A Conversation with Refik Anadol, Media Artist

Key Concepts

AI Art: Art created using artificial intelligence, often involving machine learning models trained on vast datasets.
Data Painting: A term coined by Refik Anadol, referring to the concept of information or data becoming a pigment for artistic creation.
Latent Space: A mathematical representation where AI models store learned data, allowing for exploration of new forms and concepts.
Generative Adversarial Networks (GANs): A class of machine learning frameworks where two neural networks (a generator and a discriminator) compete to create realistic data.
Diffusion Models: More recent AI models capable of generating complex and realistic image forms.
Thinking Brushes: A metaphor for AI models that can learn and create, akin to traditional artistic tools.
Data Curation: The process of carefully selecting, cleaning, and organizing data to ensure its quality and ethical use for AI training.
Ethical AI: The development and deployment of AI systems that are fair, transparent, and responsible, with data ethics being a core component.
Generative Reality: A term suggesting a new era of human experience where AI can create realities that blend fantasy, hallucination, and dreams.
AI Museum: A concept for a dedicated space to showcase and explore AI-generated art and its implications.
Slow AI: An approach to AI development that prioritizes sustainability and ethical considerations, even if it means slower processing times.
Human-Machine Collaboration: The synergistic relationship between human creativity and AI capabilities in the artistic process.

Artificial Realities: Refik Anadol's AI Art at Columbia University

This summary details the discussion surrounding the installation of Refik Anadol's AI artwork at Columbia University, focusing on the artist's philosophy, methodologies, and the broader implications of AI in art.

Introduction and Installation Context

The discussion begins with an introduction of artist Refik Anadol, highlighting his prominence in AI-driven art. The installation of his work, titled "Artificial Realities," on two buildings at Columbia University is presented as a significant event. The speaker expresses gratitude to Costis and Andrew for facilitating this opportunity, emphasizing the immediate and impactful nature of the collaboration. The initial idea was to connect Columbia with New York City, a "calling card" for the institution, and to reflect the city's ever-changing nature, a concept echoed by Columbia professor Ken Jackson's theory of New York as a "city-state." The artwork's title was changed to "Artificial Realities" to explore the concept of AI dreaming. A parallel is drawn between Anadol's work and historical figures like Picasso and Einstein, suggesting that art can sometimes precede scientific understanding, particularly on the eve of another industrial revolution.

Refik Anadol's Artistic Philosophy and Journey

Refik Anadol expresses his honor in contributing to Columbia University and inspiring the next generation. He views himself as a teacher, drawing from his family background and his own decade-long teaching experience at UCLA, where many of his team members are students or alumni. Anadol's artistic journey began in Istanbul, a city that connects continents and cultures, mirroring his interest in bridging physical and virtual worlds. His early fascination with computers at age eight, viewing them as a "friend" or "player two," foreshadowed his later work with AI. The movie "Blade Runner" served as an early inspiration, sparking the idea of buildings that could dream and remember.

He coined the term "data painting" in 2008, envisioning information as a pigment. Initially met with skepticism, this concept fueled his entrepreneurial spirit as an artist. Anadol perceives data not as mere numbers but as a form of "memory" that can take any shape, color, or speed, unbound by Newtonian physics. He believes artists have always questioned "what is beyond reality," and AI is a powerful tool in exploring these questions. His move to Los Angeles allowed him to delve into the pioneers of the "light and space movement," blending light and data in his work.

Anadol's studio has grown to nearly 30 people, speaking multiple languages and representing diverse nationalities. A core focus for the past decade has been the creation of "living walls" and "living environments" that integrate art respectfully into architecture. He advocates for art being accessible to "anyone, everyone, any age, any culture," believing it belongs in schools, hospitals, and any space needing inspiration or hope.

The Genesis of AI Art and Key Methodologies

Anadol recounts his early engagement with AI in 2015 at an AI conference, where he was the sole artist. This led to a residency at Google's AI program in late 2015/2016, where he worked with pioneers of current AI models. His initial experiments explored the concept of machines learning, dreaming, and hallucinating, inspired by Jorge Luis Borges's work. He illustrates the transformative power of AI by comparing a task that took nine people six years to sort 1.7 million documents to an AI model completing it in 90 seconds. This marked the beginning of his journey into AI art, creating what he believes was the world's first AI artwork using real-time data in public space.

His work has evolved to explore "seeing through machine," moving beyond just creating beautiful images. He views these creations not as mere paintings or sculptures but as "sculptures of intelligence," akin to the chisel and marble of the Renaissance. Anadol also sees art as a form of medicine, particularly in his work with machine learning algorithms and memory, inspired by the loss of his uncle to Alzheimer's. He highlights a project using brain activity data from children, which sold for 1.7 million euros at a UNICEF auction, demonstrating a commitment to giving back to communities.

The artwork at Columbia University draws from nine years of his early work in New York, a city that has profoundly impacted him since his first visit in 2011. He also touches upon the sensory aspects of AI art, noting the missing element of smell in visual and auditory AI. Four years prior, he began researching and experimenting with scent molecules, learning to mix half a million scent molecules in a laboratory in Geneva, with plans for a surprising scent release in an AI artwork.

The "Cityscapes" and "Artificial Realities" Project at Columbia

The "Cityscapes" piece, later renamed "Artificial Realities," was conceived with the idea of reconnecting Columbia with New York City. The artwork utilizes approximately 120 million pictures of New York, which required extensive curation. Anadol explains that the artwork employs two AI models:

GAN Models (Generative Adversarial Networks): Described as "legacy models" before diffusion models, GANs involve two neural networks: one that understands reality and one that doesn't. The conversation between them generates the art. Anadol's studio developed a new algorithmic architecture for these models, allowing forms to blend in a way that resembles a "camera in the mind of a machine" traveling through a "latent space" (a mathematical space where AI stores data). This allows for exploration of 512 dimensions, impossible for humans to conceive directly.
Diffusion-Based Models: These are more recent algorithms that create more complex and realistic image forms, contributing to the artwork's heightened realism. Anadol emphasizes the use of open-source state-of-the-art AI models, advocating for AI art to be open source.

The term "hallucination" is used to describe the process of the AI generating novel, non-real but realistic imagery, akin to "actual realities." Anadol posits that we are entering an era of "generative reality," where AI can create experiences that blend fantasy, hallucination, and dreams, impacting all our senses.

The Impact and Ambitions of AI in Art

Anadol addresses the concerns of students regarding AI's impact on their future and careers. He views AI as "anything and everything," and therefore, it must be "for anyone and everyone." He embraces the rapid pace of AI innovation, seeing it as essential for human progress. He contrasts the traditional artist's tools (brush, canvas) with the ever-changing digital landscape of AI, where the canvas, algorithms, and data are constantly evolving.

He shares an example of a recent project with Time magazine, where AI analyzed 5,000 covers to identify emerging job titles based on societal problems and innovations, generating a million new job titles. This illustrates AI's potential to solve problems it helps create.

Regarding the limits of AI in creativity and taste, Anadol highlights the challenge for individual artists to access the resources needed to train custom AI models. While tools like Midjourney and Stability AI offer breakthroughs, he believes true innovation lies in developing unique models. He sees machines as "co-creators" and a "creative force" to blend with. The primary challenge remains providing artists with the resources to train AI models on big data.

Anadol acknowledges the concerns of artists regarding data usage, copyright, and ownership. He emphasizes the importance of ethical data sourcing, citing his studio's work with libraries, open-source archives, and institutions like the Smithsonian. He is preparing to open the world's first AI museum in Los Angeles, aiming to foster partnerships and educational techniques. For the Columbia project, he ensured ethical data sourcing, including nature data from Central Park, and utilized sustainable computing resources from Google Cloud, resulting in "slow AI, but good AI."

Ethical Considerations, Sustainability, and Future of Art

The discussion delves into the ethical implications of AI art, particularly concerning data usage and copyright. Anadol stresses the importance of training one's own models with curated data to ensure ownership and avoid unauthorized use. He suggests that while opt-in/opt-out systems exist, the long-term solution for artists is to develop custom models. He advocates for using data inspired by nature and cities, which he believes belong to everyone. He also proposes the integration of AI and blockchain technology to create a system where artists are credited and compensated for their contributions.

Sustainability is a significant concern. Anadol acknowledges the immense energy consumption of AI and suggests that fusion energy might be a necessary future solution. He draws a parallel between nature's inherent AI and the need to learn from it, referencing a tribe that lives in harmony with the forest. He envisions a future with "eco-friendly agents" and "mini agents" for individuals, allowing them to curate their own AI experiences.

The future of art is seen as a convergence of media. Anadol believes that traditional categories of art may reinvent themselves, and AI can help find new languages for human expression. He describes his work as "generative reality," a new medium that can be heard, smelled, and potentially touched, going beyond VR, XR, or AR.

The Artist-Technologist Relationship and Impact

Anadol's relationship with Jensen Huang, CEO of Nvidia, is highlighted as a crucial collaboration. Their interactions, starting in 2016, have been characterized by mutual inspiration and pushing boundaries. Huang's vision for the world and his ability to connect with Anadol as an artist, rather than an employee, have fostered a unique dynamic. Nvidia has provided Anadol's studio with early access to new GPUs and technologies, enabling them to experiment and push the limits of AI art. This collaboration has led to the development of scientifically accurate species models and real-time parameter adjustments on large-scale displays, showcasing the cutting edge of AI and compute.

The installation at Columbia University is presented as a testament to technological advancement and cost reduction. The initial estimate for screens was a million dollars, but the final cost was around $300,000, reflecting improvements in technology. The artwork is designed to be a "living work" that can be continuously reprogrammed. The impact on the campus is described as transformative, changing the building's atmosphere and the daily routines of its 3,000 daily visitors. A significant achievement is that the Columbia installation uses 70% less energy than the MoMA installation, demonstrating a commitment to innovation in both artistic and environmental aspects.

Conclusion and Takeaways

The discussion concludes with expressions of gratitude and a reiteration of the profound impact of Refik Anadol's AI art. The artwork at Columbia University is celebrated not only for its aesthetic qualities but also for its ability to foster dialogue, inspire future generations, and serve as a tangible example of human-machine collaboration. The project embodies a forward-thinking approach to art, technology, and sustainability, leaving a lasting impression on the university and its community.