Machine Hallucination

Practice Area

Client

Artechouse

Industry

  1. A scene from the exhibition that showcases the GAN generated images projected onto the space

  2. These are computer generated images (GAN outputs) of New York, showcasing hallucinated images of New York

  3. A scene from Machine Hallucination which transforms each of the photos into a fluid data sculpture

  4. These are computer generated images (GAN outputs) of New York, showcasing hallucinated images of New York

  5. A scene from Machine Hallucination which transforms each of the photos into a fluid data sculpture

  6. A scene from the piece showcase a visualization of dimensionality reduction inside the mind of a machine

  7. A scene from the piece that showcases a contact page of all of the GAN outputs

  8. A scene from the exhibition that showcases the GAN generated images projected onto the space

Project Vision

Artechouse commissioned Refik Anadol Studio in the fall of 2019 to showcase an inaugural exhibition at their new state-of-the-art digital space in Chelsea. The studio debuted their largest installation to date: Machine Hallucination, an exploration of New York’s architecture through the mind of a machine. In deploying machine learning algorithms on photographic memories of New York City, Machine Hallucinations presents a data universe of New York in 1025 latent dimensions. This installation aims to uncover the ever-changing shape of the city, evoking our collective memory, and the ways that new forms of visual representations can alter people’s perception of this iconic destination.

The exploration for this piece begins with 113 million publicly available photographs of New York – both archival and contemporary – the largest raw dataset ever gathered for an art exhibition. From this enormous swath of data, a subset of 9.5 million of the most definitive images of New York were chosen – those without people, devoid of ego and distraction – resulting in a strict focus on the city’s iconic architecture. The question of why we collect, record, and share our experiences is entangled with the formal and aesthetic concerns about how to represent reality, totality, and the depth of human imagination. Machine Hallucination is a groundbreaking latent cinematic experience derived from representations of urban memories as they are re-imagined by machine intelligence. This installation is novel therefore in its generation of new synesthetic storytelling techniques derived from multilayered manipulation of vast visual archives beyond the conventional limits of the camera and the existing cinematographic techniques.

The exhibition included Machine Hallucination’s 45-minute immersive cinematic experience, a process video that explained some of the ‘behind the scenes’ techniques that allowed for the project to be a possibility, three data paintings from the Machine Hallucination series, and accompanied by a soundtrack designed by Kerim Karaoglu.

Initially, this exhibition was meant to run from September 6 to the beginning of December but because of the success of the exhibit, it was extended until the end of January. The project served to draw thousands of people to this new space, not only celebrating the cultural impact of the city but also highlighting cutting-edge innovation and new possibilities for collective storytelling.

Project Details
Design Team

Carrie He (motion design), Efsun Erkilic (executive producer), Refik Anadol (media artist), Nicholas Boss (data scientist and information designer), Pelin Kivrak (scholar), Julia Pryde Thompson (creative research), Tobias Heinemann, Danny Lee, Ho Man Leung (generative designer), Raman Mustafa (lead architectural design), Christina Moushoul (architect), HyeJi Yang (architect), Alex Morosov (AI/software design), Kerim Karaoglu, (sound designer)

Project Area

6,000 sq ft

Consultants

Kyle McLean (VVVV developer), Maurizio Braggiotti, Efe Mert Kaya, Ali Emre Kali (visual designer)