Refik Anadol doesn't create videos—he creates responsive environments. Using machine learning and neural networks, Anadol generates visual landscapes that evolve in response to music, audience presence, and algorithmic decision-making. The result is something that feels both alien and intimately human.
His collaborations with musicians and sound designers prove something radical: AI-generated visuals can convey genuine aesthetic intention. His 'Unsupervised Machine Learning' series, combined with generative soundscapes, creates experiences that transcend the typical artist-plus-technology dichotomy. The machine becomes a collaborator with actual creative voice.
Anadol's immersive installations—in museums, concert halls, public spaces—represent the next evolution of how art and sound interact. His work at LACMA and internationally shows that AI visualization isn't replacing human creativity. It's extending it.
Anadol suggests a future where the boundary between creator and technology dissolves—not into faceless automation, but into genuine co-creation.
What the Data Actually Becomes
The most important thing to understand about Anadol's practice is that the data he uses is not decorative material. He does not feed information into a system and then choose which outputs look good. The data is the subject. When he processes decades of archival photographs of a city, the visual output is not a pretty pattern that happens to be made from photographs. It is an argument about what that archive contains, what patterns emerge when memory is processed at scale, what becomes visible when a machine encounters documents that humans have preserved but cannot hold in consciousness simultaneously.
The Unsupervised series at MoMA in 2022 and 2023 made this argument most explicitly. The installation used hundreds of millions of images from MoMA's own archive as source material. The AI encountered that collection and generated continuous visual output from it. What the audience saw was not the archive; it was the archive's latent structure, the visual logic embedded in a century of modern art becoming visible through machine processing. Whether that constitutes meaning or the simulation of meaning is a question Anadol leaves deliberately open. The work does not answer it. The work poses it in visual form.
The Architecture of Immersion
Anadol's installations are always site-specific in ways that matter to how they function. The scale is not incidental. When he covers the facade of a building with real-time data flows, the building becomes a different kind of object. Its permanence, its mass, its function as civic infrastructure, is placed in relation to the fluidity and impermanence of the data layer. The contrast is the content.
The sound dimension of his practice is equally considered. He works with composers and sound designers to create acoustic environments that correspond to the visual generation rather than merely accompanying it. The relationship between sound and image in his installations is not illustrative. The sounds are not designed to describe what you are seeing. They operate on the same data material through a different channel, creating a total sensory environment where the same information arrives at the viewer's perception through multiple routes simultaneously.
This produces a specific kind of attention. The viewer cannot process the installation by looking at it in the usual way, scanning for what is important, identifying a focal point, understanding the composition. There is no composition in the stable sense. The image is continuous and evolving. The only way to encounter it is to be inside it over time, to let the patterns register, to notice what the system keeps returning to and what it lets go.
The Critical Conversation
Anadol's work has attracted genuine critical skepticism alongside the enthusiasm. The skeptical position holds that spectacular scale can substitute for depth, that the visual impressiveness of large-format AI generation can be mistaken for intellectual substance, that the collaboration between artist and machine flatters both without fully belonging to either. This skepticism is worth taking seriously.
The response to it is not that the work is beyond criticism but that the work positions itself precisely inside the questions the criticism raises. Anadol does not present the AI as a tool he controls. He presents the AI as a collaborator whose outputs he cannot fully predict or explain. The claim is not mastery of the machine but a relationship with it. What that relationship produces, and whether the production has aesthetic or intellectual value beyond spectacle, is the question the work invites the viewer to answer for themselves.
The installations at LACMA, the work for the Istanbul Modern, the collaborations with Spotify and NASA on public-scale visualisations of streaming data and telescope imagery, all of them demonstrate the same core proposition: that the combination of machine learning and human artistic intention can produce visual experiences that neither could achieve alone. The proposition is not self-evidently true. Anadol keeps testing it.
The Question of Authorship
Where does Anadol's authorship reside? Not in the generation itself, which is algorithmic. Not in the data, which belongs to the archives and institutions that preserved it. The authorship is in the framing: the choice of data source, the choice of site, the parameters of the system, the decision about what the encounter between machine and archive is for. These decisions are artistic decisions in the full sense. They determine what the audience encounters and what that encounter means.
Contemporary art has spent decades questioning authorship, delegating it, dispersing it, turning the question itself into subject matter. Anadol's practice extends that inquiry into the domain of machine intelligence without resolving it. The author is present in the work. The author is also, in a specific technical sense, not responsible for most of what the work contains. Living inside that contradiction is where the practice lives, and it is a more interesting place to be than either full mastery of the medium or pure surrender to it.
The current moment in AI-generated art is producing a lot of work that uses technical novelty as content. The fact that a machine made it is offered as sufficient reason for interest. Anadol's work does not rely on that logic. It would be interesting even if the technical method were ordinary, because the questions it poses about memory, scale, and the relationship between data and meaning are worth posing. The technology is a means. The inquiry is the point.
