Volume 66, No. 1

We have been here before, though never quite like this.

In 1755, Rousseau warned that music was becoming “a science of signs” rather than an “art of expression,” lamenting how notation displaced lived practice and moved authority from bodies to texts. More than a century later, in 1906, John Philip Sousa feared that the phonograph would turn music from a shared event into a portable object—something owned, replayed, and consumed rather than enacted together. In 2014, Ravi Shankar observed that Western conservatories had trained musicians to “read the lines” while forgetting how to “read between them,” replacing the intimacy of guru–shishya transmission with institutional credentialing.

These moments are not merely historical curiosities. They are recurring symptoms of a deeper tension between mediation and presence, convenience and craft, abstraction and embodiment. Each technological shift provoked anxiety that something essential would be lost, and each, in turn, forced musicians and educators to articulate more clearly what they wished to conserve.

Today, the disruption is both faster and wider. We are moving from an era of Pattern Recognition—classification, tagging, transcription—to one defined by Generative Models that compose, orchestrate, and notate. AI is no longer simply a tool; it is becoming an ambient condition, a background infrastructure of thought, creativity, and communication. In such a world, intelligence risks becoming something we rent from distant systems rather than cultivate through practice, attention, and community.

The crucial question is therefore not whether AI will shape music—it already does—but what must remain irreducibly human within this new ecology. I frame this tension as the difference between Infinite Generation and Human Arrival: the difference between producing without end and knowing when, how, and why to stop.

Generative AI vividly embodies David Deutsch’s claim that “knowledge-making has no natural end.” Models can generate endlessly, remix tirelessly, and vary without fatigue. In my work with the AI surf-rock band The S.E.A.L.S. (Synthetic Erudition Assist Lattice), systems such as Jukebox and ChatGPT were never treated as authors. Instead, they functioned as what I call “soft AI”—provocations that sparked human curation, rewriting, and performance. Creativity did not reside in the machine’s fluency but in the human decisions about constraint, context, audience, and purpose.

Yet infinity is a resource, not an artistic goal. Generative systems have a structural blind spot: they do not know when to stop. They cannot feel exhaustion, satisfaction, or resolution; they do not inhabit time as bodies do. As Kate Crawford has argued, much AI output resembles “slop as metabolic waste”—a byproduct of systems that consume enormous amounts of energy, labor, and data without the regulating force of intent.

In the arts, however, stopping is among the most sophisticated skills one can develop. Musicians and dancers learn through repetition, error, repair, and embodied timing. They come to recognize “arrival”—that distinctive moment when tension resolves, a gesture lands, and a work coheres. AI can offer endless generation; humans offer completion. AI can suggest; humans commit. AI can vary; humans choose.

This distinction matters deeply for education. One of the greatest risks of our present moment is frictionless creativity, in which cognitive struggle—the engine of durable learning—is quietly outsourced to probabilistic systems. As Raffi Krikorian has warned, we are drifting toward a world in which intelligence is something you rent. But dependence on massive cloud models is not the only possible future.

I advocate for architecture-first approaches to AI, such as Yann LeCun’s Joint Embedding Predictive Architecture (JEPA). Rather than optimizing only for fluent text, these systems aim to build structured world models through interaction with environments. They emphasize prediction, embodiment, and consequence, learning more like bodies do: by encountering resistance, navigating space, and experiencing outcomes.

This orientation toward embodiment anchors my work on WeStep, a project that uses real-time sonification to support people with Parkinson’s disease. Wearable sensors translate gait patterns into sound, allowing participants to hear distortions in their movement and adjust in the moment. Learning unfolds through sensation and action rather than verbal instruction. This is embodied intelligence—a continuous feedback loop between perception and correction that no text generator can replicate. If AI is to become a genuine partner in collective intelligence, it must engage with the physics of the world, not only with patterns in language.

If AI threatens to remove “useful struggle” from making, then education must deliberately reintroduce it. I describe this as friction maxing: designing work where effort builds capability rather than bypassing it. Productive friction is not gatekeeping; it is formation. It sharpens judgment, deepens understanding, and cultivates resilience.

Yet individual effort alone is insufficient. We also need social forms that resist the infinite compliance of algorithmic feeds. For this, I turn to the Milonga as both metaphor and model. In Argentine tango culture, the Milonga is a third space structured by shared norms, mutual accountability, and nonverbal negotiation. The band watches the dancers; the dancers listen back. Beginners and masters share the floor, navigating limits through simple rules that enable live coordination. A modest fee from dancers supports musicians and space, binding community to material reality.

In a Milonga, connection cannot be rented—it must be earned moment by moment. Applied to AI, this suggests a relational stance: systems should be partners we listen to, resist, and guide, not authorities we obey. From the Milonga we learn to mix abilities and disciplines, invest in shared cost, listen and adjust, own community values, negotiate limits, gather for repeated encounters, and argue respectfully. These practices depend on embodied presence—the very quality threatened by frictionless generation.

We are entering a period in which we must consciously choose where learning lives. Digital abundance offers access and efficiency, but it can also dilute attention, flatten difference, and discourage risk. We need spaces that preserve presence, deliberation, and constructive conflict alongside technological possibility.

This is the work of the Tech & AI pillar in music education: not to reject tools, but to ensure that, amid infinite options, we retain the human capacity to decide what matters. Music connects not because it is flawless, but because it is shared; not because it is efficient, but because it is negotiated in real time among bodies in space.

The tension that structures this argument can be summarized as a set of necessary pairings—a space between automation and agency:

Where does learning live
At home with the recording…but also in the danger of a public circle
In the endless stream of content…but also in the discipline of stillness
In infinite generation...but also in determining the end
In the ease of agreement...but also in the difficulty of argument

These pairings refuse a simple choice between technology and tradition. Instead, they locate education in productive friction—simultaneously private and public, abundant and restrained, generative and decisive, cooperative and adversarial.

Ultimately, the future of creativity will not be measured by how much AI can produce, but by how wisely humans can decide, gather, argue, and arrive together. Friction is where learning happens; friction is where artistry emerges; friction is where collective intelligence becomes real. In the age of infinite generation, our task is not to eliminate resistance, but to design it—to create conditions in which machines extend human capacity while bodies, communities, and judgments remain at the center of creative life.