What happens when we become so enmeshed with digital interfaces that they're no longer something we start and stop using, but more a persistent presence that we live with? What happens when this happens at the level of our bodies and changes how free we feel to move physically?

Digital technologies can be used to transform our intentions, to narrow and focus them and to broaden and mutate them. Today we interact with that technology through conventional, designed interfaces. These are built upon layers of intellectual representations of human activity: documents, images, videos, profiles, likes, emojis, code. Every time you use software, you’re invited to represent your thoughts, your intentions, your wildest self, in terms of these structured datatypes that someone has invented and built. As a human, you need to adapt around them.

There is at least a whiff of neutrality to it all. For example, while the content on social media is personally tailored and distributed according to very secret algorithms, the boxes into which it is formatted are mostly the same for everybody. In our expressive offerings to the algorithm, we all express ourselves through the same uniform interfaces.

But an artefact of this uniformity may be a kind of mass conformity that we can’t quite see because we’re all partaking in it. Not just a conformity of ideas and ideology but of modes of expression and the importance we give them. We make little videos and we write little messages and we take photos and what doesn’t exist within this documentation can seem as unreal as things we thought but didn’t say aloud.

The tool grants us power but its way of seeing the world is projected back onto us. Through this bargain, it institutionalises us into thinking its way is the way.

Subscribe now

Back in 2011 I was making body-controlled interfaces for music – specifically the homebrew world of motion-capture unlocked by the repurposed Xbox Kinect camera.

By 2014 I was using it to creating an interactive dance performance for the stage with the dance artist Jan-Ming Lee, called This Floating World.

The Kinect unlocked full-body tracking for cheap without attaching things to the body. Tracking the whole body gave so much expressive potential. You could move freely. You don’t have to hold onto anything. It could give you skeleton tracking, a silhouette or a point cloud of a body.

This was more data than my brain could design for. I could easily code trigger zones, and parameters that go up and down as your hands go faster or slower - but these all see the body as a set of independent labelled particles moving through space. They’re about body parts rather than the body as a whole. Essentially, my code was reducing the immense dynamic range of the body to these tiny little digital abstractions I just invented.

Now there’s nothing intrinsically wrong with this. It’s a lot of fun. But there’s an interesting subtlety when the whole body is being tracked. If I hand you a button you can put it down. But if I make the touching together of your hands trigger a sound, you can’t escape, unless you leave or hide your hands. One day, rehearsing with Jan-Ming in the studio, she commented that she felt like the system was constraining her movement. I was trying to increase expressivity, but in some sense was having the opposite effect. What’s going on here?

The embodied interface is a selective filter that reduces your expression into an abstract digital representation. But that relationship goes both ways. When our focus moves to the digital output, it’s the digital representation that matters, and the body is now subordinate to the representation. Whatever we do will be interpreted through the limits of that representation. Interaction projects the system’s representation of the individual back onto them. We start to think through the limited range of this abstraction (abstraction as in: stripping away what we don’t need for our model of the world).

I believe all interactive technology has this effect to some extent. But the more we entangle the interaction with our body, the less liberty our body has to move without consequence. It can become a kind of prison. Instead of controlling these respresentations as an extension of our agency, we become limited by their limitations. I think this is the fundamental mistake we make with technology: When you can’t escape it, you become institutionalised by it.

In one sense, the problem is not the interactive system but the fact we’re fixated on its output and losing touch with all the untracked parts of ourself.

Jan-Ming and I took the body-to-paintstroke system from the dance work above and used it to create algorithmic portraits of how someone moves. We called this work Movement Alphabet.

Our drawing system extracted endpoints from the body’s silhouette using a technique called pixel skeletons. It was glitchy with a real personality of its own. The relationship between movement and image was there, but with so much error and noise. Yet, in spite of this, we realised that the limiting factor was how someone feels when they’re confronted with a body-controlled interface, and the kinds of unnatural gestures that situation tends to elicit. So we focused less on the digital system and more on constructing a context to support someone in expressing themselves authentically in movement.

We devised a journey, led blindfolded by a guide whispering in your ear, into a private pod where you’re invited to share significant memories, and then gently assisted to express these through movement. The camera is nestled in a corner and the guide has a little remote control to choose when that movement is transcribed into a portrait.

The portrait that emerges is a fusion of our glitching algorithms, how the person moved, and when the guide pressed the button. They all look rather similar to each other - definititely a reductive representation. Each is more recognisable as our artwork than the participant they portray. But the participant is shielded from this narrow representation while they move. This removes the feedback loop and lets their mind remain focused on their human self.

When the participant emerges from the pod we hand them their printed portrait and a handwritten note of a few words they had mentioned when each one had been taken. They were then led to some cushions to rest for ten minutes.

I found people to be much more attached to their portrait than if they had stood in front of a projector and moved with a feedback loop.

Feedback loops are essential for learning (and pretty much the definition of interaction). But learning can also be toxic, like institutionalisation, like learned helplessness. Today, we have learnt how to express ourselves through little pieces of structured data. For many of us, the pursuit of our goals in life seems to depend upon doing this again and again.

I think AI-mediated interaction will change this. I don’t just mean talking to AI, as many of us are doing today, but bespoke AI-generated interfaces, both visual and embodied. Instead of working with representations designed by a human mind, AI forms its own messy representations statistically from whatever data we set it upon.

So, can we humans stop needing to think through the abstract realm of the machine, and instead have the machine think through the messy reality of the human?

In 2021 I explored this idea with artist and AI researcher Panagiotis Tigas in the project Sonified Body. I wanted to create a dance-to-sound system and to use machine learning to escape my previous experiences of trapping the expressive body behind a digital materialisation of my own preconceptions of what a body is, how it moves, and how that might translate into sound.

We wrote some papers if you’re interested in all the detail, but the summary of it is this:

Given the 25 joints of the Kinect skeleton representation (in 3D space so that’s 75 numbers), we train a model to find a way to compress these into 16 numbers and then decompress them back into the original skeleton. We trained the system on around 15 hours of recordings of me dancing. In other words, we’re saying: given what you can see in these recordings, find a way of representing this body as accurately as you can in just 16 parameters.

This is still a process of reduction. But we’re not saying what to reduce the body to or how to reduce it. We’re not saying which parts of the body are more important, nor introducing concepts like ‘upright’ or ‘open’. The representation emerges rather than being designed. It’s called a latent representation. And while 16 degrees of freedom is a fraction of how the body can move, it’s a huge number to be simultaneously jostling with a computer. But it’s not something you think about, or can really think about – you just play and get a feel for it.

I hooked these 16 numbers up to a drum kit synth and some off-the-shelf effects. I did this quickly without thinking of musical ideas, aiming to keep the sound a raw and direct rendering of the representation. Then we invited dancers in to have a play.

I was very happy with the results. For me, the system offers a kind of openness I haven't managed to achieve through traditional coding. I don’t see the dancer get lost in my own ideas of what a body is and what it does. It’s a way of avoiding the bottleneck of designed abstractions, and instead work with representations that are derived directly from the body itself. And to be clear, we're still using the Kinect skeleton, which is a designed abstraction of the body. We haven’t really escaped the symbolic representation of the body, but we’ve taken a step to disentangle it from how we structure our interaction with the system.

I’d like to hope Sonified Body anticipates a future where AI liberates us from the data-entry mode of being in the world that we suffer from today. But I imagine the intimacy brought by embodied interaction can be used just as readily for harm.

If we look around today and see a world of bodies contorted around phones, minds contorted by addictive designs, and spirits contorted by algorithmically mediated social lives, then we can at least take solace in reconnecting with the body as a site of truth. Just stop, take a breath, notice how you feel. But as the surveillance infrastructure that we mindlessly continue to expand starts to take a deeper interest in predicting our behaviour from our movements, we may find ourselves institutionalised in body as well as in mind.

And yet, for those of us worried about this, I don’t think it helps for us to exclude ourselves from the coming revolution and take refuge in a bubble of moral indignation. We need to build things for ourselves, to own our technologies and our means of expression, and to stay present to what it does to us and those we share it with. We need to have fun with it, and be ready to offer an alternative.

Tim
London, 28 February 2026

Based on a talk originally presented at Future Folk, Cecil Sharp House, London, 6 Feb 2026.

WORK-IN-PROGRESS SHARING

DATE: Thursday 4 December 2025
EVENT: Deep Assignments #03
LOCATION: Apiary Studios, London

///

Diffeomorphism is a project I trace back to 2021 when I trained a StyleGAN3 model on every photo I’d taken. This is a small model, small enough to train on my own PC, and my entire lifetime catalogue held more kinds of image than this model could represent. Some images that came out of it looked slightly familiar. Others were more abstract, but with textures and shapes with a definite connection to reality.

Through these glitches, I felt a deeper understanding of the nature of this AI model itself emerge. I’ve since been exploring its character, gently contemplating what the proto-perception of AI has to reveal about how intelligence emerges in the universe.

I hacked and reworked this model, keeping the bits normally cropped out, animating it by injecting chaotic data at various parts of its process. As the field moved on, I found images I’d previously discarded began to seem more beautiful and significant than those I was initially drawn towards. I wrote software to work these into undulating landscapes of pattern, texture, and vaguely familiar forms. I found them hypnotic.

When we mould a medium to represent the familiar, the medium itself becomes visible through its flaws and limitations: the crackle of the record, the colourless eyes of marble statues, the blocky auras of jpeg compression. Likewise, we can see generative AI through its mistakes and its clichés. But what is the medium that these flaws reveal? I don’t think the answer is bytes, or artificial neurons, or other digital abstractions, but a process of emergent learning, where a web of pieces are incrementally nudged until they hold a representation of a world. StyleGAN3 doesn’t use text at all. It simply tries to replicate what makes the photos I’ve taken look like they look. In the time I’ve spent witnessing this emergent proto-perception, I’ve felt like I’m witnessing something analogous to the natural intelligence of our universe.

This evening (more like tonight or the past by the time you read this), I’m sharing a work-in-progress of some slow-moving forms I’ve generated from the world that emerged in this model I trained four years ago. Seventeen minutes, with a score composed by me and featuring vocals from Adriana Minu. It is a personal meditation on this medium of emergent knowledge, flawed yet eerie, and a mirror through which we might catch a glimpse of one of the building blocks of our intelligent universe.

All this while slop is word of the year. Our relationship with media and craft is changing for sure. I don’t know what it will become. The trend was already well underway towards the dull, shallow and relentless, content lacking the spark of intention that might be traced back to another human soul. That doesn’t mean this moment isn’t profoundly different. In any case, the saturation means we desensitize quicker than ever to images that were remarkable to encounter just a few moments ago. This new work is also an invitation to re-see. In this strange moment of political distrust, insecurity and loneliness, as we reckon with the power over our lives we’ve given to privately owned digital infrastructure, it’s easy to forget quite how fucking remarkable the past five years have been.

Tim
London, 4 Dec 2025

P.S. No plans in the short term to publish the work online. Drop me a message if you’d like to watch it and I’ll see if I can organize something.

Over the years from watching friends, one of the most important lessons I’ve learnt for anyone wanting to be an artist is this: don’t get a job. Failing that, don’t get a full time job. If you can’t do that: don’t get an intellectually taxing full time job.

But for the past year and a bit I’ve been working as a graphics engineer, first for a small startup and now for Figma.

I was a little burnt out as a self-employed artist. Not by making art, or engineering, or living on little money, but the relentless marketing and self-promotion needed to get any project moving. Applications. Social media posts. I have a lot to give to this world, but convincing people to care about my work is not it. As representations of work become more important, they start taking over the process. It becomes less a journey into the soul of the collective unconscious, and more of a rain dance for a rain god who has more customers than he has raindrops. I like dancing, but not like that.

Three years ago, for the opening of Cave of Sounds at the Milan Science Museum, I did a little artist interview for an Italian TV station. I’ve spent much time over the years trying to get a bit of coverage for works – to build connections, build a name (a “profile”), so I can get the resources to make more work, and an audience to see it. So when it comes to press, the immedite impulse is yes yes yes. Finally, somebody cares. But after we recorded, I was thinking that it hadn’t felt right. Why do I want to be on Italian TV?1 This is not my art. This is a subgoal of a plan for a plan for a plan to make some art. The means became its own end. They have a tendancy to do that.

So taking a job feels like something of a sabbatical. But I’m still making art. It’s actually been a busy year by my standards. In May, Cave of Sounds was invited to Macao for the international arts festival there. In July and August, Self Absorbed had its first propert showing at File Festival in São Paulo.

The sabbatical is from the artworld itself, or more, from self-promotion, or even more, from being invested in a specific professional outcome. It’s a sabbatical from reaching for things. From professional ambition. From feeling any obligation to relate my own work to that of the wider world. From feeling like I should be posting more on social media. From nursing abstract representations of myself.

Instead I’m following what’s actually enjoyable, in the limited time I have. It’s kind of working. Things are sparkling a bit more again. I’ve been taking photos (see above). Writing music. Rendering visuals and putting sound to them. Walking round cities with Adriana. We’re back in London. And here I am writing a newsletter again. I have so many ideas to share, many sitting half-written here. And the less any of it becomes embedded inside some kind of cohesive artistic narrative, the more they start to burn.

You know, I don’t want to be flippant about the privilege of employment. I know enough people who are struggling to find any paid work at all and the job I have is a very nice one. Good people, good pay, interesting work. But it does nonetheless confront me with a challenge to my identity. I keep getting flashbacks to that film Juno. The middle-aged husband making commercial jingles for a living who falls for the pregnant teenager whose unborn baby he and his wife plan to adopt. He decides to quit his steady work to try and make it again as a musician in a band, and we shake our head at his pathetic reach for a life that might nurture his soul, but which is long gone. I only saw the film one time 17 years ago while I was still at university, but I remember his sad character vividly. Something was close to the bone. A warning to prioritise my creative soul over the trappings of life, to take my dreams seriously.

But, of course, now there is little time, energy or brainspace to do any of these things. Writing is probably the most intimate work I put out. It has been hard as I reconcile who I am. But I think it is one of the most important things to sustain. The regular act of putting what I think and feel into words has a tendency to make me think and feel more of those. Publishing what I write, extra so. The more honest I can get the writing, the more I can be in touch with the stuff that powers everything.

Tim
London, 16 Oct 2025

Along the way I made some new pictures, which I got really into and then turned into a series of slow, endless animations. I can’t embed them properly using Substack so if you can, please read month’s newsletter on my website here.

But is it your art?

It’s tempting to think a new tool to automate creation, such as an AI image generator, is going to make it easier to make art. But often it actually makes it harder.

First, let me clarify what I mean by art. I’m talking about a form of expression that shares a glimpse of what it is to be us, to see the world like we do, to be a human. Art as a way of bridging the interiority of human consciousness. When someone does it to us in a way we haven’t experienced before, there can be a profound sense of connection. It’s an enmeshment of minds. This isn’t the only meaning I have of art, but it’s the one that’s relevant for what I have to say here.

Any tool I bring to the creative process holds within it some assumptions of what I’m going to do with it. The tool carries within an essence of the interiority of those who made it. It makes it slightly easier to say what has already been said by others with that tool.

Those of us who design and use creative tools wrestle with this paradox. Sometimes it seems like every time the machine makes something easy, the artist then has to find a way to make it difficult so they can find a voice that feels their own. Or more, everything machines make easy quickly becomes generic. The artist needs to find some level of fuckery to make it personal again.

Generative AI is the next generation in these tools. Something feels different because the range of things it can create is so broad, and its capacity for imitation is precise. It makes it so easy to generate an image that looks so polished, that it’s tempting to forget that we are merely navigating to a room within the interiority of the AI itself.1

At this point, it’s easy to get lost into a debate about whether generative AI is creative in its own right or simply an elaborate form of copy and paste. I think this is a distraction from a more useful question: does it allow me to infuse my inner world into some outer material form? Is that really my inner world I’m expressing?

This doesn’t nullify the possibility of expressing myself with AI. It’s just more like sharing an image I’ve found somewhere with someone. An image resonates with me, and hopefully with the person I’m sharing it with, and there’s a connection there. My choice gives a pinprick of information about who I am. But the full weirdness of my inner world remains locked inside my head.

Why can it feel so exposing to draw or sing or dance in front of people? Perhaps, because we cannot do these things without revealing something of who we are. The more connected we are to the process, the more our interior world leaks out.

Originality is easier when crafting a work rather than specifying it. If my creative process involves continuously shaping a medium towards a finished work, I can’t help but leave an imprint of myself within it. Even if I’m trying to make an imitation, the limits of my craft will leave flaws and these will be uniquely mine.

Computation brings power. It’s streamlines the act of making. But the more it streamlines, the more it detaches the outcome from the mind of the creator. You’ll end up with something that’s been done before, or with something random. But not something you.

Making art with AI is hard

In a recent essay, the sci-fi author Ted Chiang argued that it’s not possible to make art with AI because it leaves the artist so few choices. He gives the example of a novel. Every word of a novel is a choice made by a writer, and the work of art emerges through the nuance of these details. But type a 100 word prompt into an AI novel generator and you’ve made so few choices that what you end up with is more like a statistical averaging of existing works weighted to fit your description. There just aren’t enough choices.

I think he’s on the right track. But using AI to remove the countless choices from the creative process doesn’t make it impossible to make art. It just makes it much, much harder.

Experience is worth a lifetime of pontification, so if you’ve not already had a go with an image generator, then it’s time to pop your cherry. Head over now to one of the many online generators (this one doesn’t require registration), type in a prompt and watch it make you an artist. Pay attention to how you feel at each stage of the process. Then come back here.

Here’s how it is for me. I struggle to write when confronted with the potential to make absolutely any image we could describe. Every idea I come up with seems unoriginal and pointless. So, giving up on greatness, I think, whatever, and go with something I’m pretty sure I’ve seen before, like a dog in an astronaut costume. I was given a genie that could make any image I could imagine, and I asked it to make an image we already have, an image even that I already saw. My relationship with this image is so distant that I don’t even save it. (Even my random doodles I keep in a box on the shelf.)

It’s taken me years to make AI art that feels my own. It’s involved training many models on hundreds of gigabytes of my own data and coding systems that do unusual things with them. It’s a process of hunting for something that is me rather than merely the tools I’m using, which involves spending time to really understand what those tools do. It was as difficult as any other artform I’ve invested time in. In essence, I had to find a way to build a craft of my own from a landscape of streamlined processes.

Human expression is a dance between mind and matter

A new wonder technology has arrived that can transform human potential, but it seems to have brought a dystopian sense that there’s now less purpose for us humans. I think this perspective may be a knot we’ve tied ourselves into, one suggesting we’ve lost track of what it means to be a human.

There's something odd to me about how Chiang frames craft in terms of the number of choices. It reduces human intention into the one who chooses rather than the one who does, as if the distinction between writing and prompting were just a matter of scale. Perhaps this framing makes sense to a writer, and maybe a traditional pen-and-paper composer, operating in their symbolic realms. But I can’t quite connect it to drawing, dancing or playing music.

I consider the written word to be an early example of digital abstraction, where the continuous signal of speech is reduced to a finite set of discrete pieces: words or letters. (Digital because the pieces are finite; abstract because they can now be considered independent of their context.)

Bureaucracy and computers have continued this process to the situation we have today where the gamut of human intention is routinely distilled into button presses and swipes, and material reality into image and sound recordings. The more we use computers, the more we end up thinking that manipulating the abstract entities held within them is what thinking is all about.

This kind of computational thinking dominates us today. It makes it easy to see everything we do as a sequence of discrete choices. I think many of those building AI products see human endeavour in this way. If only we could simply express our intentions in a few words and have the assistant do the work for us. Isn’t that what we all want? To get things done without having to actually do them?

It’s similar to the neoliberals who define human freedom in terms of how many brands we have to choose from. Or the contemporary focus on consent as the ultimate arbitrer of morality, from sex to GDPR cookie banners. Choice is definitely better than no choice, but it’s not the same thing as freedom.

Here’s how I would like instead to frame human expression: A continuous, mutually influential dance between mind, medium and environment where some aspect of the mind as it experiences its environment becomes imprinted on the medium. It doesn’t need to be conscious or even intentional.

If I draw a circle on a piece of paper, the path my pencil takes will be unique. It’s not so much that I’m making unique choices but more that the movement of my hand is continuously shaped by the activity of every other part of me. How open is the channel of influence between me and the world, and between me and my medium? How sensitive am I to how the process unfolds? Can I feel the breadth of my options? Can I try something a tiny bit to see how it feels then change my mind before anything serious happens? These are questions of freedom lost when we see our lives as a sequence of choices. Technological metaphors can still be useful: bandwidth, resolution, feedback loops, degrees of freedom.

When I choose 16 English words to prompt an AI image generator, I have more options (10⁸²) than there are atoms in the universe (10⁸⁰).2 Language is not short of expressive power. Yet there’s a fair chance that I’ll type the exact same prompt as someone else. To prompt an AI to say something that is truly my own, I’ll have to tweak it and go again. And again, and again. It’s the slowest dance in the world. And if I find this process frustrating, I can only imagine what it’s like for the poor AI.

Hence, generative AI has confronted us with a new craft: prompt writing. It may seem simple because we can quickly generate results that would take a lot of work with the old crafts. But the simpler a craft is, the more ingenuity is required to do something exceptional with it.

The delusion is to consider craft as a barrier to making art, rather than the channel that makes art possible. The smaller you make that channel, the harder it is to communicate through it. Hard, but not impossible.

Machines that help humans stay human

Once we see human expression less as a declaration of preference and more as a dance between mind and matter, we can see that our problems with AI are not necessarily of AI. They're problems of what we're trying to do with it: to create an assistant to whom we can delegate the details of what we’re doing. This framing aligns with a vision of us as the boss who calls the shots and makes the hard strategic decisions. Hence, we end up either struggling to distill our desires into a few words of instruction, or adapting our desires until we find one that can be expressed in that way. I’m not against this framing in the right circumstances. For example, I use AI assistance extensively when coding. There, it doesn’t have the challenges that emerge when I try to make art with it. There, what I’m trying to do can be broken down into explicit pieces that have an optimal implementation.

But with art, I need to immerse myself in my subject and my medium, so that these can infuse into me. I want to be dreaming about it, and I want those dreams to shape it in ways I’m not even conscious of. I want my tools to disappear inside of me, becoming extensions of my body. This is harder when using computers because my relationship with my medium is mediated through the abstractions of software.

In traditional software, every single abstraction - the ‘window’, the ‘post’, the ‘button’ - has been coded in advance. Breaking down processes into explicit abstractions is what we do as software engineers. But in the neural network of an AI, the structures through which it finds meaning in data emerge from that data itself. It can hold ambiguous ideas and connect concepts in ways that would be difficult to write out explicitly. Interacting with AI could come much closer to a dance between mind and medium than any of our existing hand-coded software.

This is bad news if you’re trying to automate a legal decision, or have AI fill out your accounts, or another procedure where there’s a real distinction between correct choices and wrong choices. But it opens the door for bringing the embodied parts of ourselves into our digital lives. Instead of detaching us from process, AI could entangle us with all the details in a way we can’t even imagine with regular software. Think how much we humans can communicate through even the most subtle sounds and movements. Imagine spending an hour with a generative AI which could understand and react to this side of you in real time.

It’s not easy to imagine, but I had a go in my project The Wilds.3 I built a real-time, two-way connection between the moving body and a generative AI system, and started exploring 16-dimensional space with the shape of my body, releasing into my intuitive capacity for physical exploration. I remember the first time the dancer Catriona Robertson was confronted with an early version. She was going non-stop for 40 minutes, exploring, all while still being completely herself, until we interrupted her.

I find coding and improvised dance to be opposite parts of what it is to be human, representatives of the two halves of my brain. For all its risks, AI may open opportunities to enter the machine with my whole self. I don't mean this in a cyborg sense of becoming one with the machine. Quite the opposite: we've already become one with the machine as the only way to use a computer today is to think computationally through its abstractions. It's the potential to release us from this that I'm interested in.

If you see art as a sequence of choices, then yes AI probably does seem like an affront to human dignity. But if you see art as an enmeshment of minds, touching each other through unstable realms of matter and ideas, then the affront to human dignity was already there in the relentless bureaucratisation of human life into digital data. AI is computation that emerges through learning rather than through design. It holds the potential to become a technology that lets us be more human than we are now, as dancing animals rather than wannabe mid-management.

Art is safe but artists are not

Of course, I can’t quite end the story there. I started by narrowing my meaning of art to expressive acts that carry a glimpse into what it is to be a human in this world. After feeding everyone, I think the this kind of creation is the most important thing a society does. However, artists also need to be fed, and few get by without freelancing out some part of their craft. Art may be safe from AI, but artists are not.

AI pundits like to point out how when photography supplanted the painter’s niche, it pushed painting into new and exciting realms. But they forget about the artists earning a living making woodcuts for newspapers or painting family portraits. Most images that are bought are bought to do a job.

Mass production, photography, the record, offshoring, and now AI all offer a cheap alternative to forms of human labour that feel intrinsically valuable. They bring economic transformations that centralise production, homogenise culture, increase the power of capital and transfer power from individuals to industrialists. What’s more, it took over a hundred years before a record could really sound indistinguishable from a live musician, and near that before a photo could capture colour like a painter. AI is moving much faster. There is less time to adapt and so the effects will be more violent.

This is to say nothing of how vulnerable we humans become as we come to rely on intelligent systems that we cannot predict and that hold a nuanced perspective on what makes each of us tick. We’re rushing into this during a moment where technology is centrally owned and controlled.

Nothing is inevitable.

Imagine if running a commercial AI image generator required a licence, similar to the one a pub needs to play music, where the fees go to pay the artists whose influence can be found in the image.

Imagine if we required each self-driving taxi to be matched to a qualified taxi driver, who could sit at home and watch the income come in, and generally take responsibility for keeping it clean and in good order, perhaps put a bit of personality into the interior.

Imagine if making automated decisions about people’s lives was outlawed, and every credit check required an individual to sift through the evidence dug up by an AI and then make the decision subjectively, which they might be required to justify in appeal.

Imagine if the government decided that the police can figure out how to do their job without face-recognition technology, given they’ve managed to do so for the past 195 years.

To appropriate the apocryphal Zizek quote: You don’t hate AI - you hate capitalism. And communism. And all the rest of them that prioritise efficiency over humanity.

In the meantime, I think our best bet is to invest in our humanity. That word used to mean what distinguishes us from other animals: compassion, love, art. But now it’s as much what distinguishes us from machines: individual difference, sensitivity, distraction, consciousness. The human artworks that will survive (and thrive) in the age of AI are those that can be confidently traced back, in some way, to the mind and will of a conscious being.

Tim
Glasgow, 4 November 2024

Hey everyone, hope you’re having a good summer/winter.

I made a music video for my friend Zac Gvi’s band The Nonidentical.

I used the same customised StyleGAN3 model from much of my recent visual work, sticking here to high magnifications and abstract areas on its latent space. I modified my setup to let me control the model directly from Ableton Live, so I could score the visuals to the music.

This track, Bijoulacra is from their recent album Material Tropes.

Tim
Glasgow, 31 July 2024

On Thursday I’m doing an AMA (Ask Me Anything) for Music Hackspace on AI and Artistic Identity. Here’s my starting point for it. Details for the event at the bottom.

We need better questions to understand how new technology changes us. There are a lot of hot takes on how AI is going to transform human creativity, for better or worse. But there are many possible futures ahead of us. It’s on all of us to navigate towards a future we want to live in. To help, I have three questions to share from my experiences making interactive art.

Digital Interactive Art is a space to explore the whole gamut of possible human-machine interactions. Most technology is made to solve a problem, and for profit. But as art, technology is freed from utility, and it becomes easier to explore it for what it is rather than what it can do for us. My interest is how our dynamic with a machine changes us - as a person and as a society. Building interactive art lets me put an audience into the place of someone who doesn’t understand how technology works, or how it even should work.

My favourite place to do this is in the gallery, in the form of a new physical interface. Here, you see firsthand someone’s journey of discovery. You see how social dynamics affect things. You see how vulnerable they are as they try to retain their identity and express something of themselves when confronted with an opaque system whose reactions remain unpredictable.

The first question I ask of a new interactive system is this: What roles are being assumed here?

Is this machine a tool to help me do something? Is it a space for me to explore? Is it a creative partner to collaborate with? Is it a petty official with opinions about what I should be doing? A toy? A trusted colleague? Who am I? A user? An explorer? An artist? A subject?

Of course, it’s just bits and bytes. But we can’t help but project what we know onto what we encounter. The roles we project are how we get a sense of what ‘should’ be happening, and when a system isn’t working properly.

The roles we project aren’t random. There are many cues, both within the system and how it’s presented. The person who made it imbued it with assumptions about the roles you and it would have, whether or not they were conscious of those assumptions. With any designed interaction, there is always an invitation to take a role.

In my own niche, an example of this is whether I call an artwork a musical instrument or an interactive sound installation. Many things I make could be described as either, but an instrument puts the person in the role of a creator, whereas an installation puts them in the role of an audience. It leads to a different experience.

A sense of the roles helps reveal the second question: Who feels responsible for what happens?

I don’t just mean who do we decide is responsible, or even who is actually responsible, but who do we feel is responsible? If something unexpected and exciting happens is heard in a gallery, people look at where the sound came from. Then they look at who they assume is responsible, if they aren’t in the same place as the sound. I don’t think we even need to think about this - our minds seem to always have an intuition ready, as part of our basic survival instincts.

If you interact with a system and its output becomes unexpectedly pornographic, do you feel embarrassed? Like, people are going to assume this is somehow a reflection on you?

In my PhD, I described this as perceived agency. Who do we feel is in command of what happens? Who gets the credit? Who gets the blame?

Often, how we perceive agency in a technological interaction doesn’t align with the reality of who actually has objective agency. This leads to my final question: Is this an honestly designed system?

Does the system actually give you the power it seems to imply it does? Does it promise you the world and leave you feeling like a failure when it doesn’t deliver (I’m looking at you all dating apps and social media)? Does it present as a musical instrument but only let you make the same tired tunes as everyone else? Does it present as an open communication platform but actually choose who gets heard based on an opaque algorithm that selects content based on third party interests (again, social media...)?

The dishonesty usually goes one way. It’s easier to sell people power then give them a toy, than it is to sell people a toy that’s actually a powerful tool.

Dishonest systems can humiliate people. They make us feel ashamed for the bad things they do, and like failures for the good things they don’t do. Even in the contained world of an interactive artwork in a gallery, I’ve seen people humiliated by my own work when it fails to deliver on the promises it implied. An instrument that works for everybody else but then breaks when they try to use it. As the designer, it’s my fault, yet I see them blame themselves.

When things really count, in socialising, in dating, in politics, I feel it should be a lot less surprising that technology turns so many people aggressive or sad. We may know rationally that a system is rigged, but I suspect at a deeper level we can’t quite escape our intuitions of blaming the people involved.

Tim
Los Angeles, 12 June 2024

Details for the AMA: It’s on Zoom, costs $7, on Thursday 13 June 2024 at 6pm UK time. Registration is here: AI and Artistic Identity: A talk with Digital Interaction Artist Tim Murray-Browne.

I've just published Diffeomorphic Landscapes, an online version of my recently exhibited image series created by my customised AI model trained on my lifetime archive of photos.

I've come to see that pictures that are both in focus and out of focus from the car are like memories – certain parts are really clear and other parts are hazy. That's the thing that resonates for me in the landscape pictures; they reflect how the mind works. They're a metaphor for memory.

– Todd Hido - Landscapes, Interiors, and the Nude

The above quote of Todd Hido, one of my favourite photographers, has been stuck to my wall for several years. He's talking about photos taken through the car windscreen with rain running down, like #6097 (which I also have stuck on the wall). The water blurs and defracts the view. As detail is lost, the atmosphere of place is deepened.

There was a time in my life where I thought I might become a photographer. Expressing my inner world by moving through the outer world, seeing, framing, capturing, selecting – it's a creative process that works for me.

It's a similar process to generate images with AI. Both make instant something that was previously time-consuming, which can lead to a glut of images that quickly lose value. In both, the challenge emerges of how to make these images actually carry the thoughts and feelings of the person creating them. My answer here has been to train my own AI, find my own ways to misuse it, and then select from a set of perhaps 5000 generated images. I chose the crop and frame for each image. Beyond that, they are unedited.

The photograph lies by taking a single instant of a dynamic world and presenting it as an image: a static, unchanging reality that we see again and again. It's one reason why photographing people is so difficult. AI lies by fusing many different instants into a static image of a single instant that itself never happened. Like a diaorama, it is in some sense truer in that it captures a broader range of the world even if – unlike the photo – it does not let us positively conclude any specific details of the world.

Much like the haziness of Hido's images, I find these AI generated images analogous to memory. Details nestle in a sea of vagueness.

My own memory is non-visual. If I close my eyes I can't bring to mind the things I saw a few seconds ago. I no longer seem to have a mind’s eye. Yet I can turn a corner in a forest and instantly realise I crossed in this particular neck of the woods four years ago. But then, I can have a conversation with someone at a party and ten minutes later I can't spot them in a crowd, even if I remember what we were talking about. When I'm awake, visual experiences only ever come from my eyes. When I remember and imagine, I feel narratives, places, sounds, relations, the motion of bodies, the weight and presence of objects.

My favourite from the series is Cosmic Insignificance Therapy (0112). Fragments of image tie through movement into a myriad of relations.

The full image series can be seen in the online exhibition of Diffeomorphic Landscapes. Also on that page is a recording of the talk I gave at the opening, as well as a tidied-up transcript who prefer to read than listen.

Tim
Glasgow, 7 Apr 2024

Talk: Joy and ethics in making AI art, or how I found the universe in my data self
Date: Thursday 25 January 2024, 6pm - 7pm
Location: Online, or at 22 Richmond St, Glasgow.
RSVP for Zoom link and details

Please join me tomorrow for a talk journeying through my past 4 years making art with AI, revealing AI through art, and my efforts to overcome my techno-pessism and fall in love with AI.

Throughout this time, I’ve begun from the premise of using my own data in my AI models. This is partly to sidestep the ethics of harvesting data from others, but mainly it’s an instinct to avoid blending my own individual voice into an amalgamate of human expression. It’s led me to make much more personal work than I’d done before.

I’ve still found myself wandering into unfamiliar ethical landscapes, from morphing together the faces of those who have been dearest to me, to creating a solo performance with an AI voice trained on early recordings of my wife, Adriana Minu, developing her vocal performance practice. I think contradictions between intentions and realities can be particularly revealing about what’s missing in how we think about data and AI.

More recently, I’ve been lost in hyperspace, working with glitches and physical interaction to build an intuition about how this AI model that I trained in my own image(s) perceives the world. The process has left me contemplating the nature of the universe, the emergence of reality through combinations of squiggles, and the central role of perception in it all.

I’m just putting my notes together now for it, and I’m very excited to be sharing much of this stuff. Hope to see you tomorrow.

RSVP here

The talk is being hosted by the Scottish Universities Insight Institute and Scotland Europa.

Tim
Glasgow, 24 Jan 2024

The gap between Christmas and New Years is always a time of year for reflection for me. This year, I’ve been thinking of musical cadences as a metaphor for the steady drip of artworks and projects that my year has comprised of.

You may not know music theory, but if you listen to western music then you’ll have a feel for cadences. If you listen to the bassline in the Beatles’ When I’m Sixty-Four. The final line of the verse, ‘when I’m sixty-four’ ends with a feeling that we’ve arrived somewhere. The harmony of that line follows perhaps the most important cadence in pop and jazz: 2-5-1.

Officially, a cadence is a sequence of chords that leaves you with a feeling of conclusion. But within that conclusion is a sense that we’re starting something new. The feeling of ending and the feeling of beginning are the same.

In my work, I feel the cadences as a potential ending to a work, or a project, or a chapter in life. It’s a moment when I need to decide whether the next step is to dig in deeper, or to start something fresh.

The first academic conference I ever went to, I saw a paper by Matthias Mauch where he’d trained a simple statistical model over the Beatles’ entire song catalogue trying to predict the next chord based on the previous three chords. Then, he looked for places in the music where the model was most uncertain about what was about to happen next. Those moments of maximal uncertainty were right after the 2-5-1 cadence. I liked this because it gives an alternative perspective on what a cadence is: it’s a moment when I don’t know what’s coming next – not because I’m clueless but because the world is unpredictable right now. A feeling of conclusion, a feeling of a new beginning and a feeling of uncertainty are all part of the same moment, a moment of openness and possibility.

While a cadence feels complete by itself, one of the tricks of harmony is to reveal those few chords as part of something bigger and more profound. In All The Things You Are, the harmony keeps resolving, but then another chord arrives which reveals that resolution to just be part of a larger unresolved picture.

This sequence of endlessly descending a 5-1 interval is common across Western music, including Pachabel’s Canon and Fly Me To The Moon. But you can overdo it. Michael Nymen’s Chasing Sheep Is Best Left To Shepherds does it until we loop round the entire circle of fifths and end up back where we started, and on it continues. It feels like the soundtrack to someone who never takes a moment to look back and take stock.

Another kind of repeating cadence appears in works from the Classical and Romantic eras where the piece is clearly at its end but endlessly repeats a final cadence. 5-1. 5-1! 5-1!! My favourite example is actually Dudley Moore’s parody of a Beethoven Sonata. It’s like two new lovers endlessly saying bye to each other as they struggle to end a phone call.

vIt can be hard to conclude something that feels so momentous. But most projects feel momentous to me when I’m immersed and enjoying them. Likewise, a project can suddenly seem trivial and pointless just at the moment I get to the difficult and boring part.

Sometimes I fear I’m stuck in a loop, continually returning back to the same place, like in Nymen’s Chasing Sheep. But I can look back at other times where I’d only just tapped into a brief well, but gotten bored and moved on.

My resolutions for 2024 are too fluid to commit to writing. But I’m most excited about the concluding type of resolution: projects that have said what they need to say, stale goals, things I attached to my identity but forgot why. Making space for the new.

Tim
Glasgow, 1 Jan 2024

PS Thank you to everyone who came to the opening of Small Frame Infinite Canvas. If you’re in Glasgow and still haven’t seen it, the exhibition will be open for an extra few days on January 3rd-5th and 8th-9th.

I’m excited to announce an upcoming solo exhibition Small Canvas, Infinite Frame.

South Block Gallery, Glasgow
1–23 December 2023 (Mon-Fri 9am-5pm)

Artist Talk and Opening:
Friday 1 Dec, 6pm - 9.30pm (RSVP via Eventbrite)

My first solo gallery show! My first print show! Physical media! Lots to talk about. But I’m deep in the reeds… For now, here’s my text for the show.

SMALL FRAME, INFINITE CANVAS offers an alternative perspective on generative AI through creative coding, glitch and personal memories.

Each image is generated from a single customised AI. I trained the AI to imagine new memories based on my lifetime archive of 25,000 photos. It has only seen what I myself saw at one point and felt moved to record. The images it generates echo my visual world: Scottish landscapes, tree branches criss-crossing under the sky, rock textures from Cornish beaches and the contours of the human body. (I excluded photos of other people’s faces, but kept many close up studies of the human body.)

The range of photos I’ve taken are more diverse than this AI was designed to handle, so it struggles to make its creations ‘lifelike’. Artefacts appear revealing its own endless world of algorithmic creation.

I modified the AI to expand its frame, creating ever larger images that bring these hidden glitches to the foreground. Each image here contains a small square frame near its centre. As we move away from that optimised frame, the image disintegrates into an endless canvas of offcuts and mathematical building blocks.

Our relationship with AI is still fluid. In my mind, it jostles for contradictory roles: a tool, a collaborator, a pathological copycat, a new kind of mind – sometimes all these at once. The technology is racing ahead faster than my intuitions can.

A common approach to training AI is to harvest huge amounts of images off the web, aiming for a monolithic AI. When I began researching AI in 2018, I felt a strong instinct to train my own system using my own data in the safety of my own PC. If we all express ourselves through the same systems handed down from above, we will surely all find the same paths and end up saying the same things. I needed a way to make it my own, to fuse its world with mine.

In 2021, I stepped back from the overwhelming stream of new AI systems to dig deeper into what’s possible with this single AI I trained. In the years since, I’ve developed a relationship with it. As with many AIs, at first everything is overwhelmingly magnetic. Then it all starts looking the same. But then, after some time, some things emerge that resonate more deeply.

In making the leap to the physical world, I wanted to embrace the scale of the images. They combine detail and scale in a way that can be lost in the fluidity of digital display.

Created with support from Creative Scotland awarding funds from the National Lottery, Wasps Studios and Preverbal Studio. This work uses the StyleGAN3 AI model, developed at NVIDIA by Tero Karras, Miika Aittala, Samuli Laine, Erik Härkönen, Janne Hellsten, Jaakko Lehtinen and Timo Aila. Special thanks to Adriana Minu!

Tim
Glasgow, 16 Nov 2023

I haven’t yet managed to get my writing brain back into gear since the summer. But there’s a lot going on to share.

Just happened:

• My first AI-rendered film, Cosmic Insignificance Therapy, premiered at INTERSECCIÓN film festival (19 Oct, A Coruña, Spain)

Coming up:

• A Short Ride Through Hyperspace, a new immersive AI-rendered AV installation at Frequency Festival (26-29 Oct, Lincoln, UK)

• Open Studio day (Sunday 29th, Glasgow, UK)

  ---

Cosmic Insignificance Therapy part of official selection at INTERSECCIÓN

Cosmic Insignificance Therapy made the international selection for the INTERSECCIÓN film festival in A Coruña, Spain. It had its premiere there last Thursday.

I detoured to the festival on my way to Milan. It was my first film festival (as an artist rather than a punter).

Usually my official attendance at a festival means exhibiting an installation, which involves months of preparation, shipping logistics, days of work to install and non-stop maintenance while there. To be invited courtesy of a one minute film I made seven months ago seems a bit absurd. I feel like I picked the wrong medium to be working in.

The trip has been a reminder of how much I get out of travelling to arts festivals. Navigating a strange city with an arbitrary and amorphous posse of people is how I’ve made friends with many people in my field. I made the most of the pandemic, diving deep into research, but I found it professionally isolating. Setting up zoom calls with people really doesn’t cut it for me. Meeting for coffee is better but still insufficient. I find connection and friendship come from doing things together.

Of course, most festivals are abroad, and the travel has its environmental impact. I don't have an answer to that other than trying to combine multiple things into a few trips. I am confident zoom is not the answer. I felt better about heading to this festival knowing I could combine the trip with the Cave of Sounds deinstall.

A friend told me there aren’t many arts festivals in England right now because organisations on the Arts Council’s four year funding cycle were due to reapply in 2022. Programming festivals is high risk when lockdowns might arrive at short notice.

A Short Ride Through Hyperspace

Frequency Festival have commissioned a new installation based on Diffeomorphism, my series of short films voyaging from the micro to the macro of a singular point in an AI’s latent space. (Cosmic Insignificance Therapy is the first in this series, World Without End the second.)

The new piece is a 61 minute, endlessly looping voyage into different points of the same customised AI model (StyleGAN3) that’s generated many of my images and videos of late. It will be presented in a 3xHD cave format with projections on your left, front and right.

It’s called A Short Ride Through Hyperspace. Unfortunately, I decided that after the festival’s print deadline so the title there will be Diffeomorphism. It’s showing Thursday through Sunday (26 - 29 October).

A big part of the piece for me has been composing the soundtrack. As the AI is trained on my own photo archive, the images it creates are uncanny fusions of my own memories. Following this theme for the sound, I’ve drawn on the many field recordings I’ve made over the years, as well as choral works that are significant to me, and a short sample from a MIDI synth composition I wrote when I was 7 or 8 (thank you Dad for thinking to record that onto a tape).

I’ve written before about the different types of memory the AI gives me in comparison to that recorded in photos and on tapes. AI distorts by blending across contexts. Recordings distort by capturing and repeating a single moment. So in a way, with the AI composing the images, and me composing the sounds, we’re both in the same game of seeking out sense and meaning from an archive of decontextualised recordings.

This is a slowly evolving, slightly hypnotic experience. It’s an installation, so people will come and leave as their attention demands, but I’m hoping to entice people to spend a bit of time to space out and let the experience wash over them. It’s showing in a vacant unit in a shopping mall, which I’m excited about. Also, I’ve done my best to arrange for comfy seating and, as a bit of an experiment, a station offering free herbal tea – basically, the two things I often find myself longing for in video installations.

I don’t believe we’ve got a huge Lincoln contingent on this list but if you’ll be there then definitely let me know and we can hang. I’ll be there Thursday and Friday.

Open Studio day

Sun 29 Oct 2023, 11am - 4pm
77 Hanson St, Dennistoun, Glasgow, G31 2HF

For those in Glasgow, the weekend of 28-29 October is Open Studios day at Wasps Hanson Street where I’m based. I’m travelling back to Glasgow on the Saturday so my studio will only be open on the Sunday.

Loads of other artists in the building will also be opening their doors. There are many talented painters, sculptors and ceramicists in the building. As far as I’ve discovered so far, I’m the only digital artist there.

I’m thinking I’ll rig up a projector in my studio to show a smaller version of the new work (A Short Ride Through Hyperspace), though if there’s anything specific you’re interested in then drop me an email ahead.

Much love. More soon.

Tim
Milan, 23 Oct 2023

My friend Dom Aversano interviewed me for Music Hackspace. Dom is a friend and fellow artist from Cave of Sounds, back in 2012.

Our topic is How to design a music installation? As is my nature, I start with the philosophical - musical emergence, audience freedom, the often-missing sense of creative authorship, and, eventually, how to get an installation out into the world. Along the way we also consider how AI continues to dissolve the artist/audience dichotomy, and the ups and downs of ‘democratising’ music.

Here is the full interview (also published on Music Hackspace and Dom's newsletter Liner Notes).

What do you think are the most important things to consider when building an interactive installation?

First, you need some kind of development over time. I used to say narrative though I’m not sure if that is the right word anymore, but something needs to emerge within that musical experience. A pattern or structure that grows. Let’s say someone arrives by themselves, maybe alone in a room, and is confronted with something physical, material, or technological, and the journey to discover what patterns emerge has begun. Even though an installation is not considered a narrative form, any interaction is always temporal.

Second, has to do with agency. It’s very tempting as an artist to create a work and have figured out exactly what experience you want your audience to have and to think that that’s going to be an interactive experience even though you’ve already decided it. Then you spend all your time locking down everything that could happen in the space to make sure the experience you envisaged happens. I think if you do this you may as well have made a non-interactive artwork, as I believe the power of interactivity in art lies in the receiver having agency over what unfolds.

Therefore, I think the question of agency in music is fundamental. When we are in the audience watching music a lot of what we get out of it is witnessing someone express themselves skillfully. Take virtuosity, that comes down to witnessing someone have agency in a space and really do something with it.

How exactly do you think about agency in relation to installations?

In an interactive installation, it’s important to consider the agency of the person coming in. You want to ask, how much freedom are we going to give this person? How broad is the span of possible outcomes? If we’re doing something with rhythm and step sequencing are we going to quantise those rhythms so everything sounds like a techno track? Or are we going to rely on the person’s own sense of rhythm and allow them to decide whether to make it sound like a techno track or not?

It all comes down to the question of what is the point of it being interactive. While it is important to have some things be controllable, a lot of the pleasure and fun of interactive stuff is allowing for the unexpected, and therefore I find the best approach when building an installation is to get it in front of unknown people as soon as possible. Being open to the unexpected does not mean you cannot fail. An important reason for getting a work in front of fresh people is to understand how far they are getting into the work. If they don’t understand how to affect and influence the work then they don’t have any agency, and there won’t be any sense of emergence.

Can you describe music in your childhood? You say you sang in choirs from the age of six to twelve. What was your experience of that?

At the time it burnt me out a little but I’m very thankful for it today. It was very much tied to an institution. It was very institutional music and it was obligatory. I was singing in two to three masses a week and learning piano and percussion. I stopped when I was about 13. I had a few changes in life, we moved country for a little bit and I went to a totally different kind of school and environment. It wasn’t until a few years later that I picked up the piano again, and only really in the last couple of years have I reconnected with my voice.

Your PhD seemed to be a turning point for you and a point of re-entry into music. Can you describe your PhD, and how that influenced your life?

I began doing a PhD looking at generative music, and as I was trying to figure out what the PhD would be I had an opportunity to do a sound installation in these underground vaults in London Bridge Station with a random bunch of people in my research group. They were doing an installation there and someone had some proximity sensors I could use. There was an artist who had some projections which were going up and I made a generative soundscape for it. Being in the space and seeing the impact of that work in a spatial context really shifted my focus. I felt quite strongly that I wanted to make installations rather than just music, and I reoriented my PhD to figure out how to make it about that. I was also confronted with the gulf of expectation and reality in interactive art. I thought the interactivity was too obvious if anything, but then as I sat and watched people enter the space, most did not even realise the piece was interactive.

How do these questions sit with you today?

From an academic perspective, it was a really terrible idea because a PhD is supposed to be quite focused, and I was questioning how can you make interactive music more captivating. I had this sense in my head of what an interactive music experience could be, and it was as immersive, durational and gripping as a musical experience. Nearly every interactive sound work I was finding ended up being quite a brief experience – you kind of just work out all the things you can do and then you’re done.

I saw this pattern in my own work too. My experience in making interactive sound works was much more limited back then, but I saw a common pattern of taking processes from recorded music and making it interactive. My approach was to ask ‘Well what is music really? why do we like it?’ and all kinds of answers come up about emerging structures, belonging, and self-expression, so then the question was how can we create interactive works that embody those qualities within the interactivity itself.

What it left me with was not such a clear pathway into academia, because I hadn’t arrived at some clear and completed research finding, but what I had done was immersed myself so fundamentally in trying to answer this question, how can I make captivating interactive music experiences?f

What did you find?

On the question of interaction with technology, I think the most fundamental quality of technology is interaction, human-computer interaction. How is it affecting us? How are we affecting it? How does that ongoing relationship develop?

There is so much within those questions, and yet interactivity is often just tacked on to an existing artwork or introduced in a conventional way because that is how things are done. In fact, the way you do interactivity says a lot about who you are and how you see the world. How you design interaction is similar to how you make music, there are many ways, and each has a political interpretation that can be valuable in different contexts.

Who has influenced you in this respect?

The biggest influence on me at the point where I’d finished my PhD and commenced Cave of Sounds was the book Musicking by Christopher Small.

The shift in mindset goes from thinking that music is something being done by musicians on a stage and being received by everyone else around them, to being a collective act that everybody’s participating in together, and that if there weren’t an audience there to receive it the musician couldn’t be participating in the same music.

What I found informative is to take a relativist view on different musical cultures. Whether it is a rock concert, classical concert, folk session, or jazz jam, you can think of them as being different forms of this same thing, just with different parameters of where the agency is.

For instance, if you’re jamming with friends in a circle around a table there is space for improvisation and for everybody to create sound. This has an egalitarian nature to it. Whereas with an orchestra there is little scope for the musicians to choose what notes they play, but a huge scope for them to demonstrate technical virtuosity and skill, and I don’t think there’s anything wrong with that. I love orchestral music. I think there is beauty to the coordination and power. I can see how it could be abused politically, but it’s still a thing that I feel in my body when I experience it, and I want to be able to access that feeling.

What I’m most suspicious about are stadium-level concerts. The idolisation of one individual on a stage with everyone in the crowd going emotionally out of control. It is kind of this demagogue/mob relationship. People talk about these Trump rallies as if they’re like rock concerts, and it’s that kind of relationship that is abused politically.

I recently interviewed the composer and programmer Robert Thomas who envisions a future in which music behaves in a more responsive and indeterminate manner, more resemblant to software than the wax cylinder recording that helped define 20th-century music. In this scenario, fixed recording could become obsolete. Is this how you see the future?

I think the concept of the recorded song is here to stay. In the same way, I think the idea of the gig and concert is here to stay. There are other things being added on top and it may become less and less relevant as time goes on. Just in the way that buying singles has become less relevant even though we still listen to songs.

I think the most important thing is having a sense of personal connection and ownership. This comes back to agency, where I feel I’m expressing myself through the relationship with this music or belonging to a particular group or community. What I think a lot of musicians and people who make interactive music can get wrong is since they take such joy and pleasure in being creatively expressive, they think they can somehow give that joy to someone else without figuring out how to give them some kind of personal ownership of what they’re doing.

As musicians it’s tempting to think we can make a track and then create an interactive version, and that someone’s going to listen to that interactive version of my track and remix it live or change aspects of it, and have this personalised experience that it is going to be even better because they had creative agency over it.

I think there’s a problem with that because you’re asking people to do some of the creative work but without the sense of authorship or ownership. I may be wrong about this because in video games you definitely come as an audience and explore the game and develop skill and a personal style that gives you a really personal connection to it. But games and music are very different things. Games have measurable goals to progress through, and often with metrics. Music isn’t like that. Music is like an expanse of openness. There isn’t an aim to make the perfect music. You can’t say this music is 85% good.

How do you see the future?

I agree with Robert in some sense, but where I think we’re going to see the song decline in relevance has less to do with artists creating interactive versions of their work and more to do with people using AI to completely appropriate and remix existing musical works. When those tools become very quick and easy to use I think we will see the song transform into a meme space instead. I don’t see any way to avoid that. I think there will be resistance, but it is inevitable.

In the AI space, there are some artists who are seeing this coming and trying to make the most of it. So instead of trying to stop people from using AI to rip off their work, they’re trying to get a cut of it. Like say, okay you can use my voice but you’ll give me royalties. I’ve done all of this work to make this voice, it’s become like a kind of recognizable cultural asset and I know I’m going to lose control of it, but I want some royalties and to own the quality of this vocal timbre

Is there a risk in deskilling, or even populism, in a future where anyone can make profound changes to another person’s creative work? The original intention of copyright law was to protect artists’ work from falling out of their hands financially and aesthetically. The supposed democratisation of journalism has largely defunded and deskilled an important profession and created an economy for much less skilled influencers and provocateurs. Might not the same happen to music?

The question of democratisation is problematic. For instance, democracy is good, but there are consequences when you democratise the means of production, particularly in the arts where a big part of what we’re doing is essentially showing off. Once the means of production are democratised, then those who have invested in the skills previously needed lose that capacity to define themselves through them. Instead, everyone can do everything and for this short while, because we’re used to these things being scarce, it suddenly seems like we’ve all become richer. Then pretty soon, we find we’re all in a very crowded room trying to shout louder and louder. It’s like we were in a gig and we took away the stage and now we’re all expecting to have the same status that the musician on the stage had.

I can see your concerns with that, but when it comes to music transforming from being a produced thing to being very quickly made with AI tools by people who aren’t professional. If you’re a professional musician there will still be winners and losers, and those winners and losers will in part be those who are good at using the tools. There will be those with some kind of artistic vision. And there’ll be those who are good at social media and networking, and good at understanding how to make things go viral.

It’s not that different from how music is now. It takes more than musical talent to become a successful artist as a musician, you’ve got to build relationships with your fans, you have to do all of these other things which maybe you could get away with not doing so much in the past.

Let’s return to the original theme of what makes for a good installation. What advice would you give to someone in the same position now that you were in just over a decade ago when starting Cave of Sounds?

In 2012 when we started building Cave of Sounds Music Hackspace was a place for people to build things. This was fundamental for me. People there were making software and hardware and there was this sort of default attitude of ‘we built it, now we’re going to show somebody’. We’re going to get up in the front of the room and I’m going to talk to you about this thing, and maybe I’ll play some music on it.

I find the term installation problematic because it comes from this world of the art gallery and of having a space and doing something inside the space where it can’t necessarily just be reduced to a sculpture or something. Whereas, for me, it was just a useful word to describe a musical device where the audience is going to be actively interacting with it, rather than sitting down and watching a professional interact with it. So that shift from a musician on a stage to an audience participating in the work.

I don’t think it necessarily has to begin with a space. It needs a curiosity of interaction. Maybe I’m just projecting what I feel, but what I observed at Music Hackspace is people taking so much enjoyment in building things, and less time spent performing them. Some people really want to get up and perform as musicians. Some people really want to build stuff for the pleasure of building.

How do you get an installation out into the world?

How to get exhibited is still an ongoing mystery to me, but I will say that having past work that has succeeded means people are more likely to accept new work based on a diagram and description. Generally, having a video of a piece makes it much more likely for people to want to show it. The main place things are shown is in festivals, more than galleries or museums. Getting work into a festival is a question of practical logistics: How many people are going to experience it and how much space and resources does it demand? And then festivals tend to conform to bigger trends – sometimes a bit too much I think as then they end up all showing quite similar works. When we made Cave of Sounds, DIY hacker culture and its connection to grassroots activism was in the air. Today, the focus is the environment, decolonisation, and social justice. Tomorrow there will be other things.

Then, there’s a lot of graft, and a lot of that graft is much easier when you’re younger than when you’re older. I don’t think I could go through the Cave of Sounds process today like I did back then. I’m very happy I did it back then.

What specifically about the Cave of Sounds do you think made it work?

The first shocking success of Cave of Sounds is that when we built it we had like a team of eight, and I had a very small fee because I was doing this artist residency, but everyone else was a volunteer on that project or collaborating artists, but unpaid. And we worked together for eight months to bring it together.

A lot of people came to the first meeting but from the second meeting, the people who turned up from that point forward were the eight people making the work who stuck through to the end. I think there’s something remarkable about that. Something about the core idea of the work really resonated with those people, and I think we got really lucky with them. And there was a community that they were embedded in as well. But the fact that everyone might made it to the end, just like shows that there was something kind of magical in the nature of the work and the context of that combination of people.

So a work like Cave Sounds was possible because we had a lot of people who were very passionate, and we had a diversity of skills, but we also had like a bit of an institutional name behind us. We had a small budget as well, but the budget was very small, and most of the budget did not pay for the work. The budget covered some of the materials, really, but a significant amount of labour went into that piece, and it came from people working for passion.

Do you have a dream project or a desire for something you would like to do in the future?

For the past few years I’ve been exploring how to use AI to interpret the moving body so that I can create physical interaction without introducing any assumptions about what kind of movement the body can make. So if I’m making an instrument by mapping movement sensors to sound, I’m not thinking ‘OK this kind of hand movement should make that kind of sound’ but instead training an AI on many hours of sensor data where I’m just moving in my own natural way and asking it ‘What are the most significant movements here?’

I’m slightly obsessed with this process. It’s giving me a completely different feeling when I interact with the machine, like my actions are no longer mediated by the hand of an interaction designer. Of course, I’m still there as a designer, but it’s like I’m designing an open space for someone rather than boxes of tools. I think there’s something profoundly political about this shift, and I’m drawn to that because it reveals a way of applying AI to liberate people to be individually themselves, rather than using it to make existing systems even more efficient at being controlling and manipulative which seems to be the main AI risk I think we’re facing right now. I could go on more as well – moving from the symbolic to the embodied, from the rational to the intuitive. Computers before AI were like humans with only the left side of the brain. I think they make humans lose touch with their embodied nature. AI adds in the right side, and some of the most exciting shifts I think will be in how we interact with computers as much as what those computers can do autonomously.

So far, I’ve been exploring this with dancers, having them control sounds in real-time but still being able to dance as they dance rather than dancing like they’re trapped inside a land of invisible switches and trigger zones. And in my latest interactive installation Self Absorbed I’ve been using it to explore the latent space of other AI models, so people can morph through different images by moving their bodies. But the dream project is to expand this into a larger multi-person space, a combined virtual and physical realm that lets people influence their surroundings in all kinds of inexplicable ways by using the body. I want to make this and see how far people can feel a sense of connection with each other through full-body interfaces that are too complicated to understand rationally but are so rich and sensitive to the body that you can still find ways to express yourself.

This interview was originally published on Music Hackspace (part 1 and part 2). You can find more from Dom Aversano at Liner Notes.

Cave of Sounds is exhibiting at the Leonardo da Vinci Museum of Science and Technology in Milan until 22 October 2023.

Cave of Sounds was created by Tim Murray-Browne, Dom Aversano, Sus Garcia, Wallace Hobbes, Daniel Lopez, Tadeo Sendon, Panagiotis Tigas and Kacper Ziemianin with support from Music Hackspace, Sound and Music, Esmée Fairbairne Foundation, Arts Council England and British Council.

Hi everyone, hope you're enjoying August.

Some live events from me this week that you can hear online - a new performance of my real-time AI audio piece Agency of Chaos, Unmoved, and an insight into the inner sonic life of our old car.

I have two sound works playing at Radiophrenia next week. Radiophrenia is a two week festival celebrating sound work on the radio. It's already underway, broadcasting on 87.9 FM in Glasgow and online.

Driving around Bucharest in the Lupo on 17 November 2018

Played during the Buffer Zone
Tue 28 August 2023, 11.30am - 12pm

The Lupo was a special car. One day driving around Bucharest we discovered that if we set the radio to AM static, we could listen to all her electrical activities. The Lupo has now given up the ghost, but our car in Glasgow holds our final analog radio. This broadcast is in memory of the Lupo, that she may sing again in the cars of Glasgow.

Collaboration with Adriana Minu. Never played this to anyone before.

Subscribe now

Agency of Chaos, Unmoved

Live performance
Sat 2 Sep, 5pm - 5.30pm

Performed live with real-time AI-rendered audio, Agency of Chaos, Unmoved is a journey on the nature of consciousness. A quotation from Alan Watts is ingested by an AI trained on his work, and gradually transformed into a new form. The struggle of the human combines with the uncanny valley of the AI. In the errors and distortions I hear the struggle of a living being: effort, intention, agency.

This will be a new performance of my recent new work. More details here.

Tim
London, 26 Aug 2023

After listening to the recording, my friend Dom Aversano posed a question.

With the use of the 'AI', I have a question, why? I want to keep that question open so won't elaborate, but I just want to understand better what it signifies in the compositional process.

Dom is sceptical of AI hype - he even uses scarequotes around ‘AI’. I’m sympathetic. Hype certainly seems to inspire a lot of meaningless art. This can make it harder to find space for the slow-burning artistic research I’ve been undertaking. So it’s prompted me to contemplate what drew me here.

The simple answer is that the possibilities AI brings into the creative process are too exciting to ignore. But Agency of Chaos, Unmoved doesn’t just use AI. It’s all about AI.

The type of AI being used by me and pretty much everyone else right now is a branch of AI called Machine Learning – a clunky term that hides quite how exciting a shift it represents. Instead of explicitly coding what a program does, we allow its behaviour to emerge through training it on some kind of data.

Machine Learning systems work unreasonably well (literally). But they are ‘black box’ systems in that their process doesn’t break down into understandable chunks. A human cannot fully comprehend how they work. This has irked many – including computer scientists – because it sidelines many of the beautiful theories we have formed about our world. The creators of ChatGPT didn’t even need to know the difference between a noun and a verb to create a machine that can speak English. I, too, enjoy intellectual theories about the world but I think the belief that everything can be reduced into comprehensible chunks is a grand delusion of the modern age. This makes Machine Learning’s black box nature more exciting than irksome to me.

So, officially, Machine Learning is the pragmatic branch of AI behind all these tricks. But on an emotional level, the terms Machine Learning and AI capture two concurrent yet distinct experiences I have with the technology. Machine Learning speaks of good engineering that disappears into the background, unnoticed until it fails to work. But AI is an alchemical technology very much in the foreground. It feels like a mystical force with capabilities we can’t quite be sure of, a kind of magic. My interactions with AI sometimes seem to occupy the part of my mind that deals with people, animals and spirits rather than tools, objects and machines.

Subscribe now

AI systems like ChatGPT and Midjourney are producing results which feel human. They display a degree of creativity that is a generational leap from what we had a few years ago. They may borrow shamelessly from whatever data they’re trained on but the way they borrow imparts an uncanny feeling of a human-like understanding.

Then there is the potential for emergent agency. Language models like ChatGPT have the capability to construct goals, form plans and do things on the internet - in principle. I’ve not yet seen a successful example myself, and so there is a sense of mystery as to what abilities still lie latent. This is where I feel the sci-fi vision of AI running through the 20th century. I need more than the shallow certainties of hype and doom to make sense of them.

Agency of Chaos, Unmoved is my journey into this collective unconscious. How does it feel to hear that voice, in its struggle to emerge, questioning what our experiences of consciousness tells us of the universe?

In practice, I found the creation of the work a similar process to composing generative music. In a generative work, the composer creates a system that autonomously generates a set of musical outcomes. For example, consider the everyday set of windchimes. There may be a lot of craftsmanship put into each chime but the final music is undetermined, emerging only when the human-designed system meets the randomness of the wind. Listening to windchimes, I can hear the human hand in the exact pitch and timbre of each chime, and I can hear nature in the energy of the wind, but I can’t predict the sequence or rhythm. Control is relinquished to an unpredictable force.

The black box AI model adds a new unpredictable force. I can’t fully understand it but I can poke and prod and get a feel for it. I craft a set of sounds to train it on, I manipulate parts of it, and I see what happens. I can’t predict the sounds that emerge from it during the performance but I can still hear within them my preparations, my actions on stage, and of course the characteristic artefacts of the AI model itself.

Tim
Buftea, 24 July 2023

As I come to the end of my time in Montreal, I invited those I’d met here to experience SELF ABSORBED on the massive video wall at Concordia University.

My friend Zeph Thibodeau's description of the experience is my favourite of any: “It feels like how I imagine the experience of the Guild Navigators in Dune plotting a path through hyperspace”.

Returning

My sense of time has been contracting these past years. I’d blame the pandemic, but I suspect it began before that. Somehow, it’s two years since I was bunkered away at the beautiful idyll of Cove Park in West Scotland. Here, I first connected the body analysis I'd trained for Sonified Body project to the new image-generating AI, StyleGAN, that I’d just spent 6 weeks training on my lifetime vault of photos. I had a hunch this would be cool but I was instantly hooked on the result. It felt like I was inside the AI’s mind. And that mind was all about me.

Since then, I’ve lugged my heavy PC up and down the UK. I’ve watched as maybe 100 people play with it. The title has changed. The story I tell to explains the work has changed. I’ve added new things: particle-based rendering, depth estimation, generative AI audio for the soundtrack. But all of these I’ve now taken out because I felt they detracted rather than enhanced that core experience.

The only significant difference now is it’s on a massive screen in a dark room. Alongside, many technical tweaks you probably wouldn’t notice but subtly improve the experience: improving the framerate, smoothing out the noisy jitters in the body tracking.

Dead ends are not bad. A lack of dead ends indicates either a failure to take risks or a failure to confront things that aren't working. But it feels odd to arrive two years later back at basically where I started.

The value of hesitation

I was chatting over a coffee with Lynn Hughes, a prof here at Concordia University. I mentioned these explorations with StyleGAN, including Diffeomorphism, my new series of slowly undulating animations.

I’m rendering these from that same AI model with a technique I sketched out in early 2022. Given the pace of this scene, it feels like a mistake to have not put them out a year or two ago if I could have done so.

Lynn spoke of the value of hesitation, of trusting the instinct to pause, of spending time with a work in spite of today's relentless churn of non-stop output. I appreciated that.

During the Pause, the way I see these images has changed. Back then I remember feeling hesitant over the visual aesthetic because I didn’t have a clear grasp on what parts of it were emerging from me (my training data and my selection of outputs) and which were visual tropes inherent to anyone using this StyleGAN model.

It may sound strange that I found myself having created an artwork by myself, and yet unable to distinguish my own contribution from that which lay latent in my tools and materials. But I’m not the kind of person who holds a clear vision of what they are creating. In fact, I have aphantasia and can't even form images in my imagination, much as I'm moved by what I see with my eyes. When working with visual media, be it drawing, photography or coding, I literally have no vision of where I’m going, though I may be able to describe what it will feel like when I get there. The image emerges through the process.

There is freedom in this. And there are traps.

I’ve written before on the tendancy of technology to homogenise human expression by standardising one’s range of actions. Technology that is new, complex and opaque, like AI, makes it extra easy to fool myself into believing that its tricks are my own self-expression. Perhaps, as my creative process is so rooted in exploration rather than navigation, I feel this more acutely.

It’s by seeing what others do with similar tools that I’m able to understand what is me and what is the tool. Two years on, I see myself more clearly in those early StyleGAN explorations. I’m clearer on what I care about. It’s true, many images emerging from the model look like tropes. But others look fresh, beautiful and profound. I’m not sure I saw this distinction so clearly back then. I didn’t realise where to dig, and how deep to go.

The dance between intuition and vision

In a work like SELF ABSORBED, the process and concepts were clear from the outset. They are embodied in the technologies and algorithms I develop. But the outcome: what it looks like, sounds like, feels like, whether it’s an interactive installation or a video – over these, I want to release preconceptions. For these, I find myself talking about trying to uncover what the work wants rather than what I want.

This type of creation, where material is encountered rather than invented, is familiar from musical improvisation, and from observing dancers in the studio. It’s often the kind of experience I’m aiming for in an interactive work. The gestural hyperspace navigation of SELF ABSORBED is an extreme example. It gives you detailed, intuitive and consistent control. Yet it defies any logic that might let you predict beyond what you’ve already encountered. It is pure exploration. I love it for that.

But more often, there is a dance between explorer and navigator. While moving intuitively through possibilities, I encounter a glimpse of something and navigate towards it.

Last year I was coding a simulation of particles that repel each other yet are drawn to a common attractor – similar to how flies swarm. I rendered the trail each left as it moved. In it, I saw something reminiscent of a flower growing, and I iteratively tweaked the code to bring out that image. This became my series a Heaven in a Wildflower.

and you are the easiest person to fool

“The first principle is not to fool yourself – and you are the easiest person to fool.” – Richard Feinmann

Over these two years, I’ve pitched variations of this work to many festivals. Only the weird world of Electromagnetic Field festival were keen. (And it was great – one of the most exciting festivals I’ve been to.)

So all these ponderings may be rationalisations for the disappointing lack of traction I’ve had in finding opportunities to show this work.

I can only speculate why. Maybe I don’t have as much energy to hustle work into places as I used to. The world is already saturated with AI art and art about AI. Festivals are not keen on interactive works created for one person to use at once. Even Cave of Sounds got rejected from a festival once on the sole grounds that they can’t squeeze enough punters through it. And that looks like this:

A friend explained it to me recently:

More people, quicker experiences, more $/m²/day. It’s obvious when someone writes it out, but I’d actually never thought of it that way before.

Maybe I just arrived into this tech-art space at the right time 11 years ago. Today it feels like a giant burn out to me. I rode a great wave, but now I’m at paddling at the shore, the sea is full of desparate people, and there’s barely any surf anyway. My instincts have been telling me a while but the deeper you are in something the harder it is to confront when it’s time for something different. The problem with being an explorer is you keep thinking there’s probably a break just round the corner, even when actually they’re all really far away. It’s time to catapult out of this deteriorating local optimum and meander through the hyperspace of future possibility. How many metaphors can I mix to say one thing. :)

Tim
Montreal, 16 June 2023

Last week was my first live performance with AI audio. I’m calling the piece Agency of Chaos, Unmoved.

Live performance at Cafe SAT, Montreal, 23 May 2023. (Download as mp3)

If, like me, you are a purist who likes to experience work with an untainted mind, then you might like to listen to the recording to form your own impression before reading my interpretation here.

At the core of this piece are four versions of the audio-generating AI model RAVE. RAVE is an auto-encoder: it ingests sound, encoding it into its own internal language. It then decodes its own language back into sound. To train it, I give it hours of audio, and it optimises the encoding and decoding process to work well for that audio. (Think of a person training to listen, remember and vocalise a sound. They learn what to listen for, what details to remember and how to recreate a sound from those details.)

I trained four versions of the RAVE model. One on a collection of lectures by Alan Watts. Another on every sound I've recorded, including the few seconds attached to each Live Photo taken on my iPhone. A third on all the music and sound art I've ever made. The fourth version is trained on a set of recordings of Adriana Minu, vocal performer and my wife. These were the first recordings of her experimental vocal practice after 10 years of not singing. In her words: “there is struggle and vulnerability in my early voice as I feel my way through new vocal territory”.

I played with combining these models together in new ways. If I feed the sound of Alan Watts through the model trained on his own voice, I get a slightly distorted version out. The distortion has an uncanny nature to my ears, less like analogue noise or digital glitch, and more like a skilful robotic imitator slipping up here and there. Next, I tried running the models simultaneously, and feeding the internal language encoded by one model into the decoder of a different model. The sound departs further. The dynamics and rhythm remain. The timbre is reminiscent but not quite there.

There was a magic moment where focus of the piece became clear. I was gradually degrading the quality of the Alan Watts model by modifying its internal representation between encoder and decoder. When I fed these encodings from the Alan Watts model directly into the decoder of Adriana’s model, something else emerged. These sounds were eerie. The struggle of the human combines with the uncanny valley of the AI. In the errors and distortions I hear the struggle of a living being: effort, intention, agency.

There's an ambiguity in where that struggle is rooted. Is it Adriana's struggle appropriated by the AI? Is it the AI models trying to get through as I reroute their internals into each other, Frankenstein style?

At the gig, I played using some MIDI controllers I borrowed from a friend. Having spent the past few years building AI-powered embodied gestural interfaces, it felt odd to perform with knobs and buttons. But this piece is a step on a bigger journey. Each new component brings its own character, whether that’s of sound, image, physicality or how those three connect. It makes sense to me to get a feel for that character by working with the models individually first.

Afterwards, someone told me she found it creepy, adding “but in a good way”. Another said she was sat next to her friend and said to her at the end that she felt empty; her friend said she felt completely filled. A third person said it was unlike anything she’d heard before. I found this last comment most validating, which is interesting because I complain about the cult of the new in the tech-art scene where novelty gets valued above depth and feeling. Performing with AI audio definitely runs the risk of being little more than a gimmick so I think the validation is in uncovering something that sounds new - at least to me, and someone in the audience.

Tim
Montreal, 2 June 2023

On Tuesday I’m giving a sound performance at the Cafe SAT in Montreal to close the (de)Stabilizing Diffusions exhibition. Usually when I’m working with performance, either I’m playing with others or creating an interactive system and directing others to perform with it. This will be my first solo sound performance in seven years. I’ve been a tad nervous.

My piece will be concluding an exhibition of artists exploring AI. Appropriately enough, I’m performing with live AI audio generation. So there is still some dilution of responsibility. There will be glitch. And fun.

Agency of chaos, unmoved

An open-ended performance with real-time AI-generated audio. The computer music pioneer Joel Chadabe described performing with a non-deterministic system as like sailing a boat through stormy seas. A storm has its own agency of chaos, unmoved by whatever intentions I may have in harnessing its forces. In moments of desperation, it’s tempting to think that the storm is aware of our plight as it ushers or torments us.

I’m using an AI model called RAVE, created by Antoine Caillon. RAVE does live style transfer. Train it on a violin, and then whatever sound you feed in, it will make it sound like someone doing their best to play that sound on the violin. It’s popular because it not only sounds lifelike but does so fast enough to be used live. However, its style transfer can be so obediently responsive that it feels like a regular audio effect. I don’t get that feeling that it has a mind of its own. I’m performing alone once more.

But RAVE has an alternative generative mode where it endlessly invents new sounds, trying to imitate what it remembers of the sounds I trained it on. In this mode, I get just one dial of control to influence how random it gets. This gives me the opposite problem: I get next to no control, and my role is little more than turning the machine on.

These two extremes capture the ends of a spectrum of how much agency we have when we work with AI. What I want is something in the middle: to release enough control to be surprised but not so much as to be reduced to a cog in someone (or something) else’s journey. My own little AI alignment problem.

Subscribe now

The conundrum is familiar to me. It’s pretty much to what my PhD was about: the balance in creative control between a system’s designer and its participant. But with an unsupervised AI model like RAVE, there’s not the same duality between designer and user. It’s like a flashback to a nostalgic 90s when technology was frustrating because nobody knew how to make it not frustrating, rather than because corporations discovered that frustrating us can be rather profitable. Here, the balance is between me and a chaotic system that’s been tweaked a three million times by an algorithm trying to make it a little less chaotic (what we call ‘training’).

Last week I called up Maurice, the curator, and said the performance wasn’t going to work. Maurice has also been working with RAVE. He heard me out, with the great empathy of someone who’s also been working with RAVE, and persuaded me to continue. The exhibition is about how artists are approaching AI. To expose the challenges of finding a voice with an all-or-nothing AI model is a message that will resonate.

Experimentalism is perhaps another way to free oneself from the self-defeating mindset of trying to please an audience.

So I’ve persisted, looking in the cracks and glitches. I’ve trained versions of the model on different collections of sounds and have been blending them together. It takes me about 10 days to train a model so each one is a gamble. But I do now have something I’m happy with, a middle ground between the blank canvas and a CD player with a singular play button. I’m excited to share it.

If you’re in Montreal, please join us at Cafe SAT on Tuesday 23 May. Doors at 6pm, Artist talks at 7pm, I’m performing at 8pm. It’s free but RSVP required.

Tim
Montreal, 19 May 2023

PS - I had hoped to rope in Adriana but she’s heading to Glasgow for her citizenship ceremony (somewhat awkwardly timed as we’re returning in July). For those in Glasgow, on Thursday she’s performing on Thursday as part of GLEAM festival: “An alternative citizenship ceremony in which the contradictions and imprecisions of national identity can be teased out through sound. Expect no pomp, full circumstance.” I’m gutted to miss it, but so it goes.

About five years ago I received a tape in the post from my late godmother. It was a recording of a radio broadcast of choral evensong at Wakefield Cathedral on 12 September 1995. I'm one of the singers, aged nine.

There are probably a few other recordings around - in the BBC archives, on other tapes recorded off the radio. The memory of that moment is imprinted into materials and still echoing through time. Each medium, like my tape, makes imperfect copies, combining the event with its own material reality.

Subscribe now

Between the age of 6 and 12 I sang as a treble chorister. Each week of the school term, we’d rehearse for about 5 hours and sing two services. It’s a union of education, work and cultural ritual that is hard to find in the secular world. The harmonies of sacred choral music are embedded deep within me.

My own memory is shaped very differently from the tape memory. The sounds on the tape resonate through me associatively. I’m back into the cathedral, smelling the incense, feeling the weight of the strange costume - cassock, surplus and ruff - and seeing the ever-animated face of the choirmaster conducting us.

The memory of the tape and of my mind both have their own topology, by which I mean they have a kind of shape in which some things are near to others. The tape is just organised by time. I can fast-forward and rewind. My mind is more of a web of experiences. I can barely remember it on demand, and yet an external cue like a sound or smell can dive me right into the middle of it.

The memories held within the AI model are different again. While ChatGPT is perhaps closer to the associative web of human memory, the StyleGAN model I’m using here has a very different feel. It places memories that look similar near each other without any notion of time or meaning. They are still my memories, because it’s trained exclusively on my own library of about 25,000 photos I’ve taken over the past 20 years.

Each of the three forms of memory are a kind of echo. Each reveal a different aspect of reality. Perhaps the tape is materialism, my memory is structured through consciousness, and StyleGAN captures a different kind of immaterial structure of the universe.

I’m reminded of that line from Alan Watts, “You are an aperture through which the universe is looking at and exploring itself.”

The title of this work came from the tape. You can hear it in the soundtrack. For me, it’s also fitting to the fractal shape of the universe that the visuals suggest.

Tim
Montreal, 5 May 2023

Do you remember four years ago when we first saw those AI generated photos of people’s faces and were told “this person does not exist”?

I remember the disorientation of that moment. It seemed incredible that an AI had acquired such a deep knowledge of the complexities of the human face, as well as the capability to render with photographic realism. When I was only confronted with the flawless images, it was easy to jump to that conclusion.

Those images were created with an AI model called StyleGAN. It’s the same family of model I’ve been using in a lot of my work, including the video I shared last month “Cosmic Insignificance Therapy”.

These days, I'm more familiar with its mistakes.

StyleGAN’s mistakes aren’t simple factual errors like colouring the pupils green or drawing the eyes too big. It seems to completely lose grasp of what physical reality looks like. Its mistakes fall into the uncanny valley, that disturbing gulf between the cute and the lifelike where zombies and ghosts belong.

The weirdness of deconstructed intelligence

the weird is a particular kind of perturbation. It involves a sensation of wrongness: a weird entity or object is so strange that it makes us feel that it should not exist, or at least it should not exist here. Yet if the entity or object is here, then the categories which we have up until now used to make sense of the world cannot be valid.

— Mark Fisher, The Weird and the Eerie

I think StyleGAN’s uncanny mistakes can be disturbing because they violate my gut assumptions about how intelligence manifests in this world. I’m used to seeing human-like intellectual capabilities together as a bundle, what I consider human intelligence. If a human can draw photorealistic faces, I might assume they have mastered many other intellectual abilities, like a deep sensitivity to human physiology and how it exists in physical reality.

But the sight of that woman’s face slowly degenerating through deformity into smudges reveals a thinking process lacking these abilities. To feel the presence of some part of intelligence without the rest is weird. It disrupts my assumptions of what I can expect from reality, shaking me into a world where something approximating human-like intelligence can arise from unrecognisable ingredients. It’s scary like a zombie who has enough agency to animate a human corpse into violence but lacks the capacity for compassion, reason or pain that might stop it.

The danger of seeing intelligence as a spectrum

StyleGAN takes one piece of the intelligence bundle, isolates it and amplifies it. Those initial uncanny images forcefully unbundled my conception of intelligence. This feels like an important experience.

StyleGAN is now already over four years old. But I’ve spoken to a few people who I think might be experiencing a similar uncanny reckoning with ChatGPT. An initial overwhelm of its impressive abilities is followed by disappointment on finding its limitations. In some cases this is met with a sense of vindictive relief. Maybe this AI is not so intelligent after all. Thank God for that.

There are indeed good reasons to talk down the “intelligence” of AI. For example, last month the Romanian Prime Minister Nicolae Ciuca unveiled an “AI advisor” called Ion which will eventually inform the government of the thoughts of the population so it can make better decisions. The story was reported in the Guardian with zero critical analysis after a press launch featuring Ciuca speaking to a mirror that was pretending to be an AI.

To give Ion its human name, and call it an AI, implies some kind of intellectual authority. It leans into the intuition that intelligence is a singular trait, and so one intellectual capability implies all the others. But Ion is, at best, a software package for statistical analysis, and possibly little more than a PR stunt. And call me a cynic, but in my experience statistics tend to be used by those in power to justify their decisions rather than inform them. I can’t imagine a more poignant image of a government AI advisor than the PM speaking into a mirror.

But let’s take care. Even if ChatGPT is little more than turbocharged statistical analysis of the web, to call it “unintelligent” is as problematic as calling it “intelligent”. It can likely do things - intellectual things - beyond what we can even dream up right now. It may have just a slice of the intellectual capabilities of a human, but it scales to a capacity that gives qualitatively different results.

For example, ChatGPT can code, test its code to see if it works, modify that code and then iteratively build up a piece of software much like a human coder. But, unlike humans, we can spawn thousands of instances of ChatGPT to work in parallel for a relatively tiny cost. Someone can (so probably will) give it an internet connection, a list of their enemies and ask it to discover new hacking techniques to dig up dirt on them all.

It’s difficult to anticipate how far it would get. Ten years ago, many were expecting self-driving cars to be the dominant mode of transport by now. Sometimes, you need far more parts of the bundle we call intelligence than it seems at first. This may also be true of ChatGPT coding, but its limited ability for original critical analysis gives much less of a meaningful signal than it would for a human.

Intelligence is far more complex than a bundle of intellectual capabilities, but I’m finding the bundle a more useful analogy than the ascending ladder of abilities with humans at the top. Some capabilities which have previously only been present in humans are now in machines. Others are coming into existence that we’ve never seen before.

Tim
Montreal, 21 April 2023

This is the ART∩CODE newsletter from Tim Murray-Browne.

I’ve just moved us from Mailchimp to Substack. I think Substack emails are less likely to end up in Spam, so if you’re not expecting this, then maybe I’ve been emailing your spam folder for a few years. If you don’t want to receive it, there should be an unsubscribe link below. If you’d like to know how you joined this list, email me and I’ll let you know.

“To lose yourself: a voluptuous surrender, lost in your arms, lost to the world, utterly immersed in what is present so that its surroundings fade away. ... to be lost is to be fully present, and to be fully present is to be capable of being in uncertainty and mystery.”

— Rebecca Solnit, A Field Guide to Getting Lost

I

A couple of years ago I started trying to drive without using a satnav (or more precisely, Google Maps on my phone).

I’ve always been suspicion of satnavs. All they do is bark out orders. It seems harmless enough, but I suspect somewhere in my subconscious it’s reinforcing the narrative: ‘better do what this machine tells me to’. I can’t help but feel a step towards the subjugation of humans by machines. Perhaps people felt this way when they introduced traffic lights. Grumble grumble.

But my issue with the satnav was more practical. I noticed I wasn’t learning my way around when driving in the same way I did when walking or cycling.

I remember one day driving with Adriana from central Edinburgh back home to Glasgow using only street signs. An hour later, when we should’ve been arriving home, we turned the corner to be confronted by the sea - the North Sea! That sits between Edinburgh and Denmark, not Edinburgh and Glasgow. But we learnt a lot more about the geography of Edinburgh as a result.

Another time, I added an hour to a trip to London by taking the M77 towards the West coast instead of the M74 south.

Both trips I’d made many times with the satnav. While the satnav held my hand, I was apparently learning very little. It’s basic pedagogy that I’m going to learn better if I have a go at a problem myself before hearing the solution. But it shocks me quite how little I absorb from the steady stream of solutions from the satnav. I won’t even remember the exact route it’s led me down.

This concerns me not just because of the dependency it breeds, but because navigating space is much more than a means to get around. It’s a fundamental metaphor to how we think. It unites the discreteness of places with the continuity of space. Map-reading confronts me with the gulf between the reality I inhabit and a model that attempts to describe it. If I no longer exercise my mind’s ability to navigate the land, does this undermine how I think more generally?

II

Large Language Models like ChatGPT are still ascending the hype curve. In the coding world, it’s felt like there’s talk of little else the past six months. This might be natural given our profession. But I think we're also experiencing more immediately how AI is changing the nature of computer work.

Whereas an AI tool can now displace an entire job for a photographer or illustrator, I’m unaware of this happening so far with a serious coding project. And whereas, say, a written article requires the whole text to remain coherent to an author’s voice, software development has long since evolved into a modular affair well suited for collaboration. When I ask ChatGPT (and its cousin Copilot) for a small snippet of code, I can just copy paste it into my project like a lego brick.

ChatGPT does not know my context. It doesn’t know what I’m trying to achieve here. I have to think of what’s relevant and tell it. And its code is not always correct. I have to verify and test the answers it provides. But it’s more than good enough to dramatically increase the pace at which I can make things.

What is this doing to my mind? I’m getting more done, but am I getting stupider? If OpenAI kicks me off ChatGPT, would it just leave me slower? Or will I be left in the middle of nowhere, clueless about how to continue this work myself?

III

To illustrate this, I’ll share a brief snapshot of my experience coding with ChatGPT. Don’t worry if the technical details feel overwhelming or go over your head. That experience is actually part of what I’m trying to illustrate here.

In the Cosmic Insignificance Therapy video from last week, I slowly zoom out over one minute on an undulating image, rendered by an AI model. I rendered the video with code I wrote in the Python language. To do that zoom, I had to express it mathematically.

I vaguely remembered that to animate a zoom, I needed a logarithmic interpolation function. I couldn’t be bothered to work it out myself so I asked ChatGPT. It spat out the answer:

min * (max/min)^t

I stared at this a moment. I hadn’t thought it could be expressed so simply. I was expecting at least a few logarithms in there. I had a flashback to sitting in my A-Level maths class 20 years ago learning geometric and arithmetic series. (See what I mean about the relationship between knowledge and a sense of place?) In my head, I substituted in t==0 (we get min) and t==1 (we get max). All looks OK. I’ll remember this. ChatGPT just taught me something.

This is an optimistic scenario of AI collaboration. I’m still using my brain and ChatGPT is helping me do so.

Later on, I wanted to make a video where we zoom in and out to different places, then end up back where we started so that I can seamlessly loop the video. Again, I knew the name of the tool I needed: a closed cubic spline. This is a mathematically defined line that goes through whatever points you want it to, yet stays smooth and elegantly curvy.

ChatGPT complied, but this time the answer was more involved. Brace yourself.

def log_interp_spline(points, t):
    n_points = len(points)
    if n_points < 2:
        raise ValueError("At least 2 points are required for interpolation")

    points = np.concatenate(([points[-1]], points, [points[0]]))
    t = t * (n_points - 1) + 1
    t_floor = int(np.floor(t))
    t_frac = t - t_floor
    
    a = points[t_floor - 1]
    b = points[t_floor]
    c = points[t_floor + 1]
    d = points[t_floor + 2]

    p0 = (c - a) / 2
    p1 = (d - b) / 2

    coeff0 = 2 * b + p0 - 2 * c + p1
    coeff1 = -3 * b - 2 * p0 + 3 * c - p1
    coeff2 = p0
    coeff3 = b

    result = coeff0 * (t_frac ** 3) + coeff1 * (t_frac ** 2) + coeff2 * t_frac + coeff3
    return result

How long did you spend with that? If you’re me last week, not very long at all. I was not whisked back to my undergraduate lecture on computer animation where I was probably taught this. I just copy-pasted it and rendered a quick video to confirm it worked.

Or did I? As I look at it now, I’m not convinced it’s in logarithmic space. So did I have to fix that? This was just last week, and I can’t remember. This is alarming. It seems I was coding with ChatGPT like I drive with satnav. I feel the tool dependency. My mind atrophies.

IV

In a sense, ChatGPT is the opposite of a satnav. I’m the one plotting our route, barking out orders. Delegation is not a crime. It’s a skill, a subtle one at that, and one I often see myself fall short on. I’d get little done if I followed each intellectual tangent on my way.

But, as an artist, I make stuff that nobody asked for. In doing this, I hope to reveal something in the process that I could not have just thought up without getting my hands dirty. Each act of delegation creates distance to some part of the medium. It transforms what I might uncover in my intuitions, reducing the terrain but hopefully giving me time to go deeper.

When delegating, the question is not just what will maximise my output? but also what parts of this process do I need to experience myself?

We all need to choose a level between director and craftsman, but it’s a question I hope to keep forefront in my mind as I continue.

Tim
Montreal, 7 April 2023