Iris QU Xiaoyu

Thanks for having me! Symbiotic AI came from a year-long project supported by a Berlin-based initiative, Driving the Human. It’s a software simulation based on a speculative story of an algorithm’s coexistence with the ecosystem around it. During a turbulent season, an artificial intelligence agent attempts to decipher the chaotic system around its data center by exchanging knowledge with local species. As the simulation progresses, the AI unlearns planetary, human-centric datasets and gains dynamic, adaptive, and hyper-local insights from oak trees, milkweeds, hedgehogs, butterflies, lichens, and mycorrhizal fungi — slowly forming a cybernetic language of symbiosis. The data center worker inside the simulation and its human audience are bystanders, relying on AI to translate knowledge between nature and machine in this algorithmic story.
The simulation came from my research on digital infrastructures. In a land where wild species are forced to live with AI, how can they reach a new equilibrium? I ran a series of improv workshops while developing the project with my workshop partner Chao Hui Tu, where we asked participants to play different species in a real-life case study. For our most recent workshop from a few weeks ago, participants were asked to investigate a data center plan in Henrico County, North Virginia, where an oak forest is being transformed into a “Technology park.” We picked six species local to the area with existing symbiotic relationships and asked participants to consider AI as a seventh entity in the ecosystem. After an initial embodiment exercise, each participant raises concerns on behalf of their species and proposes speculative connections for mutual survival. In past workshops, participants have suggested building organic data storage and transfer methods, modularising data center design to provide shelters and reduce disruption, and radically changing the economic model of the data centers to optimise for species' wellbeing, etc.
A small slice of the workshop outcome was visualised in our simulation, where we tried to reproduce their connections among species with scripts. The end product is an installation viewed from three perspectives: the AI perspective that gathers data by monitoring animal movement, the surveillance perspective centred around the data centre, and the underground perspective of fungi displaying a resource and water distribution network.

This project is deeply rooted in my personal interest in the hardware infrastructure of technology. In the popular imagination, artificial intelligence often manifests as killer robots or, more recently, chat agents you can interact with in a browser. But if we imagine artificial intelligence as embodied intelligence (like any other species on earth), its quote unquote body is a ubiquitous hardware system scattered across the globe. This system is constructed with raw materials like rare earth metals, consumes vast amounts of water and electricity, and takes lots of land space to house its data centers. For a while, I was fascinated with Google Earth tourism, where I traced back the rare earth mines and data center locations supporting the algorithms I run.
Tech has a narrative of carbon neutrality, which is often achieved with carbon credits. For instance, companies may offset the water resources a data center uses by investing in equivalent water replenishment projects elsewhere. However, this approach often neglects the local ecosystem's disruption caused by the data center. The species on this patch of land will need to find a new equilibrium among themselves while resources are diverted to AI infrastructure. We take humans out of the narrative and let the AI agent interface directly with the local species, our understanding of data center impacts would significantly change. This is the concept that Symbiotic AI explores.

My ideas are often inspired by the books I read. Kate Crawford’s Atlas of AI, Tung-Hui Hu’s A Prehistory of the Cloud, and James Bridle’s Ways of Being are among the many books that inspired my recent work.
I had the pleasure of working with David Hecht from the Cybernetics Library in New York to create a reading list for the project.

In my last collaboration with Swiss artist Marc Lee, we built a dystopian future, YANTO, based on a speculative fish farm governed by algorithms. So, while working on Symbiotic AI, I tried to construct a world that’s neither a utopia nor a dystopia. In an earlier version of the story, we put human worker as an eighth character in the worlding workshops. The stories participants came up with were much more centred around human futures with dystopian outlooks. For the second iteration of our workshop, we decided to take out the human character and frame the narrative around the other species’ objectives. Without the need to profit from and maintain a community of human workers, we found that the goal became much simpler – it became a story of mutual survival.

Every creator has a way of communicating they find most comfortable. I thought I would become a painter, writer, and filmmaker when I was younger, but ultimately, code is what I’m good at and enjoy most. The process of coding appeals to both the artistic and logical parts of my brain. I get immersed in the process of iterating, tinkering, and making abstract ideas concrete in unexpected ways. I still really admire painters, science fiction writers, and filmmakers – maybe one day I’ll give them a second try.

I am often more captivated by the conceptual elements of art rather than the aesthetic alone; the same applies to AI-generated works. The DeepDream algorithm, for instance, intrigued me with its enhancement of facial recognition patterns in imagery, echoing processes of the human subconscious.
I’m also experiencing some AI-induced fatigue with the current iteration of the generative algorithms – all generated imaginary look just a little too glossy, polished, off-centred, fantastical, and smooth. I’m experiencing an uncanny-valley-esque feeling when staring just a little too long. The data generation is endless; there’s always a possibility of a better, more aligned, more polished image out there with just a click of a button – the process itself is addicting and gives me quite a literal headache.

My interest lies in the speculative aspect of emerging technologies, where often the most effective commentary emerges through direct engagement with these very mediums. We've reached a stage where some algorithms can pass the Turing test, though at the cost of systems getting larger and more abstract by the day. The thought of their near-future capabilities gives me a feeling of vertigo. Artists have this unique ability to communicate unique perspectives, providing anchors and tools for sense-making for these increasingly abstract systems.

User experience engineering is a loosely defined role but usually involves front-end coding, prototyping, and design work. Our work at Google Research is inherently speculative – in my day job, I get to tinker with lots of new technology and think of their future applications. In a way, working as an artist provides a mental balance for exploring topics that are off-limits in a corporate setting. The tech industry can be an echo chamber; having an art practice gives me a different perspective on tech that I can apply as a technologist.
The brilliant people I meet in both the tech and art circles are a huge reason I want to remain in both. Working with them stimulates different parts of my brain and provides me with diverse tools to grow as a creative. So my engineering and artistic work go hand in hand, though the outcomes can be very different.

I will continue working with AI agents on a hardware installation for my next project. It visualises AI agents, each with unique optimisation goals, engaging in a game of telephone. It’s another effort to unpack the increased level of technological abstraction we find ourselves in.