Inside the Dreams of a Machine
“When humans dream, our minds are hard at work, processing memories to form a new combination of images and ideas”
While the virtual musings of a machine may seem out of touch with reality, Refik Anadol envisions a future where machine intelligence touches our minds and souls. A young Turkish-American media artist, Anadol recently unveiled his exhibition Latent History, a visual investigation of Stockholm’s past and present. His work features the visions of machine learning algorithms that have been inspired by photographs from 150 years of Stockholm’s history. Through the dream-like ebbs and flows of this ever-changing landscape, Refik Anadol’s work both remembers and re-imagines the memory of Stockholm.
Anadol’s work uses a class of machine learning systems called generative adversarial networks (GANs). GANs consist of two neural networks that work together to create images, music, or prose that are indistinguishable from our own. One system generates examples, such as a hand-written number; the second network attempts to determine whether examples are real (from the training set) or fake (produced by the first network). The first network attempts to deceive the second. This system can also be understood as a game between a counterfeiter and a cop, or an actor and a critic. These two networks then learn in tandem until an acceptable result is reached.
In the case of Latent History, Anadol’s algorithms have been trained using both archival and contemporary photographs from the City of Stockholm’s archives, Stockholmskällan, and the Swedish National Heritage Board’s archive, K-samsök (SOCH or Swedish Open Cultural Heritage). To better explore the architecture of the city, his team used particular classifiers to remove all elements of people from the photographs, leaving only the environment behind. By discerning patterns in the pictures, the algorithms create a stream of images that combine to reveal new aspects of the city.
Although his exhibit was born from lines of code and cold datasets, Refik Anadol shows us that algorithms can provide personal insights into our emotions and memories. In an interview, he explained that to make art from data, “you have to be on the next level of creativity. You have to find [an] algorithm to make the invisible visible.”
Anadol’s Latent History accomplishes this goal; it captures the unseen consciousness of the city by blending photographic memories with cutting edge technology. The artist believes that this new AI-driven storytelling will become a new form of cinema, one that can “reconstruct common memories for humanity.” Refik Anadol is not only a revolutionary media artist and designer, he is also an inspiring storyteller and spatial thinker who holds a unique vision for the future of machine learning.
Written by Alex Sheen, Edited by James Mueller & Alexander Fleiss