Challenges for Musical Education in the Age of AI and Digital Transformation

Synthesis: This paper maps the challenges that generative AI, streaming algorithms, and digital audio workstations pose for music education. Three converging transformations are examined: the changing nature of music creation and consumption, shifts in the public for music shaped by algorithmic curation, and the democratization of music production through digital tools. The paper surveys implications for both curricular content and pedagogical methods, arguing that music education must adapt to a landscape where AI can produce complete, stylistically coherent pieces from text prompts. Crucially, it distinguishes autonomous AI generators that position students as consumers of machine output from interactive composition assistants that extend student creative agency — arguing that only the latter is educationally productive — and it situates the economic precarity of musicians as a crisis intensified, but not created, by Generative AI.

Key Findings

1. Dematerialization of music — the shift from physical carriers to MP3 and streaming — converted music from scarce to abundant, collapsing per-unit recorded-music value by roughly two orders of magnitude well before generative AI became commercially relevant.
2. A small number of top artists and major labels capture a disproportionate share of streaming revenue, while per-stream payouts to the long tail of working musicians remain fractions of a cent, making streaming a "discovery layer" rather than an income source.
3. The paper sharply distinguishes autonomous AI generators (Suno, Udio) that produce end-to-end, opaque, compositionally uncontrollable output from interactive composition assistants (FlowComposer, the Continuator) that support and extend the human composer's intentionality — only the latter is educationally productive.
4. Education must shift toward production and DAW and AI fluency, while foregrounding the embodied, social dimensions of ensemble performance as the aspects most resistant to AI substitution.
5. Income support experiments, such as Ireland's Basic Income for the Arts pilot, show that decoupling creative income from the atomized sale of output can be economically constructive, offering a rare empirical alternative to streaming-era value collapse.
6. AI systems that are trained recursively on their own output risk "model collapse" toward low variance, amplifying the homogenization of musical style that algorithmic curation already accelerates.

The Nature of Music Has Changed

Each major technological shift — from the printing press that democratized musical scores to electric amplification — has forced educators to reconsider what, how, and why they teach. The paper argues three transformations are now converging simultaneously. The nature of music itself has changed: MP3 freed music from its physical carrier and streaming (Deezer, Spotify) replaced unit sales with subscription- and attention-based access, so that by 2024 streaming accounts for the overwhelming majority of recorded-music revenue in Western markets. Dematerialization dissolved scarcity: a fully equipped recording studio can now be approximated by a laptop, a microphone, and Open Source software, and better diffusion has come paired with lower value. Pachet's "paradox of creators" describes how digital tools give musicians unprecedented reach while structurally destroying the economic value of the content they distribute — a dynamic generative AI now deepens by flooding channels with machine-produced content at near-zero marginal cost.

This carries a direct educational implication: technical mastery of an instrument and its social prestige are no longer a sufficient differentiator. Barriers to creating and distributing music have fallen, but barriers to being heard and valued have risen. Music education must equip students for this duality — tools to reach audiences widely, alongside the artistic distinctiveness and economic literacy needed to sustain a creative life amid hyperabundance.

The Public Has Changed: The Rise of Functional Music

Streaming platforms have actively shaped listening habits through algorithmic recommendation, nudging users toward music suited to functional contexts — studying, exercising, sleeping, relaxing. The result is the growth of "functional music," valued primarily for its effect on mood and concentration rather than for the aesthetic and expressive qualities that have historically been the focus of music education. This shift is both quantitative and qualitative. Studies of popular Western music over five decades document a restriction in melodic variety, a homogenization of timbre, and increased dynamic compression — consistent with optimizing music for background listening. A tension emerges between music as autonomous artistic expression and music as functional content, a distinction with profound consequences for what educators choose to teach and value.

Music-Making Has Changed: DAWs and Generative AI

Digital audio workstations (Ableton, Cubase, Logic Pro) have transformed production practice from a top-down, score-centered approach into a bottom-up, sound-centered one built from loops, samples, and sonic textures. Students today may have spent years in a DAW environment, developing sophisticated intuitions about timbre and production aesthetics that traditional conservatory curricula were not designed to recognize. On top of this, generative AI has irrupted.

The paper insists on distinguishing two approaches with very different relationships between human and machine:

• Autonomous generators (Suno, Udio) produce complete, stylistically coherent pieces from a text prompt, well suited to functional music where speed and stylistic competence matter more than originality. But their generation is end-to-end: structure is opaque and uncontrollable, and they function as statistical interpolators over a learned stylistic space, structurally biased toward conformance. Trained on existing recordings, they also raise unresolved questions about copyright, consent, and fair compensation.
• Composition assistants (FlowComposer, the Continuator) support interactive practice, letting the musician impose constraints and retain full intentionality. FlowComposer assisted in the 2018 album Hello World; the Continuator learns a player's style in real time and responds as a stylistically attuned improvisational partner, generating strong engagement and a heightened sense of musical Learner Agency.

The paper's key educational claim: autonomous generators position students as consumers or curators of AI output, whereas composition assistants position AI as an extension of the student's own creative agency — only the second is educationally productive. It also flags a lingering divide between audio-based systems (rich but opaque) and symbolic systems (interpretable and controllable), with a future "Graal" of managing both simultaneously.

The Economics of Music Has Changed

The deep transformation of musicians' economics was largely complete before generative AI became commercially relevant. The shift from unit sales to pooled subscription revenue collapsed per-unit value, and per-stream payouts remain on the order of $0.003–$0.005 on most platforms. While aggregate payouts have grown (Spotify reported ~$11 billion in royalties in 2025), this growth is highly concentrated; streaming functions as a discovery layer rather than an income source. Generative AI did not initiate this dynamic; it prolongs and intensifies it, making recorded music abundant and cheap to produce as well as to access.

On rights, generative AI poses a qualitatively new problem: systems that absorb the statistical signature of a genre without copying any single identifiable work erode the market for all of them, a harm that copyright doctrine — designed to protect specific fixed expressions — is ill-equipped to address. As of 2026, litigation over training data is resolving toward negotiated licensing (e.g., Universal Music Group's and Warner Music Group's 2025 settlements with Udio and Suno), but the downstream musician whose stylistic niche is approximated has no specific infringement to point to, while purely AI-generated output is itself largely unprotectable. AI also offers positive prospects: automatic plagiarism detection, certification of musical creations, and royalty-management tools (e.g., URights by SACEM, CertCon by Cedro Rosa Digital).

For survival, musicians converge on diversification across income streams — licensing for TV/film, live performance as the most resilient revenue source, direct fan relationships via Patreon and Bandcamp, and teaching. Notably, teaching itself may become an economic refuge: at the very moment AI disrupts creation and distribution, human mentorship offers something increasingly scarce in an AI-saturated content economy — direct, accountable, personally invested transmission of musical knowledge. The paper also reviews Ireland's Basic Income for the Arts pilot, which returned an estimated €1.39 in social and economic value per €1 invested, as rare empirical evidence that decoupling creative income from direct sale can be economically constructive.

Education Has to Change

Online Platforms and Access

Before AI dominated educational technology, online platforms transformed access. YouTube has been the most consequential single development in informal music education, democratizing access to musical knowledge for learners where private instruction is inaccessible. MOOCs (Coursera, edX, Berklee, Juilliard) add structure, with the online music-education market estimated near $5 billion by 2030. Yet MOOCs face structural challenges: completion rates typically under 10%, and the embodied practice, real-time feedback, and social dynamics of ensemble playing resist short video lectures and self-graded quizzes. Research on social MOOCs suggests peer interaction and collaborative Problem Solving are essential, but hard to engineer at scale.

AI-Based Training: Tools, Opportunities, Limits

A systematic review by Sánchez-Jara et al. spans virtual and augmented reality to assistive technologies, with four consequential areas for teaching:

• Personalized Learning and Intelligent Tutoring: adaptive systems adjust difficulty, pacing, and content in real time. Research by Ou et al. found AI-assisted practice applications improved performance, Self-Efficacy, and Self-Regulated Learning. Tools like SmartMusic, Tonara, and Violin by Trala (over 400,000 users) provide the feedback loop practice requires even without a teacher.
• Automated Assessment and Feedback: historically limited to pitch accuracy and rhythmic precision, current research pushes toward assessing expressive qualities like phrasing and dynamics. AI-powered Learning Analytics dashboards let teachers monitor many students simultaneously, shifting the teacher's role toward supervision and interpretation.
• AI as a creative partner: generative tools can serve as musical sparring partners — a student asks for chord progressions or a "virtual Bach" harmonization to critique. But Doshi and Hauser found AI-assisted outputs more similar to each other, warning that AI-mediated creativity can become mimicry rather than genuine exploration; AI as scaffold is valuable for struggling students but teachers must guard against homogenization.
• The teacher's evolving role: no near-term AI can provide the context-sensitive, emotionally attuned, culturally grounded mentorship of expert human teaching — scale, consistency, and patience are not substitutes for modeling musical passion and sensitivity to a student's state.

The paper's most productive framing: not autonomous replacement but intelligent assistance — AI relieves teachers of repetitive, quantifiable feedback while freeing them to focus on what only humans can do.

Rethinking the Curriculum

Music education curricula need reconsideration, not marginal amendment. The paper proposes several directions: DAW and production literacy as a standard competency at all levels; AI Literacy for musicians as a practical and critical capacity (understanding what generative AI can and cannot do, evaluating outputs, and distinguishing AI-as-shortcut from AI-as-creative-catalyst); more prominent critical and contextual knowledge about rights, economics, and the social role of music; and defending ensemble and performance — the embodied, social, spontaneous dimensions most resistant to AI substitution.

What this means for practice

• Instructors. Treat DAW proficiency and AI fluency as core musical competencies rather than an optional add-on, and assign interactive composition assistants — FlowComposer, the Continuator — in preference to end-to-end generators such as Suno and Udio: only tools that let the student impose constraints and retain full intentionality extend creative agency, while a generator positions students as consumers or curators of output whose structure is opaque, and an "AI module" bolted onto existing curricula cannot meet a change this deep.
• Instructors. Assess for stylistic distinctiveness whenever AI is part of the task: Doshi and Hauser found AI-assisted outputs more similar to each other, so scaffold struggling students with these tools but do not reward polished conformance to the model's learned style — the homogenization risk extends to AI-mediated assessment.
• Instructors. Defend ensemble, performance, and improvisation as non-negotiable curriculum time, since the embodied, social, and spontaneous dimensions of musical practice are what no near-term AI supplies and what online delivery tends to efface — a caution the paper grounds in MOOC completion rates that typically sit under 10%.
• Instructors. Teach the economics and rights landscape as content: per-stream payouts on the order of $0.003 to $0.005, royalty growth concentrated at the top, negotiated licensing settlements, and the near-unprotectability of purely generated output, so students plan diversified careers (licensing, live performance, direct fan relationships, teaching) rather than streaming income.
• Administrators. Consider structural income responses and equip music students to advocate for them: the paper cites Ireland's Basic Income for the Arts pilot, which returned an estimated €1.39 in social and economic value for every €1 invested.

Limitations

• This is a single-author conceptual article, not an empirical study: there is no sample, intervention, or comparison group, so its claims about what works in classrooms rest on cited secondary literature and the author's synthesis rather than new data.
• The literature is surveyed selectively, with no stated search protocol or inclusion criteria, and the technologies and policies it examines are individual illustrative cases (FlowComposer, the Continuator, SmartMusic, Violin by Trala, the Irish pilot) rather than systematically compared evidence.
• The economic and policy figures are drawn from industry and news sources — platform per-stream payouts, Spotify's reported royalty total, the €1.39-per-€1 estimate for Ireland's Basic Income for the Arts, the roughly $5 billion online music-education market projection — and are not independently verified in the paper.
• The argument originated as a round-table presentation at a Brazilian music-education meeting and was constructed with the assistance of Claude Sonnet 4.6, and its cases are drawn mainly from European and Brazilian contexts; transfer to other systems and national curricula is untested.

Citation

Briot, J.-P. (2026). Challenges for Musical Education in the Age of AI and Digital Transformation.