Refik Anadol’s data-fuelled 3D digital animations seem almost too dazzling for the building exteriors or giant walls of LEDs that have been his canvas, pulling huge audiences at art spaces across the world, over the past decade. A lodestar at the intersection of technology and creativity, Anadol has launched his biggest project yet this month with the long-awaited opening of DATALAND, the first museum for AI-generated art. Housed in The Grand LA (the USD 1 billion Frank Gehry-designed downtown Los Angeles development), its five galleries add up to a 2,300 m² space where the debut exhibition, ‘Machine Dreams: Rainforest’, offers visitors a spectacular multisensory experience. It’s the behind-the-walls custom-built tech that you can’t see though – processing immense databanks, reading visitors’ emotions, and producing art – that’s perhaps the most jaw-dropping. Here, in conversation with Hans Ulrich Obrist, Anadol and DATALAND co-founder and painter Efsun Erkılıç, reveal the scope of their ambitions.

Hans Ulrich Obrist: How did the idea for DATALAND come about?

Refik Anadol: The origins go back several years. When we presented ‘Echoes of the Earth: Living Archive’ at Serpentine in London in 2024, we were developing an early version of the AI systems that power DATALAND today. Around the same time, Frank Gehry, whose building is home to DATALAND in Los Angeles, continued to be a profound source of inspiration. When we projected WDCH Dreams (2018) onto the exterior of the city’s Walt Disney Concert Hall, designed by Gehry, he called it a living skin. That phrase stayed with me. It led us to ask a series of questions: ‘Can a building learn? Can it dream? Can our emotions become part of an artwork?’ For thousands of years, art has existed on walls. We became interested in what might come next. Could we create experiences that move beyond seeing alone and engage our senses, memories, and emotions in entirely new ways?

Efsun Erkılıç: Initially, we searched for an existing space where we could realize some of the technologies and ideas we were developing. But eventually we considered a different question: What would happen if we designed the space we were searching for?

RA: When we first entered the space that would become DATALAND, it was completely empty; Gehry had left us a beautiful open canvas, a concrete shell for free imagination. Rather than thinking about concrete, glass, and steel, we wanted to create an environment shaped primarily through light, data, sound, and human presence. There are few physical objects here. The medium is the experience itself.

HUO: Let’s talk about the ethical use of large datasets.

RA: The moment we learned that the next generation of AI models would be trained primarily on human generated information, we chose a different path. We wanted to focus on nature because we believe it is one of the most urgent subjects of our time. This decision led us to build long-term collaborations with institutions such as the Smithsonian National Museum of Natural History in Washington DC and other scientific and environmental organizations. Together, we assembled a vast archive of ecological data, always with permission, and through formal partnerships. DATALAND is powered by one of the largest nature-focused AI archives ever created, bringing together billions of ecological data points and environmental recordings. There are four petabytes of data held in a supercomputer in the building.

EE: We also wanted DATALAND to be a place where people could develop a deeper understanding around AI and have public conversations. We wanted visitors to engage directly with these systems, understand how they work, and participate in discussions about their future. To have a voice and demand different regulations, they need to be part of the game.

HUO: You created many alliances early on, like the artist residency at Google’s Artists + Machine Intelligence (AMI) program, 10 years ago. Your friends at Google have since been very helpful for many big projects. Nvidia has been a partner. This institution brings together art, science, AI, and cutting-edge technology. Who are your partners here?

RA: Many of these relationships began more than a decade ago. The people I met during Google’s AMI residency remain some of our closest collaborators. Their support has enabled projects that would have been impossible otherwise. Here at DATALAND, all galleries rely on highly complex computational systems developed in partnership with Google, allowing machine-generated worlds to evolve in real time. They also supported our field research across 16 rainforest ecosystems. Nvidia’s co-founder and CEO, Jansen Huang, has been a collector of our work for 10 years and has contributed the computing infrastructure necessary to realize projects at this scale. He’s donated a remarkable number of GPUs (graphic processing units) to create our supercomputer to run 1.2 billion pixels.

HUO: The experience never repeats. It’s not a loop: each visit is different.

RA: Finally, we were able to create a truly immersive room. Within it, a generative bird flies, exploring the world, discovering creatures and connecting with them. Each time, it encounters a unique experience: the scent of molecules, the way the forest receives it, and the kinds of memories that it might carry. There are data tunnels that weave these beautiful worlds together, imagining the memories of living beings.

EE: It can be a little disorienting.

RA: Yes, it’s a powerful immersion. We are experimenting with live algorithms, as well as scent and sound. The machine discovers memories of the forest, its flora, fauna, and fungi. Each time, a unique data tunnel with a different symbiotic relationship is created for the viewer. Resolution-wise, we really pushed the technology to its limits. You can see entire fungal systems. We have created a fungal memory of a machine dreaming those creatures.

HUO: Each sensory level you add means more time [in terms of people’s engagement], so I think you might have trouble getting people out of the room!

RA: Our final gallery is the AI Sanctuary. Whoever enters this room allows the AI to read their emotions in real time and turn it into the final artwork. The room also features a special sound system that transports us into the forest through recordings from 50 million birdsongs. Everything you hear and feel emerges from the large nature model. Sixteen rainforest sound archives come together with a beautiful song gifted by the LA Philharmonic.

EE: We have partnered with Cornell University, which maintains one of the world’s largest bird migration and biodiversity databases, and has shared its data with us. We trained the model using this archive.

RA: The brain of DATALAND is powered by 150 supercomputers. That’s how we run all of the AI systems in real time. It’s a completely custom-built infrastructure because very few organizations require this kind of computational architecture. It is connected to Google Cloud Platform in Oregon and runs on sustainable energy. We hope that each day brings new ways of sharing stories and creating experiences.