New Technologies, New Visions
On view April 12–14 at Printed Matter’s LA Art Book Fair 2019
In 2019, Printed Matter’s LA Art Book Fair returned to the Geffen Contemporary building at MOCA in downtown Los Angeles. Together with Anteism Books, AMI organized an exhibition focusing on works made from 2015 to 2018 to trace the emergent practices of artists creating with machine intelligence.
About the Exhibition
Developments in technologies have long influenced image-making. From photography to aerial and satellite imagery to digital printing techniques, the possibilities for visualization afforded by new technologies have always spawned new forms of expression.
In recent years, advances in machine learning are similarly revolutionizing how artists make pictures. Algorithms inspired by the structure of the human brain, known as deep neural networks, enable features like image search and language translation. When run in reverse, these networks can generate novel images and text.
Focusing on works made from 2015 to 2018, New Technologies, New Visions traces the emergent practices of artists creating with machine intelligence. Drawn primarily from works produced in residency at, or in collaboration with, Google’s Artists + Machine Intelligence program, the show begins with early experiments in understanding and giving form to the algorithmic gaze. It showcases artists who use generative techniques to create portraits and landscapes that come from our world, but are, nonetheless, otherworldly. The show includes a text essay by Hito Steyerl that considers the implications of early AI technologies, setting the scene for new technologies we may want to build.
New Technologies, New Visions is organized by Eva Kozanecka, Kenric Allado-McDowell and Anteism Books.
Works Shown List
Refik Anadol
Archive Dreaming, 2016 — Architectural installation with 4-channel video, 8-channel audio, custom software
Refik Anadol calls himself a data artist, meaning that data is his medium. To visualize more than 1,700,000 digitized documents in the SALT Research Collections, Anadol collaborated with artist and engineer Mike Tyka to employ a technique known as t-distributed Stochastic Neighbor Embedding (t-SNE). Digital files were organized by attributes, and then projected onto a six-dimensional space, and represented in three dimensions. Developed in 2008 by Laurens van der Maaten and Geoffrey Hinton, the t-SNE algorithm is unique in its ability to capture the local structure of the high-dimensional data very well, while also revealing global structure such as the presence of clusters at several scales. The resulting set of data points in space is called a point cloud, or “data universe.” When presented in its native digital form, it allows one to explore new connections between objects in the archive. The final work (not shown) is an immersive architectural space that is a means, Anadol explains, to “reframe memory, history and culture” through the lens of machine intelligence.
Mike Tyka
Portraits of Imaginary People, 2017 — Chromogenic color-prints
In recent years, algorithms have become proficient at producing and manipulating images, video, and sound to a degree that makes it difficult to distinguish generated media from real media. In this series, Tyka creates a novel type of photographic portraiture, namely, photos of people who don’t exist. To produce this work, Tyka sourced thousands of photographs of faces from Flickr to train a machine-learning program called a generative adversarial network (GAN). The images shown here are a result of a simulated “arms race,” as Tyka describes it, between two neural networks: the first network (the “Generator”) tries to generate increasingly convincing output, while a second network (the “Discriminator”) tries to learn to distinguish real photos from the artificially generated ones. At first, both networks are poor at their respective tasks, but with time, the images generated become increasingly realistic. The result is an algorithm that can create new faces based purely on emergent rules about what constitutes a face.
Tom White
Perception Engines, 2018 — Two-color risograph prints
To “see the world through the eyes of a machine” is how Tom White describes his intention for this series. To make Perception Engines, White trained a neural network using ImageNet, a database of images organized by Stanford Vision Lab and used by the industry to train and test machine vision algorithms. The outputs are what a computer believes to be a pair of binoculars, an iron, etc. The artist’s agency lies in selecting a number of parameters for the network, like the colors and thickness of lines, and refining the output by rejecting prints that he doesn’t find aesthetically pleasing. “I am trying to free the algorithm so it can express itself, so people can relate to what it’s saying,” explains White.
Ross Goodwin
WordCar, 2017 — A Cadillac XTS mounted with an Axis M3007 surveillance camera; GPS unit; custom sofware, and a laptop
Ross Goodwin describes himself as “not a poet,” but “a writer of a writer.” In 2017, he trained a character-level text prediction neural network model, known as a long short-term memory recurrent neural network (LSTM RNN), on an undisclosed corpus of text, and set off on a road-trip to generate machine poetry. He outfitted a Cadillac car with a surveillance camera, a GPS unit, a microphone and a clock, and connected it to a portable AI-writing-machine that fed from these input data in real time. As he drove, a manuscript emerged line by line from the machine’s printer on long scrolls of receipt paper that filled the car’s rear seats over the course of their journey. The resulting text was published as a book of poetry in 2017, titled “1 the Road.”
Casey Reas
Untitled Film Stills, Series 2, 2018 — Chromogenic color-prints
For Casey Reas, software is more than a tool: it is an extension of his internal abilities to think and explore. Inspired by the photographs of Cindy Sherman and the films of Stan Brakhage, Reas uses generative adversarial networks to make pictures. To make a single picture, Reas inputs a list of 100 numbers between -1 and 1 into a trained model. Each new value generates a new picture. When Reas moves across this latent space of all possible pictures in an orderly manner, he can generate a series of related stills. In this way, each image in Reas’ Untitled Film Stills is a frame from an imagined film. The result is “ambiguous, dreamlike images that oscillate between clarity and the ephemeral.”
Anna Ridler
Untitled (Tulips), 2018
Named after a virus that causes a stripe pattern variation in tulip petals, “Mosaic Virus” is a film by Ridler that shows tulips blooming, whereby the appearance of each tulip is influenced by the price of bitcoin. To create the film, Anna Ridler took ten thousand photographs of tulips over the course of tulip season, and categorized each by hand to form the training data set for a generative adversarial network. The output are novel flower paintings, or “botanical impossibilities,” as described by Ridler. For Printed Matter’s LA Art Book Fair, Ridler reimagines the film as a print work, displaying 52 imagined tulips as playing cards. “I wanted [the print work] to echo ideas around speculation, risk and chance,” says Ridler.
About LA Art Book Fair — Printed Matter, Inc., is the world’s leading non-profit organization dedicated to the dissemination, understanding and appreciation of artists’ books. The LA Art Book Fair, initiated in 2013, is Printed Matter’s companion fair to the NY Art Book Fair. Held at the Geffen Contemporary building at MOCA in downtown Los Angeles, the LA fair features over 300 exhibitors from the west coast and some 20 countries, and is attended by more than 35,000 visitors annually. The Geffen Contemporary at MOCA is located at 152 North Central Avenue, Los Angeles, CA 90012.
