Generative Systems: Learning to Ask and Listen
There comes a moment, when working with generative systems, in which the question ceases to be at the centre of the process. Not because there is no longer any need to question the system, but because something shifts: attention is no longer focused solely on what the system produces, nor on the way it responds, nor on the precision with which it carries out a task. Instead, another form of relationship begins to emerge, subtler and more unstable. One stops, if only for an instant, asking, and begins to listen.
This is not a technical transition. It does not depend on the model, the dataset, the latency or the sophistication of the interface (however comforting it may be to pretend that everything hinges on measurable parameters, so that we may continue to sleep soundly within our safe domestic positivism). It depends, rather, on the way one enters the process and on the kind of attention one is willing to exercise.
I have spent years working with artificial intelligence as part of an artistic practice: first through projects, then through my doctoral thesis, and later through research. I have observed systems produce outputs I had not anticipated, sonic trajectories that seemed to deviate from the initial form of the project, classifications that called into question the very categories through which I had tried to organise the material. Each time, a question remained suspended, difficult to dismiss through the orderly vocabulary of human–machine interaction: was I guiding the process, or was I merely recognising, after the fact, something that had already happened without me?
And, above all: who was listening to whom?
Beyond the Tool: Building a Relationship with Generative Systems
To describe AI as a tool is correct, but only up to a point. It is correct because, of course, a generative system can be used to do something. It produces, transforms, organises, suggests, accelerates. Yet it is reductive to the extent that it encourages us to imagine that the most interesting variable is the tool itself, as though the question were simply one of learning to press the right keys, formulate the best prompt and extract the most efficient output. A reassuring perspective, and for that very reason already a suspect one.
Anyone who has studied the piano knows perfectly well that the instrument does not play by itself. But they also know that technique comes after something else. First comes listening. The ability, for instance, to distinguish a musical phrase that breathes from one that merely occupies time. That ability is not developed by increasing the speed of one’s fingers; it is formed through an education of perception that no method, no exercise and no manual can ever fully replace.
With artificial intelligence systems, something similar occurs, though with a further complication. The system is not a neutral object waiting to be activated. It carries with it a history, a statistical structure, a distributed memory, a set of sedimented relations that are never simply passive. It does not respond in the same way to different interlocutors, not because it possesses some mysterious interiority (a hypothesis we are happy to leave to the enthusiasts of TED-conference animism), but because the process that is set in motion is relational. It changes according to the quality of the question, the context evoked, the clarity of intention, the ability to inhabit ambiguity without closing it down too soon.
The most interesting results, in fact, emerge when those working with these systems are not merely seeking confirmation; when they are not using AI to obtain a rapid, plausible version of what they already had in mind, but instead accept that the encounter with the non-human may shift the centre of the process. Not in the naïve sense of delegating some autonomous creativity to the machine, but in the more radical sense of recognising that the material one works with is not inert. It responds, resists, deviates, returns something.
The Qualities to Cultivate When Working with Generative Systems
Certain qualities seem to recur in those who are able to turn this encounter into something more than a matter of efficiency.
The first is a curiosity that is not confined to a single field. It is the willingness to consider relevant what, at first sight, may not appear to be so. To bring into the process lateral references, cross-pollinations, deviations, materials drawn from different fields. Not as interdisciplinary ornament, but as a method of listening. As a way of recognising connections before they become evident.
The second is a form of presence that we might call contextual awareness. Knowing one’s own disciplinary field is not enough: one must be able to perceive the wider cultural, social, technological and epistemic movements that make what is being produced meaningful, or conversely problematic. Every output exists within a context. Every generated image, sound or text carries with it a network of conditions, implications and genealogies. To pretend that the result is merely the product of a well-formulated request may be comfortable, but it is rarely interesting.
The third quality is a refusal to settle for the first plausible solution. Generative systems are remarkably good at producing coherence. And it is precisely there that they become dangerous: coherence reassures, making the result recognisable and presentable. For that very reason, the first convincing output is often the point at which the process risks stopping before it has even begun. Moving beyond that surface requires a certain obstinacy, but above all it requires the habit of asking uncomfortable questions of one’s own processes before asking them of the system.
The fourth, perhaps the most difficult to describe, is imagination, understood not as a generic capacity for invention, but as a point of view brought into the process. This does not necessarily mean knowing exactly where one wants to arrive. Indeed, the work often becomes interesting precisely when the initial destination begins to shift. But a direction is needed, however provisional. It requires a sensibility capable of recognising when a deviation is merely noise and when, instead, it opens up a possibility more precise than the one imagined at the outset.
Artistic Intelligence: Beyond AI Literacy
These are not digital skills. They do not belong to the rhetoric of AI literacy understood as a new compulsory form of literacy required to survive yet another technological transition, since humanity is evidently rather fond of turning every change into an introductory course with a certificate at the end. We call it Artistic Intelligence: a set of approaches that artists, designers and creative practitioners have always cultivated in their relationship with their own materials and processes. Artificial intelligence is a complex, reactive, layered and often unpredictable material. And, like all materials worthy of attention, it returns something in proportion to what one is able to bring into the relationship.
Perhaps it is here that the discourse on AI becomes less interesting as a discourse on technology, and more interesting as a discourse on the formation of the relationship. It is not simply a matter of learning to use a system better; it is a matter of understanding what the system reveals about the way we think, select, desire and recognise value.
Artur Schnabel, as early as 1958, observed that he did not treat the notes better than many other pianists, but that his art lay in the pauses between them. It is a remark that immediately shifts the problem: producing correct notes is not enough, just as producing coherent answers is not enough. What matters is what happens between one note and the next, between a gesture and its consequence, between the output and the capacity to recognise its necessity or insufficiency.
Perhaps it is worth asking under what conditions something similar might occur with generative systems: when we are able to stop thinking only about the correctness of the prompt or the output, and begin instead to listen to what the process is transforming in us, while we believe, with admirable human presumption, that we are the ones transforming it.