes sense of the similar human ability. Because, given we haven’t found that magic creative neuron, really that’s got to be computation only (or at least we don’t know what else it is). If we eschew a kind of mysticism that sees human creativity as some sort of transcendental thing, and think a little about that superbly creative task, that we all master, learning our first language, well, how do we do it? By being exposed to lots and lots of examples! A very very large training set.
Similarly, artists produce art based on the other artwork they have seen. If they didn’t, they simply would not be relevant. Looking at it from a purely pragmatic perspective, if a machine can make humanly surprising, stylistically new kinds of art, I think it is foolish to say well it’s not really creative because it doesn’t have consciousness. And I think we are at that point where we are indeed starting to see really fascinating examples of what AI art can produce
I don’t know if art is indeed the final frontier that AI much conquer. Certainly, I don’t think art, given what a complex beast it is, and how human-directed it is, will ever be “conquered” by AI, but I certainly think it will be profoundly modified. Much like how photography profoundly changed painting.
What has art by AI accomplished so far, and what remains its greatest challenges still?
Not quite sure what the first question is. AI art, aside from its value of creating a kind of art with a deeply different practice involved in its making, hasn’t quite ‘broken’ any formats. It still is partly formal partly conceptual art presenting fairly traditional media like digital video, physical prints and so on.
But AI still has a lot of challenges in art. So far the AI we have seen only excels at a kind of formal visual pattern making. But formal relations within an image is not all. There is also the conceptual dimension. Here another kind of AI, the so-called symbolic AI, that understands semantics and concepts and how logic and language work, will certainly have a role. And once AI starts to produce rich concepts (rich for humans) there will always be the fact that the artwork is an object intended for the human and thus the ethics of the human relation will always have a role.
What are the key similarities you see in the way AI and the human mind draws, imagine or remember? What are the key differences that may still set the two apart?
Let’s take memory. Human memory is partial, imperfect and directed by the affective significance of the content. Machines, on the other hand, have perfect memories. As for imagination, surely a child that has only seen apples will not be able to imagine oranges. Similarly, a machine with a specific training set will not be able to imagine things completely outside that image universe. Those were similarities.
The differences have to do with the fact that the machine draws, in a sense, the whole picture at once, it doesn’t really have a sense of the part-whole relationship. Because it creates the entire image pixel by pixel and the colour and intensity values for each pixel are correlated with those of every other. The human on the other hand typically (and I say typically because it’s really not well understood how humans draw) draws whole as an explicit coordination between the parts. The other difference is that the human has a sense of materiality, of paint, of stone etc, in general, of the object that an artwork is. Whereas a computer, given it works primarily only with information, has a sense of the artwork as an image.
The first ever portrait by an algorithm will soon be auctioned. Is there a way to classify this as progress, like human evolution, even though it feels like replicating the self? What future does AI hold in the field of culture and aesthetics?
First off, the portrait by Obvious is not the first portrait by an algorithm (Mario Klingemann was making portraits with AI before the Collective). Nor is it the first AI work to be sold. Private collectors have been buying the work of AI artists for the past year. Also, I don’t think I would call it progress. The narrative of progress is a difficult one for the arts, because it is not clear that value in the arts is uni-dimensional and ordered (the two qualities necessary for the progress of any kind).
All that said, what perhaps I would call it, is an occasion of profound redefinition of the human. We have always thought artistic creativity as an exclusive domain for the human; with this moment I think we are acknowledging that we might need to redefine what creativity can be and more profound what the human truly is.
For arts and aesthetics first, let’s say this: the traditional arts are not going to die out because of AI art. But, 1) the AI art movement is barely two years old, so it is in its infancy in terms of what it can achieve. We can expect it to become a full-fledged genre (like lens-based art) of its own with great conceptual and aesthetic richness; 2) photography if we remember had an indirect but enormous impact on painting: suddenly the aim to paint realistically from nature and the importance of linear composition seemed not so compelling; photography could do those better, or at least more obviously, than painting!
Painting as a field changed. This was the birth of Impressionism, of abstraction, of everything we know as contemporary art. We can expect something similar with AI. As AI starts getting very good at producing compelling aesthetic images that are also conceptually rich, humans have to start asking what art can humans make? What can we focus on? What will distinguish the mark of the human in the art of the future? Well, one way forward at least is provided by the AI artists: work with the machines. Embrace being a cyborg. We have other theories on how art can respond, which involve focussing on the under-theorised ethical dimension of art, but that’s for another time.
Gradient/Descent is on show at Nature Morte Gallery, Delhi till 15 September
All images courtesy of Nature Morte, New Delhi