Machine Dreams: Rainforest

Refik Anadol’s Machine Dreams: Rainforest, the headlining debut exhibition at Anadol’s Los Angeles museum Dataland, is a highly ambitious attempt at using AI and immersive projection techniques in the service of sustainability and climate communication. The exhibit aims to prove to skeptics that machine learning can offer something radically different from generic ‘slop’ by dedicating the 25,000-square-foot space to endangered ecosystems.

Instead of using a neural network on human art or text, the art studio trained a custom “Large Nature Model (LNM)”, an AI model built entirely on environmental data, exclusively on natural data, including billions of ecological data points from coral reef topography, climate records to extensive birdsong audio archives from the Cornell Lab of Ornithology. While the LNM is an effective framing device – given the current omnipresence of LLMs in public discourse – it should also be viewed cautiously because, while the term suggests a fundamental orientation towards natural principles and structures, the technology itself does not appear to be significantly different from traditional GenAI systems, which partially fulfills the criteria for greenwashing despite the fundamentally positive intent. The exhibition takes the ‘raw’, otherwise often invisible metrics of vulnerable ecosystems, specifically data collected across 16 global rainforest environments, and translates them into a tangible, reactive sensory space. It encourages the audience to perceive the natural world as a complex, hyper-connected entity that is actively shifting but also degrading in response to human activity.

The most glaring irony of AI art is the massive amount of energy and water required to run data centers. Dataland treats this as a central design constraint as LNM runs on a Google Cloud server located in a low-carbon zone in Oregon powered by, according to official sources, 87% renewable energy. A visitor’s entire multi-sensory journey allegedly consumes about the same amount of electricity as charging a single cell phone.

The exhibit also aims to frame datafication as a practice to facilitate a different for of human-nature relationship. Visitors wear specialized biometric sensors that track real-time physiological signals like heart rate and skin temperature. The generative artistic projections across the walls shift and react to these inputs, attempting to forge an intuitively felt feedback loop between human biology and the simulated ecosystem.

Based on cutting-edge spatial audio systems (L-Acoustics Ambiance and L-ISA), the galleries change their physical room acoustics in real time based on what is happening on screen. Soundscapes blend the rainforest data with recordings of ‘sacred’ healing songs of the Amazonian Yawanawá community, thus aiming to create a familiar but still ‘authentic’ immersive experience.

One of the strongest points raised by critics is the uncomfortable ethics of ‘digital extraction’. The studio traveled to the Amazon to collect “troves of material”¹ and recordings from the indigenous Yawanawá community (“5 petabytes worth of raw data that we collected by ourselves”, according to Anadol, allegedly “sourc[ed] […] with the consent and participation of researchers”) to create a high-tech gallery inside a billion-dollar mixed-use luxury complex in downtown Los Angeles, The Grand LA designed by Frank Gehry. Arguably, translating an existential environmental crisis into a polished, commercial and branded attraction (e.g. a prominent cooperationCooperation This is a dummy entry about cooperation. with L’Oréal Luxe) in itself evokes greenwashing.

While the technology is undoubtedly impressive, the visual language is clearly reminiscent of the signature swirling data clouds known from Anadol’s earlier works. Because the visuals are so abstract, they are awe-inspiring but not suitable as a concrete, urgent call to environmental action. The sensory overload risks numbing the audience rather than sparking real-world empathy.

The rhetoric used to promote the exhibit further solidifies this point. Framing the venue as the “world’s first museum of AI arts” feels like a deliberate attempt to monopolize the conversation by relying on superior financial and technological resources.