Large Language Model
Media artist Refik Anadol fine-tunes Llama for ‘Large Nature Model’
September 13, 2024

Meta CTO and Head of Reality Labs Andrew “Boz” Bosworth recently spoke with internationally renowned media artist and director Refik Anadol for the latest episode of his podcast, Boz To The Future. Anadol is a pioneer in the aesthetics of data and machine intelligence. His work locates creativity at the intersection of humans and machines and uses the massive datasets around us as his primary material, with the neural network of a computerized mind as a collaborator.

Anadol’s studio trains their own models, leveraging Meta’s open source Llama collection of models in the process. “We’re fine-tuning all the text models and learning a lot,” he says. “Your open source movement is so important—like, so essential. I do not believe we can have a breakthrough in the art because if you think about the complex systems and products and services, they are the breakthroughs, which is amazing. But for artists, we love to break things. We don’t like something real. We allow the fantasies, dreams, you know, glitches. But artists have to take a system that exists and then turn it into the thing that they have a vision for. I have a lot of respect for PyTorch and all that. Open source information is like gold for art making.”

Anadol’s recent project Large Nature Model leverages more than a half billion images of nature collected over four years from partners like the Smithsonian, National Geographic, and others. “Right now, everything is about large language models and about human reasoning, but we thought that perhaps nature is also very important for controlling what happens. What happens if we think, ‘Nature first?’ So we’re fine-tuning Llama now with an amazing nature-focused, rainforest-focused dataset.”

Together, Bosworth and Anadol discuss the immense responsibility that comes from working with these datasets. Anadol not only inspires people with his work, he also brings them along on the process behind the creation because he believes that demystifying the complexities of artificial intelligence will help people better understand and accept it.

“I believe AI is full of possibilities, but possibilities come with responsibilities,” Anadol says. “AI is a mirror for humanity. If we know who we are, I’m very comfortable and confident about AI.”

Listen to the full podcast episode to learn more.


Share:

Our latest updates delivered to your inbox

Subscribe to our newsletter to keep up with Meta AI news, events, research breakthroughs, and more.

Join us in the pursuit of what’s possible with AI.

Related Posts
Computer Vision
Introducing Segment Anything: Working toward the first foundation model for image segmentation
April 5, 2023
FEATURED
Research
MultiRay: Optimizing efficiency for large-scale AI models
November 18, 2022
FEATURED
ML Applications
MuAViC: The first audio-video speech translation benchmark
March 8, 2023