A sense of the unfamiliar surrounds the visitors when they step into Gradient Descent. Curated by 64/1, an art and research collective, the exhibition fascinates and unsettles in equal measure as seven artists come together to focus on the interface of art and artificial intelligence. As we find ourselves on the brink of a post-human future, Gradient Descent is an ongoing interrogation into what artistic practices can mean in a world inferred through streams of information and the clockwork of algorithms. Held at Nature Morte, and bound together by a sense of phantasmagoria, the exhibits suggest the deep dreams of machines playing with the ability to emotionally move the human subject.
As the country’s first Artificial Intelligence art exhibition, it reflects global preoccupations; while ZKM Center for Art and Media Karlsruhe, hosting ‘Open Codes’ at the Grand Palais, Paris, opens its interiors to ‘Artists and Robots’. Adding to this conversation on what makes an image artificial, the curatorial team of 64/1 -Raghava KK and Karthik Kalyanaraman- explore the human-machine dynamics to observe if the uncanny world of machine-generated imagery can provide a new turn in aesthetics. The foreboding sense of uncanny fills the dark chamber where Mario Klingemann’s 79530 Self-Portraits plays on a video loop, revealing the seamless mutation of images that unceasingly form fragile or grotesque human faces. Trained on thousands of painted “old masters” portraits from various European collections, the AI (Artificial Intelligence) re-imagines the self-portraits of the artist into startling and haunting configurations. While the artist feeds the AI with a data set, the task of interpreting, arranging, and producing remains with the AI, thus taking the work of conceptualizing outside the human mind and into the machine.
The unpredictability of this process often manifests itself into a playful situation like the dialogue between the AIs in Jake Elwes’ Closed Loop. One talks with images that the other seeks to translate into words, which are then embodied in a crisp, new set of imagery by the first. The two neural networks absorbed in their own perceptions of meanings often produce laughable and poetic miscommunication, not very different from when, as the concept note reminds us, ‘two humans, perhaps from different cultures, meet’. Gracefully pushing the limit of how words can be represented or images can be read, the two AIs sift through the newly learnt human language to create exchanges, at once mysterious and lyrical.
An eerie stream of consciousness surrounds the four-channel video, Deep Meditations by Memo Akten. The first of his two AIs, harvest the vast field of human documentation of life, from landscapes to star clusters as well as abstract notions like art, love, god and faith. The second is fed with soundtracks sourced from YouTube. A continuation to and a fusion of his earlier projects, ‘Learning to see’ and ‘Learning to Listen’, its fluid ephemeral shapes recall the work of memory that floats between images, combining and recreating impressions, receding and emerging, and connecting dissimilar thoughts into an uninterrupted progression. The exercise of watching the video is hypnotic as painterly shapes rise to the surface, leaving a surreal imprint on the eye which is rich in detail and introspective in capacity. The dataset is the inspiration in the art of this information age as Harshit Agrawal’s The Anatomy Lesson of Dr Algorithm locates a conceptual flowering within videos of surgical dissections.
As the AI simulates the labour of a human in diverse fields, the pressing question is if it can claim dexterity over the critical human faculty of creativity. While the authorship is still synergistic, with the human artist offering a carefully curated ‘training set’ on which the AI works with a certain level of autonomy, are there ways in which the machine can generate patterns and possibilities still inconceivable for humans? From the early years of the 1970s when the artist Harold Cohen collaborated with his self-coded software AARON, we have now turned to AIs that appear to create imagery seemingly on their own. Anna Ridler’s exciting Fall of the House of Usher, for example, is achieved by exposing the neural network to the artist’s ink drawings of stills from a 1928 film version of the Edgar Allan Poe short story. The AI then proceeds to make its own reconstruction of the film to compose an evocative inky world with fleeting faces.
Beautiful and yet vexing.