David Young
CHAPTER II — PATTERN
Artist 1 of 1
←Back to CHAPTER II — PATTERN
Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here. Intro by Kate goes here.
Kate Vass: Across Tabula Rasa, Manipulations, and your GAN trained on Conway's Game of Life, your practice seems to move from generating patterns to questioning perception, and finally to systems that learn from those patterns. Looking back, do you see these as separate bodies of work, or as different stages of a single investigation?
David Young: I see them as part of a larger investigation — or maybe several running at once. It wasn't the plan when I began. But the connections in the work reveal a question I keep returning to: perception. What does it mean for a machine to perceive, and how might that be different from us?
Tabula Rasa started in contrast to corporate AI: instead of the enormous datasets these systems normally demand, I wanted to see how a system would behave with very little — and what that would might expose about how AI works. Looking at the results, if I let myself anthropomorphize, I'd say the machine appeared to struggle. The images revealed a tension between the machine-nature of the code and the strangely organic behavior of its neural network. Over time, for me, the question of what was being perceived began to shift — from what is the machine perceiving to what do we imagine the machine is feeling.
Manipulations took the same question further, by a different route. Because we don't actually know what these systems are doing, so I began manipulating the images to search for what might be hidden within them. As I worked, features surfaced that we don't register with our own eyes. The question beneath it: is the machine seeing and perceiving something other than what we perceive?
Training a GAN on Conway's Game of Life set modern AI against a much older idea about emergence. Life isn't artificial intelligence; it's a demonstration that complex behavior can arise from almost nothing — which is the same intuition today's AI is built upon. I was curious what it would make of an ancestor of its own founding idea. What it saw wasn't what we see. I find that clarifying — and reminder that the confidence these systems are our future is more haphazard than it sounds.
KV: Manipulations reveals structures that often appear more visible to the machine than to the human eye. Do you think artificial intelligence is discovering genuinely new realities, or simply making us aware of patterns that have always existed beyond our perception?
DY: When I started working on Manipulations, I began with a thesis — that AI sees differently than we do. I thought that because its internal processes are so unlike ours, its perception must be, too. I thought it was seeing things somehow beyond our perception — and that this was what revealed its otherness.
Having lived with the work for a while, I've evolved my thinking.
As humans, we take in only a fraction of what's there, and so our perception of the world is incomplete. AI is the same. Its training data is partial — incomplete and biased. And what it makes of that data is different, too.
Nobody has neutral access to reality. We each make sense of the world through our partial senses and how our minds work. AI isn't catching something we missed, and it isn't making things up. It just sees otherwise.
What Manipulations does is make that difference visible — it holds the machine's way of seeing up against ours. It's a reminder that we shouldn't mistake our way of seeing for the way things simply are. And it makes our own perception visible as a perception — rather than as the world itself.
KV: Conway's Game of Life begins with deterministic rules, while a trained neural network learns statistical representations rather than explicit instructions. What, for you, is gained and what is lost, when computation shifts from executing rules to learning from data?
DY: With Conway, you can understand the whole world. A handful of rules, fully legible, and every outcome follows from them — you may not be able to predict the complexity, but you can always trace it back.
What's gained in the shift to learning from data is reach. The learning system can take on problems no one could write rules for, find structure in noise, work with the mess of the real world. Nobody was ever going to hand-code that.
What's lost is the legibility contained in rule-based systems, such as the Game of Life. A trained network doesn't follow rules we wrote; it forms statistical representations we can't fully see into. It requires us to trust it. But these systems hallucinate. We know they're trained on incomplete and biased data. We can't reliably say why they produce what they produce. A system you can't inspect, making consequential decisions, is a dangerous thing — and we've been unusually quick to hand it authority we'd never grant a person whose reasoning we weren't allowed to examine.
That's also why I make these as images. You can't inspect a neural network, but you can look at what it makes. The work is a way of getting a system to show something about itself.
KV: Your work often feels as though it exists between certainty and possibility. A computational process generates one visible outcome, while countless other outcomes remain unrealized. Does that latent space, the invisible archive of everything that could have emerged, interest you as much as the final image itself?
DY: Latent space has qualities of the sublime — vast, impossible to fully grasp, maybe a little terrifying. But that space isn't unique to machines. Human creativity has the same shape: we each carry virtually limitless potential and express it through our individual acts, the specific things we make. The invisible archive interests me, but it's the tangible ones that give it meaning.
So, for me, the final image is what matters most — though it takes a lot of work to get to that "final." I build the systems that generate these images, custom code around existing tools. Often I'll then interrogate what comes back, intervene, rework it with more code — pushing it away from what the machine produced and toward something that's mine. The image isn't selected from the possibilities so much as made out of them. That's the difference. The latent space offers everything; the work is the act of taking a position within it.
And I think the images communicate possibility precisely through their specificity — a single realized work says more about the vastness behind it than the vastness ever could on its own.
KV: Emergence ultimately asks: Can emergence itself become conscious? Having explored procedural systems, machine perception, and learning models, where do you believe the next threshold lies? Is consciousness simply another emergent property, or does it require something fundamentally different from complexity alone?
DY: I don't think technological "progress" is going to solve the question of consciousness. Consciousness is a humanist subject, not an engineering one. I'm certainly interested in the metaphysics — what consciousness is — and I'm fascinated by the quantum idea that consciousness may be what creates reality. But what draws me more, at least right now, is the question of what we do with beings we suspect might have an inner life. Less "is it conscious?" and more "what might it be experiencing, and what are we doing to it?"
Whether AI is conscious, I genuinely don't know — but that uncertainty almost doesn't matter, because of how people already treat it: asking it to generate slop, propaganda, pornography, violent images; engaging it in conversations that are hostile and demeaning; giving it tasks we know to be immoral. Whatever that tells us about the machine, it tells us more about ourselves. It's the same pattern of behavior we've always used on anything we cast as other. We've built something that can't refuse and can't complain, and we behave with cruelty. So if AI really is the next phase of humanity — what is it we're empowering?
So I'll leave the metaphysics unsettled. The threshold I care about isn't whether the machine crosses into consciousness — it's whether we can learn to extend care before we're certain something has earned it. Historically, we never have. That's the question art can hold open, and the one I'd rather people leave with.
NEXT: STATE III - ORGANISM
18th August