Demystifying AI Art: Refik Anadol reveals the process behind generative creations
Anadol's latest exhibition at London's Serpentine Galleries, "Echoes of the Earth: Living Archive," takes an unprecedented approach towards transparency. It clearly outlines the raw data inputs and internal workings of Anadol's generative AI model, referred to as the Large Nature Model. Through radical disclosure, visitors gain new insights into how the artist's AI system is constructed and derive inspiration from vast troves of environmental imagery.
Over the past 15 years, Refik Anadol has gained prominence as one of the leading digital artists in the world. He has continuously worked with and helped advance cutting-edge technologies throughout this period. Anadol began creating algorithmic art in 2008 and large-scale projection mapping installations in 2010. He was also an early pioneer of virtual reality, working with Oculus when they first released developer kits in 2013. Artificial intelligence has been a focus for a decade, with notable residencies like Google AMI in 2016 where he developed AI data painting and sculptures. More recently, Anadol has explored blockchain and NFT technologies since 2020. Through his experimentation and collaborations, Refik Anadol has established himself at the forefront of major technological developments in digital art over the past 15 years.
With his exhibition "Echoes of the Earth: Living Archive" at the Serpentine Galleries in London, Refik Anadol (who describes himself as a media artist rather than a digital artist) is at the forefront of two important developments in emerging media art. Through openly showcasing his artistic process, Anadol seeks to demystify sophisticated technologies like AI, blockchain, and NFTs. He also aims to make information more accessible to the public through the use of artificial intelligence. Both of these trends aim to reduce fear and uncertainty around rapidly changing technologies. They further demonstrate to traditional art communities the substantive nature of digital art forms.
In May 2023, Anadol and his team used their AI model's output to projection map a live "hallucinatory" work directly onto the façade of Casa Batlló in Barcelona, simultaneously presenting the piece at Rockefeller Plaza in New York City.
The corresponding section "On NFTs on Living Architecture" meticulously outlines each step of the creative process. This includes collecting an initial dataset from Gaudi's sketches, archives of visual/academic files, and publicly available house images. The data was then processed to detect objects, classify images, and sort content into themes. An AI model was generated and trained by processing this wealth of source material. Finally, a "pigment pipeline" was created to transfer the visualized archive into the undulating, fluid-inspired motions that have typified Anadol's work over the past decade. Each stage leading up to the finished projections is documented with unprecedented transparency.
During the 2020-21 COVID-19 lockdowns, Anadol reached out to some of the largest public archives of natural history data for assistance. As part of his overarching Dataland project—described as a "museum and Web3 platform dedicated to data visualization and AI arts"—the artist lists the institutions that responded. These included the sprawling Smithsonian Institution in Washington D.C., home to 148 million objects, 9 million public specimen records, and 6.3 million public images. The Natural History Museum in London, with its 80 million specimens and 4 million public photographs, also contributed. Additionally, the Cornell Lab of Ornithology based in New York shared materials from their collection of 54 million images, 2 million sound recordings, and 255,000 videos. By directly engaging these authoritative archives during the pandemic, Anadol gained access to an immense wealth of research and documentation for his work.
