Refik Anadol has spent the better part of a decade turning museum collections and city-scale data into slow-moving, ambient generative visuals — most famously “Unsupervised,” the piece he made for MoMA in 2022 by training a model on 200 years of the museum’s collection. In June, he stopped putting that work inside other people’s institutions and opened one built specifically to house it.
Watch: Refik Anadol — Art in the Age of Machine Intelligence (TED) (YouTube)
What Dataland actually is
Dataland opened June 20, 2026, inside The Grand LA — Frank Gehry’s roughly billion-dollar mixed-use complex in downtown Los Angeles — with five galleries spanning about 2,300 square meters. Anadol co-founded it with Efsun Erkiliç. The museum’s inaugural exhibition, “Machine Dreams: Rainforest,” runs through January 31, 2027, and is built around an immersive, multisensory piece generated from millions of images and recorded sounds of the natural world.
The part worth paying attention to
The most interesting decision isn’t the visuals — it’s what they’re trained on. Anadol’s studio built what it calls a Large Nature Model, trained on data gathered first-hand from sixteen rainforests, combined with data partnerships with the Smithsonian, the Cornell Lab of Ornithology, Getty, iNaturalist, and London’s Natural History Museum. In other words: no scraped web images, no unlicensed training set. Every source is either collected directly by the studio or licensed through an institutional partner.
That’s a direct response to the criticism that’s dogged AI art generally — that most generative models are trained on work nobody agreed to include. Anadol is using Dataland to argue that permission-based training data is viable at museum scale, not just as a defensive footnote.
Reception has been mixed rather than uniformly glowing — LAist’s review, for instance, describes enjoying the visual experience while remaining unconvinced by the larger premise. That tension is worth sitting with rather than resolving one way or the other: a technically ambitious, ethically framed project can still leave critics unsure whether the art justifies the apparatus around it.
Someone else worth knowing about
If Anadol’s work is about ingesting collections at institutional scale, poet and artist Sasha Stiles works at the opposite end of the same idea: since 2018 she’s been developing “Technelegy,” a personal AI system trained specifically to emulate and extend her own writing voice, most recently shown at MoMA under the title “A Living Poem.” Where Dataland is about what a museum can generate from the natural world’s data, Stiles’s work is about what a single, consenting individual’s own words can become when fed back into a model built just for her.