Pindar Van Arman

2020 / Generative AI System, Paint on Canvas

Can deconstructing your artistic process and teaching robots to paint help you better understand yourself? Pindar Van Arman thinks so, and has put the theory to work for 15 years in more than 2,000 portraits.

Developer Portraits Gallery

gtc20-fall-ai-art-gallery-pindar-reveal-training-portraits-500x500

Training Paintings

Pindar’s robot trained for the final painting by practicing on individual portraits submitted by NVIDIA GTC AI Art Gallery visitors and registered developers during the days leading up to its unveiling.

Final Portrait

Artonomous used feedback loops, CNNs, and GANs to generate the final painting.

The Experience

gtc20-ai-art-gallery-overview-pindar-van-arman-experience-cover-1cCW-D

NVIDIA recently reached the 2 million registered developers milestone. Pindar’s robot helped commemorate this milestone by painting a portrait during the week of GTC—creating the face of the NVIDIA developer community.

WATCH THE ROBOT

The Process

gtc20-ai-art-gallery-overview-pindar-van-arman-process-1a-3c33-D

The NVIDIA Jetson-based autonomous robot begins each portrait with a set of reference photographs. Pindar’s generative AI system then goes to work mimicking the artist’s creative process.

gtc20-fall-ai-art-gallery-pindar-process-1b_410x230

Artonomous uses more than two dozen artificially creative algorithms, including multiple generative adversarial networks (GANs), convolutional neural networks (CNNs), and Feedback Loops to paint the portrait one brush stroke at a time. The robot’s creativity is modeled on Marvin Minsky’s The Society of Mind.

gtc20-fall-ai-art-gallery-pindar-process-1c_410x230

Dozens of independent creative agents are put into competition with one another, each continuously evaluating the painting and trying to influence the next mark being made. At the completion of each work, Pindar critiques the painting, adjusts the algorithms, and begins the cycle all over again.

Pindar Van Arman

Pindar Van Arman’s artwork is an exploration of the artistic process, and how close artificial intelligence comes to human creativity. Over 15 years, his painting robots evolved from making backgrounds to using AI algorithms. Then he was inspired by Minsky’s The Society of Mind, which theorized that the human mind was a collection of smaller interactive intelligences, each of which surfaced when needed. As his robot’s algorithms started competing for control of the brush, they became creative. While Pindar remained the artist, his robots had become studio assistants capable of creative autonomy, even producing artworks completely on their own.

This is artonomous in collaboration with Kitty Simpson, an internationally published photographer with a passion for close-up, black-and-white portraiture that captures the core of each model by tuning out “the noise” and revealing the beauty.

www.cloudpainter.com
www.artonomous
Twitter
The AI Podcast

Sessions at GTC