In Conversation with the NFF Fellows: Simone C Niquille

Every year, three artists work on their own innovative research at the intersection of technology and society in the NFF Fellowships Programme. In this special programme, everything is up for discussion: the questions being asked, the methods used, even the outcomes are not fixed. To give insight into their work forms, dilemmas, and motivations, Paulien Dresscher, curator of the programme and researcher in the field of digital culture, engages in conversation with the Fellows.

The second Fellow this year is Simone C Niquille.

Fellows 2024-2025
Part 2: Simone C Niquille

Chair Don’t Care

Simone C Niquille is a researcher and designer who joined the Fellowships Programme with her research project Chair Don’t Care. With Chair Don’t Care, Niquille investigates the boundaries of storytelling with and about machine learning. While the character Chair wanders through a seemingly simple environment, it turns out that navigation is no easy task. In fact, Chair Don’t Care is a study of computer vision, and how the computer’s gaze sees and influences our world.

The philosophical approach to the non-human gaze is not new, and originates from the world of film: in the early years of film theory, film was seen as the art form that shows or constructs reality, with the “eye” of the camera as the registering element. With the famous film Man with a Movie Camera (Vertov 1929, Soviet Union), the camera is explored as a Kino-Eye: the machine gaze. The underlying idea was that the camera sees more, but also differently, than the human eye, and thus could also present a different version of reality. Computer vision goes a step further than the Kino-Eye: it is an algorithmic eye that can not only record images, but also read them, classify them, and then respond to them: it doesn’t just represent, it can also have an impact.

Computer vision is the field within artificial intelligence (AI) where the boundary between the world and how it is perceived blurs: it enables computers to automatically “see,” analyse, and interpret digital images and videos so they can recognize objects, people, scenes, or patterns and act upon them. Computer vision not only captures the world, but also influences how we understand it and move within it. And today, it is ubiquitous: in healthcare to detect tumours, in vacuum cleaners that navigate your home, but also in VFX for lip syncing and deepfake techniques. And of course, in animation, which is Niquille’s area of work.

Is it a bug or a feature?

A computer camera doesn’t know what it’s looking at; they are the eyes of a robot. We have to teach it, Niquille says: “It’s like an automatic vacuum cleaner: before it can function, it has to learn to recognize objects and make decisions about them: what do I need to go around, what do I need to do?”

This learning is a crucial part where ethics and responsibility come into play. Just as filmmaking revolves around the representation of people, bodies, and identities, computer vision can also involve bias; a subtle distortion where certain people or groups are favoured or disadvantaged. The question then becomes: who is responsible if the “machine gaze” discriminates or distorts?

For Niquille, computer vision is also about the queer experience: “How do you, as a body, move through a world that looks at you through stereotypes and standards? If I cut my hair shorter, I’m addressed as ‘sir’; if my hair is longer, as ‘madam.’ Stereotypes determine how you’re seen by society.”

This idea of stereotypes also plays a role in computer vision, where the question quickly arises: Is it a bug (an unintended error) or a feature (designed that way on purpose)? It can be a bug if it’s based on a poor dataset, or it can be a feature, where the system deliberately excludes groups. When something goes wrong with machine learning, it’s an important question whether it is a technical problem or social in origin.

Asking questions

With Chair Don’t Care, Niquille builds on her earlier work, The Beauty and the Beep, an animation project in which she taught a chair to walk. Through its fumbling and stumbling, the protagonist – the chair – elicited increasing empathy and made her think further about the fate of a chair: “It’s such a simple object, always part of our lives, that we hardly think about it and often take it for granted, which is also reflected in the expression ‘to be part of the furniture.’”

The chair in Chair Don’t Care is in its teenage years, frustrated because nobody wants to see it for what it wants to be. As a result, it lashes out in an aggressive, adolescent way.

For Niquille, it is clear that much more is possible with machine learning and animation than initially appears. “In traditional 3D animation, you first create a virtual skeleton – also called a rig or armature – a term that also refers back to the chair. With this, you can animate a character’s movements. This process is time-consuming and complex, but you can give the character a lot of personality: how someone walks or moves can, for instance, show what kind of person they are.” Creating such animations is a true craft, and there are artists who dedicate their entire careers to it. But, as in many creative fields, AI and machine learning are being used to simplify such processes: within animation, online platforms use algorithms to calculate how someone moves. That is useful for games, for example, to quickly populate worlds with NPCs (non-playable characters). In Unity, you can give a character a target: the algorithm then makes it move toward that point. Ideally, the system learns to do this better and better.

For her research, Niquille further developed this technique as an independent storytelling device. Machine learning itself becomes both the subject and the script, including the moments when it becomes absurd or goes wrong: “This way, not only the end result is influenced by machine learning, but the process itself is used as a narrative element.”

One of the challenges of this way of working is adapting the character to the virtual world. Niquille explains that in our real world there are all sorts of ergonomic assumptions: “You know how high a step is, or how a door handle works. This can be seen as a form of ‘ableism’, most systems and spaces are designed for ‘average’ bodies. In the virtual world, you can play with this: high steps, short or long legs, big or small steps.”

This way, Niquille was able to create absurd struggles for the character that reveal how the body relates to the world. For most people, the real world works fine, but in the virtual world you can experiment and show, in a playful way, what happens when these assumptions no longer hold.

Space for serendipity

For Niquille, the Fellowship was an important moment, because after her earlier work The Beauty and the Beep it gave her the luxury to look back and look ahead. It gave space to explore how to continue. The chair came along, as a moment of reflection and the start of something new, where time was an important factor: often you get support for a specific project, but having the trust to spend time on exploration without immediate measurable results is essential. Without that space, nothing can happen.

“For me, this moment really marked a mid-career reflection: it gave me the chance to calibrate, to think about how I want to reach people and with whom I want to collaborate.”

Part of Niquille’s research was about the Chair Don’t Care project itself, but also about how to proceed afterwards. It offered the space to ask questions and think about the direction to take, without an immediate answer or result being demanded. There was room for serendipity, which is essential to the process.

What, then, is the end product?

“It feels strange not to have a finished end product for such a long time. The question ‘what have you done all year?’ inevitably arises. But for me it was precisely a period in which I learned and developed a lot.”

The outcomes of this process are interesting. In filmmaking, many decisions are often made during production, because there’s no time to prepare everything fully: character profiles, world-building, preparing soundtracks. This project gave her the chance to do that preparatory work and lay a solid foundation before actually starting the animation.

For Chair Don’t Care, Niquille set up Unity worlds within the programme in which the character can navigate and experiment—including a small interface to trigger earlier versions of the “brain”—allowing her to switch back and forth between different versions while animating. Instead of a character simply getting better linearly, it can now be used as a dramatic-narrative element. By working closely with a musician, she also developed a shared language, created sketches, and researched how music can support the story.

The Town Musicians of Bremen

Finally, Niquille worked on character development: the chair got friends, forming a small ensemble, comparable to the Town Musicians of Bremen.

The bottom character is a rhinoceros, above it the chair, and next to it a squirrel. The choice of these animals relates to the history of machine learning and the origins of archetypes.

The rhinoceros is based on the first scientific drawing by Albrecht Dürer in 1515. Dürer had never seen a rhinoceros but made this sketch based on descriptions. Thus emerged the image of the “archetypal rhinoceros.” Niquille: “For me, this early version of the rhinoceros is an example of what we might now call AI-slop: an image built through visual language. It looks believable but has little to do with the reality of a rhinoceros.”

The second animal is the Janus-squirrel: a 3D mesh (a digital skeleton or framework that forms the basis of a 3D model) generated with one of the early text-to-3D models such as DreamFusion. The squirrel has multiple heads, a consequence of how the model handles perspective. The idea that an animal or human has exactly one head is not a concept that an AI model inherently understands.

And finally, on top, the chair: it is the Bertil chair from IKEA, which appeared in the 2006 catalogue as a synthetic image: a computer-generated rendering instead of a photo of a real chair. Nobody noticed; it didn’t stand out and wasn’t experienced as special, unlike the spectacular AI images we see everywhere today, sometimes even dismissed as “AI-slop.”

This success – hiding in plain sight – led IKEA to use synthetic product images more and more often. Companies like IKEA are thus unknowingly producing enormous amounts of potential training data by replacing traditional product photography with CGI images: computer programmes are no longer trained on images of the real world, but on the output of other programmes.

The three archetypal figures – the rhinoceros, Bertil, and the Janus-squirrel – together form an “Image Stack”: a way of investigating how images are generated and viewed, and how the computer helps us, but sometimes also determines, how we look at things. With her work, Niquille also asks how you involve people in such a story: how do you make something technical, nerdy, and experimental still accessible and entertaining?

Accessibility

Niquille’s goal is to present technology in an engaging way, without losing its complexity. What makes her work special is the confusing combination of deep philosophical questions about machine learning and AI, delivered in the form of childlike animated characters. By translating such complex concepts into short films and interactive stories, she invites both children and adults to look critically and with wonder at the ways technology shapes our world.

Our adolescent Chair no longer wants to be taken for granted and challenges its environment—an echo of Niquille’s own drive, urging us not to take computer vision’s gaze for granted: pause to consider what things are, and question what seems so self-evident.

Written by Paulien Dresscher