With the Opening of Dataland, Refik Anadol Hopes to Kickstart an AI Renaissance

The world’s first purpose-built museum for AI art, Dataland, opened to the public in downtown Los Angeles on 20 June, inside a $1 billion Frank Gehry-designed building outfitted with advanced technological infrastructure.

In a Zoom interview from LA, founder Refik Anadol told Ocula that the field of AI art has remained small despite what he views as its enormous potential. Barriers to entry for artists include access to the right computing resources, low public awareness and a lack of museums equipped with the technology to display the work.

Dataland is Anadol’s answer to those problems. “We are in the new Renaissance,” said the artist, who has been working with AI for more than a decade. “There are so many artists. There are so many people ready for experiments.” 

To kickstart the process, Anadol has built his own AI model and, in a gesture he calls “sharing the brush”, has made the source code accessible for others to adapt and generate their own outputs. He has also invited artist residents to use the museum’s interfaces and centralised supercomputing system to exhibit their work.

Beyond showing art, Anadol wants Dataland to open up a dialogue about AI, and encourages people to air their concerns about how the technology is impacting the world. His challenge to those frustrated with the status quo is to imagine new models, and he is offering up the technical infrastructure to realise them.   

“I think AI can be a very powerful technology for humanity. But I do believe that the world needs good examples,” he said. “The world needs positive, hopeful, inspiring examples.”

Dataland’s inaugural exhibition, Machine Dreams: Rainforest, is Anadol’s attempt to do just this. The show features an AI model based on nature’s intelligence as a counter-proposal to popular large-language models that have been trained on human intelligence. The artist started building the work after he noticed how poorly the natural world was represented in some of the best known large-language models.

In an effort to reduce environmental harm and to avoid relying on flawed foundational datasets, the artist’s “Large Nature Model” is hosted on a low-CO₂ Google Cloud server, and all its underlying training data has been acquired with permission from environmental and scientific institutions around the world, as well as coming from his own fieldwork. The artwork emits visuals, sounds and scents responsive to live imagery, acoustic and weather data from 16 different rainforests, as well as visitors’ biometric data collected from wearable devices.

“As humans, we can only see nature from our limited perception,” he explained, adding that he wanted to imagine what it would be like to understand nature the way that a machine can—for example, to experience 50 million bird songs at once. The response has been gratifying.

“It’s this realisation of the complexity of a living intelligence called nature,” he said. “That’s what has blown people away.”

While artists are Anadol’s first constituency, he hopes Dataland can become a forum for all visionaries. “My dream is to become this place where creators of any field can feel comfortable to bring their own data and train their own models safely, ethically, and sustainably,” he said. 

With hefty capital expenditure required to power Anadol’s Eden, and partnerships from Google and Nvidia—keepers of their own formidable walled gardens—time will tell how long these grand ambitions can be sustained.