That IBM may replace Picasso as the 21st century’s generator of art may seem like a laughable proposition, or a foregone conclusion. The opening of the first exhibition of Artificial Intelligence (AI) art in New Delhi’s Nature Morte gallery suggests that this eventuality is within the realm of possibility — except that there is no universal acceptance of the idea that AI art is art at all.
With seven international artists, in the exhibition Gradient Descent: A New Practice, the curators, brothers Karthik Kalyanaraman and Raghava K.K., seek to make a case for the “deep aesthetic and conceptual practice” that AI art can generate. In the process they open up our minds to the close interface between science and art that seems to be on the horizon of art production.
Together, these works point to the fundamental debate on whether AI provides an aesthetic alternative, or whether it is augmented intelligence, which can enhance and augment rather than take over the processes of art-making.
Art and science
It is pointed out that the histories of art and scientific invention have always worked in tandem — at least in the arts of the Western world. Using the converging lens and viewing screen of the camera obscura, Vermeer in the 17th century famously gained the “optical way” of constructing his paintings. While the Indian artist of that time showed little curiosity in introducing perspective, the European painter was experimenting with an apparatus to create modifications in the construction of a painting.
This is not very different from the artists’ use of media such as 3D bioprinting today, to enhance the depth of their work. AI, however, represents a phase in human history when technology is more dominant than ever before — an equal collaborator with the human brain. Its actual role in the domain of art, then, becomes an open question.
To the artist who serves as intermediary between AI and the viewer, one may ask, why use AI at all? In a telling demonstration of what AI can achieve, IBM has put out on Google a mash-up of different images created from blending a reworking of famous paintings — ‘Starry Night’ by Van Gogh, ‘The Scream’ by Munch, ‘Christ’ by Jamini Roy. By integrating these images, the computer can come up with any number of works as permutations and combinations that carry traces of the original masterpiece.
The conventional use of AI in art goes through three possible intermediary stages. The first is when AI copies or impersonates the artist’s own work and comes up with alternatives or extensions to that vision. The second is image-to-image translation in which single frames can be treated to change the time of day, the sense of distance etc. within the original, thus significantly altering the image.
In the third stage AI plays the role of collaborator. Algorithms that are made to create images, visualise sounds or music are deconstructed to allow the artist to rework them, and then allow for AI’s interpretive play in fully realising the work.
Clearly some innovative uses of machine art are possible — and the market is not impervious; companies and collectives dealing in and designing AI visual art have sprung up. Thus ‘Le Comte de Belamy’, the portrait of an imaginary 18th century French count made by the French group Obvious (which works with AI), complete with a heavy gilded baroque frame, created a buzz for the price that it fetched. In place of the name of the artist is an algorithm that approximates the image.
Computers in dialogue
The valid question that arises is where is the artist’s creativity in what finally emerges as a mechanical product. In conversation, the curators suggest that the creativity lies in the process, that osmotic space for exchange where man and machine produce an unpredictable third entity. It is in the choice of content or what are known as ‘training sets’.
Kalyanaraman suggests that this creates a “poetics of metonymy” — that the computer is “not a mixer or a matcher but creates its own aesthetics of perception and reception.” In the present exhibition, Mario Klingemann, the most-well known among the participants, created a database of thousands of faces from pre-Giotto European painting, which were then read in conjunction with 64 biometric points of his own face.
The result is a series of faces which are also self portraits that appear to morph and distort in a continuous loop. Perhaps an even more interesting exercise was undertaken by London-based artist Jake Elwes, a graduate from the Slade School and School of the Art Institute Chicago, who engages two AI computers in dialogue. Speaking to each other as image and text, that appear like a diptych on the wall, they simulate a very real conversation of mutual exchange, producing poetic responses to simple visuals and text, like ‘snow on the ground’ or the ‘passage of a black bird’. The value of his work lies in how the machine simulates the effect of conte on paper, and the delicacy of the trace, even as it creates a complex language of abstraction.
In putting together this exhibition, the curators sought Indian artists working in AI and found only one, Harshit Agrawal. In a way this is symptomatic of the stop-start-stop approach to AI in India, the limited funding it has attracted, and the complete absence of public engagement.
Although AI is considered the fourth industrial revolution, with its use of deep intelligence and mass data, there is a fear that the AI revolution may skip India altogether while China seeks to dominate the global AI industry. In India, the everyday use of AI products — such as Apple’s Siri or Amazon’s Alexa — are too few to be significant. Until such time as AI percolates into public consciousness, exhibitions of art generated by AI may be few and far between, and only confined to conventional art spaces.
Gayatri Sinha is an art critic and curator who, while preoccupied with her art website www.criticalcollective.in, is also contemplating a book