Abstract
Every major change in how music is made and shared has also changed how we think about originality. This paper looks at generative artificial intelligence as the latest such change, using two thinkers who analysed earlier ones. Walter Benjamin argued that recording and mechanical reproduction weakened the “aura” of the unique work, the special presence of a single performance or object, while the copy still pointed back to an original. Jean Baudrillard argued that the process went further, until copies referred only to other copies and the idea of an original became a kind of nostalgia. The paper claims that AI-generated music is neither a further stage of reproduction nor a deeper level of simulation. It is a “second original”: sound that has never been performed or recorded, produced from patterns learned from a vast archive of existing music. It is new in time yet collective and derivative in origin, and it works by generating rather than representing. The paper then examines what this means for musicians, including economic pressure, creative identity, unequal access and unsettled questions of authorship, and for listeners, whose music becomes adaptive and personalised, which calls for a “dual literacy” of fixed works and open-ended generation. It argues that the outcome depends on the choices musicians, audiences and institutions make, and that neither resistance nor uncritical celebration is the right response.
Keywords: generative AI, music, originality, authorship, Walter Benjamin, Jean Baudrillard, aura, simulation, listening, creativity
Introduction
The history of music in the modern period is inseparable from the successive transformations of its technical conditions of production and reception. Each major shift in the means of making and circulating sound has altered not only the practical possibilities available to musicians and listeners but also the very concepts through which originality, authenticity and presence are understood. The present moment, marked by the rapid deployment of generative artificial intelligence systems trained on vast collections of recorded music, constitutes one such transformation. To grasp its significance, it is necessary to return to two earlier diagnoses of the fate of the artwork under conditions of technical mediation, those offered by Walter Benjamin and by Jean Baudrillard. Benjamin examined the loss of aura that accompanies mechanical reproduction. Baudrillard traced the further dissolution of the original into a regime of pure simulation and hyperreality. With the advent of artificial intelligence, a new phase has emerged. What emerges is neither a simple continuation of reproduction nor a further intensification of simulation but a paradoxical return of the original and an originality that arises precisely from the prior dissolution of every stable origin. This second coming of the original carries musical practice and musical experience into a domain beyond representation. The following study examines that claim in detail and then considers its consequences for those who make music and for those who hear it.
The Withering of Aura in Mechanical Reproduction
Walter Benjamin, writing in the nineteen thirties, confronted the cultural consequences of photography, film and the phonograph. These technologies made it possible to multiply images and sounds without limit and to distribute them across space and time in ways that earlier handcrafted or live performed works could never achieve. The unique painting hanging in a particular gallery or the singular performance unfolding in a specific hall possessed what Benjamin called aura. Aura was the quality of distance, the sense that the object occupied a unique place in history and that any encounter with it required a certain reverence or at least a recognition of its singularity. Mechanical reproduction destroyed that distance. A photograph of a painting could be held in the hand, examined at leisure and compared with countless other images. A recorded performance could be played in any room at any hour. The work of art was thereby emancipated from its dependence on ritual and cult, yet it was also stripped of the authority that had once flowed from its uniqueness.
Benjamin did not regard this loss as purely catastrophic. He saw in the withering of aura the possibility of a new collective mode of reception. The masses could now approach cultural objects on terms of equality rather than of hierarchical contemplation. Film, in particular, trained the viewer in a distracted and critical mode of attention that differed from the absorbed gaze demanded by traditional painting. Yet Benjamin remained alert to the political dangers that accompanied the new technical regime. Reproduction could serve progressive ends, or it could be harnessed by authoritarian powers that sought to aestheticise politics and to mobilise the masses through spectacle. The decisive point for the present argument is that mechanical reproduction still operated with reference to an original. The photograph referred to a scene that had once been present before the lens. The phonograph record referred to a performance that had once taken place in a studio or concert hall. The copy might be degraded or incomplete, but it remained a copy of something that had existed before the act of reproduction. The original was distant and its aura diminished, but it had not yet disappeared.
Music occupied a special place in this analysis. Unlike a painting, a musical work had always existed in a certain tension between the score as an ideal structure and the performance as a concrete event. Recording intensified that tension by fixing one performance and then circulating it as a stable object. The recorded track became the primary form in which most listeners encountered music. Live performance continued, but it now stood in a secondary relation to the recorded version that audiences already knew. The aura of the unique concert evening gave way to the familiarity of the disc that could be replayed indefinitely. Benjamin’s framework thus already pointed toward a music whose presence was mediated and whose originality was no longer guaranteed by the singular event.
Simulation and the Loss of the Original
Jean Baudrillard radicalised the diagnosis. In the order of simulation, the copy no longer refers back to an original that can be recovered or even clearly remembered. Signs circulate among themselves, generating a hyperreality in which the distinction between model and reality collapses. The map precedes the territory. The image produces the world it claims to represent. Music provided Baudrillard with a privileged example. The recorded track already separated sound from the living body of the musician. Sampling, digital editing and later algorithmic selection intensified the detachment until the listener encountered not a trace of a prior event but a self-sufficient sonic surface whose relation to any originating moment had become irrelevant.
In this condition, the original is not merely distant. It is lost, absorbed into an endless chain of mediations until the very idea of origin becomes a nostalgic fiction. One no longer asks whether a particular recording is faithful to a performance. One asks only whether the sonic surface is coherent, satisfying or useful. The hyperreal is not an illusion that conceals a real. It is a generation of models of a real without origin or reality. Music under these conditions becomes a pure play of signs detached from any grounding in a unique creative act or a unique historical moment. The listener moves through a landscape of tracks, playlists and remixes in which every element is already a replica of a replica. The search for authenticity becomes itself a simulated desire, a longing for an origin that the system has rendered impossible.
Baudrillard’s analysis remains indispensable because it can describe the cultural atmosphere in which music produced by artificial intelligence first appears. The training data on which generative models are built consist of millions of already mediated, already reproduced tracks. The models do not discover a hidden original. They extract statistical regularities from a corpus that is itself the product of earlier regimes of reproduction and simulation. In that sense, artificial intelligence inherits the hyperreal. Yet it also does something that neither Benjamin nor Baudrillard fully anticipated. It produces new sonic material that has never been performed or recorded by a human musician before the moment of generation. The output is original in a strict chronological sense. It is also thoroughly derivative, shaped by patterns extracted from prior human labour. This double character constitutes the second coming of the original.
Artificial Intelligence and the Second Coming of the Original
The generative model does not copy an existing track. It synthesises new sequences of pitches, rhythms, timbres and formal structures according to probabilities learned from the training data. The result can sound strikingly like the styles and emotional contours associated with human composition. It can also sound alien or surprising depending on the degree of constraint imposed by the prompt and the parameters of the system. In either case, the material has no prior existence as a performed event. There is no studio session, no concert hall, no living body whose gestures produced the sound. The originality is therefore not the uniqueness Benjamin associated with aura nor the pure invention of an unmediated creative subject. It is an originality born from the dissolution of the original and an emergence of the new from within the hyperreal itself.
This second coming is paradoxical. The new material is original because it has never existed before. It is also the product of a system that possesses no experiential relation to the act of making music. The model has never listened with pleasure or frustration, never practised an instrument, never felt the pressure of a deadline or the elation of a successful performance. It operates solely on statistical correlations. Yet the correlations are drawn from the accumulated labour of countless human musicians. The originality of the generated track is therefore a collective and anonymous originality, a distillation of historical patterns rather than the expression of a singular consciousness.
The second coming takes musical practice into a world beyond representation. Representation, whether in Benjamin’s mechanical copies or Baudrillard’s simulacra, still operates within a logic of reference, however attenuated. The photograph refers to a scene. The sample refers to a recorded moment. Even the most abstract simulation still plays with the memory of an absent real. Generative models, by contrast, do not primarily represent. They generate. The prompt may evoke a style or a mood, yet the resulting sound is not a depiction of an external musical event. It is a novel acoustic object whose conditions of possibility lie in the statistical regularities of the training data rather than in any correspondence to a prior performance. The listener confronts something that feels familiar and yet has no determinate source. The experience is no longer one of recognising a representation of music but of encountering music that has been produced by a system that itself has no experiential relation to the act of making music.
Beyond Representation
To speak of a domain beyond representation is not to claim that all reference has vanished. Listeners still hear echoes of known genres, known harmonic progressions, known vocal timbres. The models are trained on human music and therefore remain bound to the historical archive of human sound. What has changed is the mode of relation between the sounding object and any supposed origin. In representational regimes, the work points back toward an absent cause: a performance, a composer’s intention, a historical moment. In the generative regime, the work points primarily toward the process of its own production and toward the statistical space from which it was drawn. The question that arises is no longer “what does this represent” but “what does this make possible.”
This shift has consequences for aesthetic judgment. Traditional criteria of originality that rested on the uniqueness of an author’s vision or the singularity of a performed event become less decisive. New criteria emerge that concern the quality of the prompt, the skill with which outputs are selected and refined, the coherence of the larger structure into which generated material is inserted and the capacity of the resulting work to open unexpected perceptual or emotional pathways. Judgment moves from the recovery of an origin to the evaluation of an emergent event. The work is assessed less as a faithful or inventive representation of something prior and more as a successful or unsuccessful intervention in the ongoing flow of generated sound.
The same shift affects the temporal experience of music. A recorded track in the older regime was a fixed object that could be revisited identically. A generative system can produce endless variations, each slightly different and responsive to new prompts or even to real-time feedback from the listener. The musical object becomes less a stable artefact and more a continuously unfolding process. Presence is no longer the presence of a completed work but the presence of an ongoing generation. The second original is therefore not only an original that has never existed before. It is an original that exists only in the moment of its generation, and that can be regenerated differently at any subsequent moment.
Impact on Musicians
The consequences for musicians are uneven and often contradictory. For many professional and aspiring composers, the technology appears first as an economic threat. Generative systems can produce functional music for advertisements, game background tracks and even extended compositions at a fraction of the cost and time required by human labour. Commissions that once sustained session musicians, film scorers and producers of atmospheric or genre-specific work now migrate toward automated tools. The pressure is real, and it falls most heavily on those whose livelihoods depended on the production of music that could be described in terms of style rather than in terms of singular authorial vision.
Beyond the economic register lies a challenge to creative identity. If a model can generate material that satisfies the stylistic expectations of a given genre, what remains distinctive about the human contribution? Technical skill alone is no longer a sufficient answer since models already outperform average human competence in pattern recognition and recombination. The distinctive contribution may instead reside in the capacity for intentional risk, for deliberate deviation from statistical norms, for the infusion of lived contingency that no training corpus can fully capture. Musicians who treat the technology as a collaborator rather than a competitor often report that it expands rather than contracts their range. A composer may use a model to generate dozens of harmonic progressions, then select, transform and embed them within a larger structure that remains governed by human judgment. In such cases, the machine functions less as an autonomous author than as an instrument of unprecedented flexibility, one that can suggest possibilities the human mind might not have reached unaided.
Access to the most powerful models, the computational resources required to fine-tune them and the cultural capital needed to navigate the emerging market are unevenly distributed. Musicians already marginalised by industry structures may find themselves further displaced while those with institutional support or technical fluency gain new leverage. The technology therefore risks amplifying existing inequalities even as it democratises certain forms of production. A musician working alone with modest equipment can now generate orchestral textures that once required a full ensemble. The same musician may struggle to secure recognition or remuneration in a marketplace flooded with machine-generated content. Distinction shifts from the act of composition itself to the act of selection, framing and contextualisation. The musician becomes less a creator of notes than a director of generative processes, a role that demands new competencies and new forms of authorship credit.
The question of credit remains unsettled. When a track is produced through a combination of human prompting, model generation and subsequent human editing, who is the author? Legal frameworks designed for human creators and for fixed recorded objects prove inadequate. Ethical frameworks that emphasise the intentionality of a conscious subject struggle to accommodate a process in which intention is distributed across human and nonhuman agents. Musicians must therefore invent new practices of attribution and new ways of narrating their own creative labour. Some choose to foreground the collaborative character of the work, acknowledging the model as a partner. Others insist on the primacy of human selection and refinement, treating the model as a sophisticated tool. Both strategies are provisional. The deeper transformation is that authorship itself is being redefined under conditions in which the origin of musical sound is no longer exclusively human.
Impact on the Audience
Listeners experience a parallel transformation. The recorded track, long the primary medium of musical encounter, is no longer a fixed object. Generative systems can produce endless variations of a piece, each tailored to a user’s preferences or even to biometric feedback. The listener is no longer a passive recipient of a finished work but an active participant in the ongoing generation of sonic material. Playlists that once reflected human curation now incorporate algorithmic improvisation that responds in real time to attention patterns. The boundary between composition and consumption blurs. In the extreme case, the audience may never hear the same piece twice, yet the pieces remain statistically coherent within a chosen aesthetic frame.
This fluidity alters the emotional and social functions of music. Shared cultural reference points, the songs that once provided common ground across a generation, become more elusive when every listener can inhabit a personalised sonic environment. The loss is not absolute. Human-curated playlists, live performances and intentional releases of fixed tracks continue to create moments of collective attention. Yet the default mode of listening increasingly favours the private, the adaptive and the infinite. Baudrillard’s hyperreal finds a new intensity here. The music one hears is not a representation of a social world but a simulation tailored so precisely to individual desire that the social world itself begins to recede. At the same time, the second coming of the original reintroduces a strange form of authenticity. Because the generated material has never existed before, each listening event can claim a kind of uniqueness. The aura that Benjamin associated with the singular object returns not as a property of the artwork’s history but as a property of its unprecedented emergence in the present moment of generation.
The listener must therefore develop a dual literacy. One must remain capable of attending to fixed human compositions with the depth of concentration they invite. One must also become capable of navigating open-ended generative environments in which the musical object is never fully stable. This dual literacy is not merely a technical skill. It is an aesthetic and ethical disposition, a willingness to move between the security of the completed work and the uncertainty of continuous generation. Audiences that cultivate only one of these modes risk missing half of the musical present. Those that cultivate both stand to experience an enlargement of the range of possible musical pleasures.
The social dimension of listening is likewise transformed. Music has long served as a medium of shared identity, collective memory and public emotion. When each listener receives a uniquely generated stream, the common reference points that once anchored those functions grow weaker. New forms of commonality may emerge around particular models, particular prompting styles or particular communities of practice that share and discuss generated outputs. Yet these forms of commonality are more fluid and more fragile than those that once attached to a widely known recorded track. The audience is no longer a public that gathers around a shared object. It is a dispersed set of individuals, each moving through a personalised sonic landscape that occasionally intersects with the landscapes of others.
Toward a New Musical Condition
The historical sequence from Benjamin’s mechanical reproduction through Baudrillard’s simulacra to the generative present does not describe a simple decline. It describes a series of transformations in the conditions under which music comes into being and is received. Each stage dissolves a previous form of authenticity only to make possible a new one. Mechanical reproduction dissolved the aura of the unique object and made possible the critical mass audience. Simulation dissolved the distinction between original and copy and made possible the pure play of signs. Generative artificial intelligence dissolves the necessity of human performance as the sole origin of musical sound and makes possible an originality that is statistical, collaborative and perpetually emergent.
The task now is to inhabit this new condition without nostalgia for the lost original and without naivety about the power relations that shape access to the means of generation. Musicians must invent forms of practice that treat the generative model as a genuine partner while preserving the space for human judgment, risk and contingency. Audiences must cultivate the dual literacy that allows them to move between fixed and generative modes of listening. Institutions of education, criticism and legal recognition must adapt to a situation in which authorship is distributed, and originality is no longer the exclusive property of a conscious human subject.
None of these adaptations will be automatic. They will require deliberate effort and ongoing debate. The second coming of the original is not a technological inevitability that will automatically enrich musical life. It is a possibility that must be shaped by the choices of those who make and those who hear. If those choices are guided by a clear understanding of the historical transformations that have brought us to this point, the domain beyond representation need not be a domain of loss. It can be a domain in which new forms of musical presence, new forms of collaborative creation and new forms of attentive listening become available. The original returns not as a recovery of what was lost but as an invention of what has never yet been heard.
Conclusion
Walter Benjamin showed that mechanical reproduction altered the status of the unique artwork by withering its aura and opening new modes of collective reception. Jean Baudrillard showed that the further development of technical mediation dissolved the original itself into a hyperreal circulation of signs. Artificial intelligence does not simply continue these processes. It produces a second original, an originality that arises from the prior dissolution of every stable origin and that carries music into a world beyond representation. The impacts on musicians and audiences are already visible in economic pressures, shifts of creative identity, new forms of authorship and new modes of personalised and adaptive listening. The ultimate significance of these changes will depend on the practices and institutions that musicians, listeners and cultural authorities choose to build. The second original is neither a threat to be resisted nor a novelty to be celebrated without qualification. It is a transformation of the conditions of musical possibility that demands careful thought and deliberate shaping. Only through such thought and such shaping can the generative present become a genuine expansion of the human capacity to make and to hear music.
