When NVIDIA’s CEO Jensen Huang is ready to roll out a new GPU (one of its specialized chips that enables complex computing for AI and data processing), one of his first calls is not to a scientist or an engineer, but to an artist. Refik Anadol, the Istanbul-born media artist known for turning raw data into immersive digital environments, often gets early access to NVIDIA’s experimental chips before they’re released to the public.

Anadol recently described one of these early-adoption moments to an audience gathered in Columbia Business School’s Henry R. Kravis Hall. The artist had taken the stage at a recent NVIDIA GPU Technology Conference conference, before 15,000 people, and was manipulating live AI parameters on a 16K screen powered by unreleased hardware. “Can you imagine?” he asked the attendees of CBS’s latest Distinguished Speaker Series event, co-hosted by the CBS Arts Society.  “I’m touching a GPU that’s not even been produced yet!” 

Anadol recognizes his relationship with Huang and NVIDIA to be an incredible opportunity for someone like himself, an artist entranced by big questions about the future of AI and the new portals of possibility it may open for humanity. In Anadol’s estimation, the interplay between AI’s development and humanity’s—much like the relationship between his own artistic vision and that of Huang’s technological one—has the potential to be greatly symbiotic. 

That optimism animates Anadol’s new permanent installation on CBS’s Manhattanville campus. Made possible through the generosity of Andrew Gundlach ‘01, recipient of the Horton Award at the Tamer Institute’s 2025 Awards Breakfast, Artificial Realities: New York City (2025), is now glowing in the lobby of Kravis Hall.The piece is built from 120 million publicly available images of the city, which is transformed into a continuously morphing, dreamlike vision of New York. 

“It has changed the ambience of this campus,” said CBS Dean Costis Maglaras during the conversation. “You can feel it from across the courtyard.”

This is in part because of the deeper conversations the new piece is engendering about AI across the campus, which Maglaras welcomes. He has frequently highlighted AI’s role in shaping the rapidly evolving workplace, and accordingly, CBS has introduced new AI-focused courses and research to enable its students to keep pace. 

Anadol explained that he, too, sees Artificial Realities: New York City (2025), which is part of his ongoing Machine Hallucinations series, as more than mere decoration: it’s a living and evolving experiment. It represents a collaboration between human imagination and machine learning, one that suggests the future of AI might look less like cold automation and more like a shared act of dreaming.

From Istanbul’s Merging of East and West 

Anadol grew up in Istanbul, a city that literally bridges continents. 

“Where East and West connect,” he said, “you feel cultures and histories overlapping.” He attributed his early understanding of connection—between physical and virtual, past and future—as eventually coming to define his art.

At eight years old, he received his first computer, and quickly came to think of it “as a friend, as my ‘player two.’ I felt so connected with that machine as a child,” he said. 

Even before he began working with AI, Anadol was already questioning the limits of what could be called art. In 2008, as an undergrad, he started asking professors and other students, “Can information become a pigment?” Some of his professors and peers pushed back on this possibility, or responded with absolutes like ‘computers can never make art.’

But for Anadol, data was never just numbers. “When I see data,” he said, “I see a form of memory, and it’s a form of memory that can take any shape, any color, any speed.” Over time, that idea evolved into a philosophy: data as pigment, memory as material.

During his 2015–2016 residency with Google’s Artists + Machine Intelligence program, Anadol worked alongside engineers and scientists developing the models that would later power today’s generative systems. He coined the term “AI Data Painting,” which he used to animate one of his guiding questions: “If a machine can learn—can it dream?”

This question is at the core of his ongoing Machine Hallucinations series, which now spans major installations at MoMA, The Sphere in Las Vegas, Paris’s Centre Pompidou-Metz, and CBS’s Kravis Hall, with Artificial Realities. Each project invites viewers to peer inside the “mind” of a machine as it processes enormous datasets—visualizing, remembering, and imagining in its own nonhuman way.

“I call this piece ‘Artificial Realities’ because I believe what’s happening with AI right now is really interesting,” Anadol said. “We’re entering a new era of humanity’s journey, where we see things, we read things, we smell things, and we will taste things, we will touch things. It’s somewhere between fantasy, hallucination, and dream.”

During the pandemic, Anadol began training AI models not only on images and sound but on scent molecules, pulling on data from 75 million flowers. It’s one example of how his studio, now a 30-person team in Los Angeles representing 15 languages and 10 countries, continues to push sensory boundaries, creating immersive experiences that feel both futuristic and human.

Cameras in the Minds of Machines

Anadol’s process begins not with code but with curation. The Kravis Hall installation, for instance, scraped its millions of source photos from public archives and sorted them through two different AI models. The first, a generative adversarial network , learns patterns by pitting two neural networks against each other—one that “knows what’s real” and one that doesn’t. The second, a diffusion model, produces increasingly sophisticated, lifelike textures. Together, they create what Anadol describes as “a camera in the mind of a machine.”

He holds his creative process to strict ethical standards. For one thing, Anadol challenges his technology partners to use computing powered by sustainable energy, even if it means sacrificing speed. “It’s slow AI, but good AI,” he said. 

He also insists that his studio use only open-source or publicly licensed datasets—often working with public institutions like the Smithsonian.

“I’m an artist; I will never touch another artist’s work,” he said. “That’s why we train our models with the data that I know. I know exactly which folders are in there; I know exactly which cluster we are using.” 

He encourages other artists to train their own models rather than rely on commercial systems like Midjourney, because “when the tool itself is the breakthrough, the artist is limited,” he explained. “But as soon as one of us wants to create a custom AI model, immediately that person is ahead of the curve, because that person has curated information.” 

A question naturally follows from this, he acknowledges: will all artists continue to have access to the resources, data, and compute power necessary to push AI art forward? This broad and democratized access to AI’s power is a key feature of the future Anadol envisions.

“I think AI is becoming anything and everything—which means it has to be for anyone and everyone,” he said.  

This open-source-tech attitude extends to his philosophy of the role of art itself. He rejects the idea that art belongs only in galleries or museums. “Art should be for anyone, everyone, any age, any culture,” he said. “The school, the hospital, any environment where we need inspiration or hope—that’s where I think art belongs.”