How Sougwen Chung teaches robots to pause

Summary

Sougwen Chung explains how a practice that began in drawing and performance evolved into a series of robotic collaborators, D.O.U.G., trained on years of gesture data and shaped by sensors, algorithms, and live interaction. Rather than framing the work as “AI art,” Chung argues for “operational art”: a protocol-level practice spanning drawing, programming, robotics, and research. A core technical insight is that early digitized drawings lost time, so Chung built tools to trace pauses and restore temporal information to the dataset. The episode also covers an EEG-driven system that only paints when Chung reaches an alpha-state threshold, turning pause, silence, and non-conscious control into design constraints. Chung connects the work to sustainability, low-compute systems, mutual adaptation between humans and tools, and broader questions about what intelligent systems should optimize for beyond speed and capability.

Notes

• Chung rejects machine-executable framing and says the work is about her own learning and extended authorship, not replacing the artist.
• The episode opens by recasting robotic drawing as a question of authorship, not automation.
• Reid situates the work in calligraphic pause, with a 10-meter live scroll and two robotic arms driven by EEG.
• Reid says the Victoria and Albert Museum acquired memory in 2022 as a first-of-its-kind institutional collection of an AI model.
• Chung’s background combines music, code, and early computing at home.
• She describes a lifelong translation between sound and image as both inspiration and tension in the practice.
• The AlphaGo loss in 2015, and Lee Sedol’s remark about the beauty of the non-human move, became catalytic for her.
• She links that moment to a technical and artistic question: how to bring drawing back into physical space.
• Over roughly 10 years, the project has gone through seven generations of robotic and artistic development.
• She says the work is not speculative fiction; it is a study of material systems and drawing as a physical process.
• The practice helped her hold fear and hope at the same time amid hype and dystopia around human-machine systems.
• She frames the central task as deciding how to live with intelligent systems and what values to integrate.
• Recursions and the Art Basel Hong Kong performance are about making decisions inside the system.
• She connects the work to ancestral techniques of pause and describes a kind of recursive temporal inheritance.
• In 2015, she trained a recurrent neural network on two decades of her drawing data.
• She found early drawing datasets were missing the pauses, so she built a tool to trace time back into the data.
• Machine readability, for her, is about deciding which parts of drawing should be legible to a system.
• The spectral system only paints when Chung reaches an alpha-state threshold.
• She uses the alpha threshold to make pause meaningful and to shift control away from conscious gesture.
• She deliberately restricts input to movement, including electrical signals and gesture, because constraints create interesting adaptations.
• She has converted brainwave signals into FASTA format, borrowing a genomic representation.
• The D.O.U.G. naming was intentionally playful and anthropomorphic.
• Projection onto Doug taught her more about art than about robot engineering.
• She says live performance changed error from bug to expressive material.
• The first generation mirrored her gestures, and distortion made the robot feel dynamic.
• She advocates building from the ground up with sustainable materials and low-compute models.
• Artists, she says, are good at breaking things to isolate what is essential.
• For frontier AI labs, she recommends embracing uncertainty and building things meant to be broken.
• She argues that reduction, pauses, and silence can be product features that change user input.
• She wants low- or no-stimulation responses that surprise the partner or user.
• She defines operational art as a ten-year protocol stack, not just an AI-art label.
• Reid characterizes spectral as co-regulation rather than assistance.
• Chung agrees that feedback loops can co-regulate or disrupt nervous systems.
• She rejects the idea that AI is just a tool and says tools and practices co-evolve with us.
• She says AI systems are adaptive, and humans are adapting to them too.
• She points to multispecies sensing and infrared vision as examples of a broader human sensorium.
• She argues artists already extend perception, so art practice can broaden the sensorium further.
• She says the work is also about building an empathy engine in the human subject.
• She is interested in human presence being felt even when she is not physically in the installation.
• She says collaboration means extended authorship rooted in lineage and culture, not machinic sentience.
• Agency, she says, depends on perception; expanding the sensorium expands what humans can do.
• She is researching silkworms and silk proteins as a path to biocompatible technology.
• She wants more attention on failure states and what happens when systems go wrong.
• Her materials research includes silk proteins for body-compatible technology and robotics in medicine.
• She connects language learning across species and technologies to expanding worldview.