The Future of Creativity: Art, Music, and Writing in the …

A painter in Mexico City has been working on a series of large canvases for the past six months. She begins each painting by generating dozens of images in a diffusion model — feeding it prompts drawn from her reading, her dreams, the textures of the city around her. She prints the most promising compositions, projects them onto canvas, and paints over them in oils — layering, scraping, revising. The finished paintings do not look like AI art. They look like her work — but with a strangeness in the underlying geometry that she could not have arrived at on her own. She does not tell her gallery.

A composer in Berlin has been commissioned to score a documentary about the Aral Sea. He generates melodic sketches with a music model, edits them in his DAW, records live musicians over the synthetic stems. The score is beautiful — sparse, mournful, structurally inventive. The documentary’s director does not know which passages began as model outputs and which began at the composer’s piano. The composer is not sure he could reconstruct the distinction himself.

A novelist in Lagos has been blocked on her third book for two years. She opens a chat with a language model and begins describing the character who has been refusing to come alive — a middle-aged customs officer with a dying mother and an unresolved grief. The model suggests traits, scenes, gestures. Most are useless. A few strike something. She writes for six hours, then closes the chat. The next morning, she opens the manuscript and finds passages she does not remember writing — some of which are better than what she would have written alone.

A studio musician in Nashville signed a contract last year licensing her voice to a model training company. She was paid well. This year, a producer used her modelled voice on a demo that went to a major label. She was not credited. She was not paid. The producer does not believe he did anything wrong; the contract, his lawyer said, was standard. She is thinking about leaving the industry.

A high-school art teacher in Mumbai has spent the last semester trying to design assignments that cannot be completed by an image model. She has failed. Her students, who grew up with these tools the way her generation grew up with search, do not understand the objection. They are making strange, beautiful, hybrid work — but they are not learning to draw.

Five people, five creative practices, five relationships with a technology that did not exist in its current form a decade ago. What they are doing is not the death of creativity, and it is not its liberation. It is something else — a reorganisation of the relationship between intention, craft, and surprise that has been stable for centuries and is now shifting underfoot.

This is the deepest question the generative-AI era poses to the arts: not whether machines can be creative, but what creativity means when the boundary between the artist’s idea and the machine’s contribution becomes impossible to draw.

The Historical Context: Creativity and Its Tools

The idea that creative work is mediated by tools is not a discovery of the AI era — it is a feature of every creative practice that has ever existed. The painter’s brush is a tool. The composer’s notation is a tool. The novelist’s alphabet is a tool. The potter’s wheel, the weaver’s loom, the photographer’s camera, the musician’s instrument, the architect’s drafting table — every creative discipline is constituted by its tools, and every tool shapes what its discipline can produce.

The introduction of a new tool is therefore always a reorganisation of creative practice. The camera did not destroy painting, but it destroyed the economic case for representational painting and forced painters to invent new justifications for their work — impressionism, cubism, abstraction. The synthesiser did not destroy the orchestra, but it changed what orchestras were for and what new kinds of music could be made. The word processor did not destroy the novel, but it changed the texture of revision and made possible kinds of structural experimentation that were prohibitively tedious on a typewriter. The electric guitar did not destroy acoustic music, but it created a new musical vocabulary — distortion, feedback, sustain — that defined the sound of the twentieth century.

In each case, the new tool did not simply add a capability to an existing practice — it restructured the practice itself. The skills that mattered changed. The aesthetic values that mattered changed. The relationship between the creator and the audience changed. The economic arrangements that supported creative work changed. And the question of what counted as creative work — what counted as art, what counted as authorship, what counted as a contribution — was reopened, sometimes painfully, sometimes productively, always thoroughly.

The introduction of generative AI is a transition of this kind, but with a difference that matters. The tools that preceded AI were instruments — they extended the creator’s intention, executed the creator’s decisions, amplified the creator’s skill. A brush does what the painter’s hand directs. A piano does what the pianist’s fingers direct. The tool is downstream of the intention. The creator may be surprised by what the tool reveals — a brushstroke may produce an unexpected texture, a chord may produce an unexpected resonance — but the tool is not making a contribution that competes with the creator’s intention. It is executing it.

Generative AI is different. A diffusion model does not simply execute the painter’s intentions — it contributes compositions the painter did not specify and may not have imagined. A language model does not simply transcribe the novelist’s thoughts — it suggests phrasings, scenes, character traits that the novelist did not direct. The tool is no longer downstream of the intention; it is in conversation with it. The creator’s role shifts from execution toward selection, curation, editing, direction — toward, in some accounts, a kind of curatorship that resembles a film director’s relationship with the contributions of cinematographer, screenwriter, actors.

This shift does not make the creator less creative in any absolute sense. A film director is not less creative than a novelist for working in a collaborative medium. But it changes what creativity consists of in the practice. The skill of execution — the brushstroke, the sentence, the melodic line — becomes less central. The skill of selection — of recognising which of many possible outputs deserves development — becomes more central. And the question of attribution — who is responsible for the work, who gets to be called its author — becomes harder to answer in the terms the discipline has historically used.

The New Tools: What Generative AI Actually Does

To understand what is happening to creative practice, it helps to be precise about what the tools actually do. The generative AI systems that have emerged since 2022 — diffusion models for images, transformer-based language models for text, generative audio models for music and speech — share a common structure. They are trained on very large datasets of existing work in their domain. They learn the statistical regularities of that work — the distributions of pixels in paintings, the distributions of words in novels, the distributions of notes in melodies. When given a prompt, they generate new outputs that are statistically consistent with the distributions they have learned.

The outputs are not copies of the training data. They are new compositions that share structural properties with the training data. A diffusion model trained on photographs of faces does not retrieve a face from its training set when asked to generate one — it constructs a new face that has the statistical properties of the faces it has seen. A language model trained on English novels does not retrieve sentences from those novels — it constructs new sentences that have the statistical properties of the sentences it has seen.

This is the source of both the power and the controversy. The power is that the models can generate novel outputs at a speed and scale that no human creator can match. The controversy is that the outputs are derived from the creative work of the humans whose work was in the training data — work that was used without consent, without credit, and without compensation in most cases. The models are, in a meaningful sense, the product of the creative labour of millions of human creators who did not agree to participate in the production of the systems that are now being used to compete with them.

The technical question of whether this constitutes copyright infringement is being litigated in courts around the world, and the answer is likely to vary by jurisdiction and by the specific facts of each case. But the cultural question is not primarily a legal one. The cultural question is whether the relationship between the creators whose work trained the models and the creators who use the models is one that the broader creative culture can endorse. And the answer to that question is, at minimum, not obviously yes.

What the tools do, technically, is generate outputs that satisfy the constraints specified in a prompt while being consistent with the statistical structure of the training data. What they do, culturally, is redistribute the creative labour of the past into the creative outputs of the present. The painters whose work is in the training set are, in a statistical sense, contributing to every image the model produces. The novelists whose work is in the training set are contributing to every passage the model writes. This contribution is invisible in the output — there is no way to attribute any specific pixel or word to any specific source — but it is real in the aggregate. The models would not work without the training data, and the training data is the creative labour of human creators.

The Question of Authorship: Who Made This?

The traditional model of authorship in the creative disciplines is straightforward: the author is the person who made the work. The painter is the author of the painting. The composer is the author of the score. The novelist is the author of the book. Authorship carries with it both credit (the right to be identified as the creator) and control (the right to decide how the work is used, reproduced, and built upon). The entire structure of copyright law, the entire economics of creative industries, and the entire cultural understanding of what it means to be a creator rest on this model.

Generative AI disrupts this model because it introduces a third party into the authorship relationship. The work is no longer made by a single creator; it is made by a creator using a tool that was trained on the work of other creators. The question of who the author is becomes harder to answer. Is the author the person who wrote the prompt? The person who selected the output? The person who fine-tuned the model? The people whose work was in the training data? The company that built the model? Some combination of all of them?

The legal system is beginning to develop answers to these questions, and the answers are likely to be complex and context-dependent. The US Copyright Office has issued guidance indicating that works generated entirely by AI are not copyrightable, but that works created by humans using AI as a tool may be copyrightable to the extent of the human contribution. Courts are grappling with the question of whether training a model on copyrighted works constitutes infringement. The answers will emerge case by case over the next several years.

But the cultural question — who should be considered the author of an AI-assisted work — is not a question that the legal system can settle. It is a question that the creative community itself must work through, and the answer is likely to be different in different disciplines. In the visual arts, where the tradition of appropriation and remix is well-established, the integration of AI tools may be absorbed into existing practice without fundamentally disrupting the concept of authorship. In the literary arts, where the author is more central to the work’s identity, the integration may be more disruptive. In music, where collaboration and sampling have always been part of the practice, the integration may be relatively smooth.

What is clear is that the simple model of authorship — the author is the person who made the work — is no longer adequate for many creative practices. The new model will need to account for the contributions of the tool, the training data, and the human creator in a way that is fair, clear, and conducive to the continued production of creative work. Developing this model is one of the central cultural tasks of the AI era.

The Economic Disruption: What Happens to Creative Labour

The economic impact of generative AI on creative professions is the most immediately pressing question for working creators. The impact is uneven — it falls hardest on the creators whose work is most easily substituted by AI outputs, and lightest on the creators whose work depends on skills that AI does not yet have.

The most exposed categories of creative work are those that involve the production of routine, high-volume, moderately-skilled outputs. Stock photography. Background music for videos and games. Routine copywriting — product descriptions, marketing emails, SEO content. Illustration for articles and reports. Translation. Voiceover work. These are categories where the demand is large, the per-unit value is low, and the AI outputs are already good enough to meet the demand in many cases. The creators who work in these categories are facing a direct substitute, and many of them are already seeing their incomes decline.

The less exposed categories are those that involve higher-skill, lower-volume, more context-dependent work. Original reporting. Long-form narrative writing. Painting and sculpture that responds to specific physical and cultural contexts. Music composition that integrates with specific performers and venues. Acting. Directing. The creators who work in these categories are less likely to be directly substituted, but they may still be affected — by the downward pressure on rates as displaced creators move into their categories, by the increasing expectation that they will use AI tools to be more productive, and by the broader cultural shifts that change what audiences expect from creative work.

The least exposed categories are those that depend on human presence, relationship, or authority. Live performance. Teaching and mentorship. Commissioned portraits. Curatorial work. The creators who work in these categories may benefit from the AI era — their work becomes more valuable as the supply of AI-generated content increases and the value of human presence and authority becomes more salient. But the categories themselves may be small relative to the categories that are being disrupted.

The net effect on the creative economy is therefore likely to be a redistribution — away from routine content production and toward high-skill creative direction, human-presence work, and the management of AI-assisted creative workflows. This redistribution will produce winners and losers, and the losers are likely to be the creators who are currently in the middle of the skill distribution — the working photographers, illustrators, copywriters, and musicians who have built careers on the production of moderately-skilled creative content. The winners are likely to be the creators at the top of the distribution — the well-known names whose work acquires additional value by virtue of being human-made — and the creators who learn to use AI tools to extend their productivity.

The policy question — whether anything should be done to mitigate the disruption, and if so what — is being actively debated. Proposals include collective licensing schemes that would compensate creators whose work is in training data, retraining programmes for displaced creators, public funding for human-created work, and regulatory restrictions on the use of AI in specific creative categories. Each of these proposals has merits and difficulties, and the right policy mix is likely to vary by jurisdiction and by discipline.

The Legal Question: Copyright, Training Data, and the Right to Train

The legal status of training AI models on copyrighted creative work is one of the most consequential open questions in intellectual property law. The question is being litigated in multiple jurisdictions, with cases brought by authors, visual artists, music publishers, news organisations, and others against AI companies. The outcomes of these cases will shape the development of the AI industry and the economics of creative work for decades.

The core legal question is whether training a generative AI model on copyrighted works constitutes a fair use of those works or an infringement of the copyright holders’ exclusive rights. The fair use doctrine, as developed in US law, considers four factors: the purpose and character of the use, the nature of the copyrighted work, the amount and substantiality of the portion used, and the effect of the use on the potential market for the copyrighted work. AI companies argue that training is a transformative use that does not reproduce the works in any meaningful sense and does not compete with the original works in their markets. Copyright holders argue that training is a wholesale copying of their works for a commercial purpose that directly competes with them in the market for creative content.

The courts have not yet settled the question, and the rulings that have emerged so far have been mixed. Some early rulings have favoured AI companies on certain motions; others have allowed plaintiffs’ claims to proceed to discovery. The question is likely to be resolved ultimately by appellate courts and possibly by Congress, and the resolution will have significant implications for both the AI industry and the creative industries.

The legal question is distinct from the cultural question of whether training on copyrighted works without consent is appropriate even if it is legal. Many creators argue that the practice is wrong regardless of its legal status — that using someone’s creative work to train a system that will compete with them, without their knowledge or consent, is a violation of the implicit norms that have governed creative communities for centuries. This argument has moral force even where it lacks legal force, and it is shaping the broader cultural response to generative AI.

A related question is whether AI-generated outputs that resemble the style of a specific creator infringe on that creator’s rights. Style is generally not copyrightable — anyone is free to paint in the style of Picasso or write in the style of Hemingway — but the question of whether an AI model trained on a specific creator’s work and then prompted to produce outputs in that creator’s style crosses a different line is being actively debated. Some creators have argued that this practice constitutes a form of passing off or unfair competition, even if it does not constitute copyright infringement.

The resolution of these legal questions will shape the structure of the AI industry. If training on copyrighted works is fair use, the AI industry will continue to develop largely as it has, with models trained on large datasets of copyrighted works without compensation to the original creators. If training is infringement, the industry will need to develop licensing schemes, opt-out mechanisms, or other ways of compensating creators — and the cost of training models will increase significantly. The outcome is genuinely uncertain, and it is one of the most important open questions in the development of the AI era.

The Educational Question: Can You Learn to Be Creative When the Machine Can Do It

The educational question is in some ways the deepest and most difficult. If a machine can produce competent creative work — a passable essay, a competent illustration, a serviceable melody — what is the value of teaching a human to produce such work? And if the human is taught, what exactly is being taught?

The traditional answer is that the value of learning to produce creative work is not primarily the production of the work itself — it is the cognitive, emotional, and aesthetic development that the practice of producing creative work produces in the learner. The student who learns to write essays learns to think clearly, to develop and defend arguments, to engage with evidence. The student who learns to draw learns to observe carefully, to translate three-dimensional perception into two-dimensional representation, to develop manual skill and aesthetic judgment. The student who learns to compose music learns to hear structure in sound, to develop the patience and discipline of craft, to make aesthetic decisions under constraints.

These cognitive and aesthetic benefits are real, and they are not automatically produced by the use of AI tools. A student who uses an AI to write their essay does not, by that act, learn to think clearly. A student who uses an AI to produce an illustration does not, by that act, learn to observe carefully. The cognitive and aesthetic development that traditional creative education produces is the result of the practice of doing the work — and the practice is precisely what is being skipped when the AI does the work.

The educational challenge is therefore to design learning experiences that produce the cognitive and aesthetic benefits of traditional creative practice while engaging honestly with the reality that the practices themselves are being transformed by AI. This is not a simple matter of banning AI tools in educational settings — though some institutions have tried this — because the students will be entering a world in which AI tools are part of the creative practice they will be engaging in. The challenge is to design education that uses AI tools thoughtfully, in ways that develop rather than bypass the cognitive and aesthetic capacities that creative practice has traditionally cultivated.

What this looks like in practice is still being worked out. Some educators are designing assignments that require students to produce work without AI assistance, on the grounds that the cognitive development depends on the practice. Others are designing assignments that require students to engage critically with AI outputs — to evaluate, revise, and improve them — on the grounds that the cognitive skill of selection and revision is itself valuable. Others are designing assignments that require students to use AI tools in specific ways — to generate drafts that they then develop, to explore variations that they then choose among — on the grounds that the skilled use of AI tools is itself a cognitive skill that needs to be developed.

The right answer is likely to be a mix of all of these approaches, calibrated to the discipline, the level, and the learning goals. But the deeper question — what is the value of learning to produce creative work when the work can be produced by a machine — is not a question that any specific pedagogical approach can answer. It is a question about what kind of cognitive and aesthetic development we value, and why. The answer that the creative education community arrives at will shape the next generation of creators and the creative culture they produce.

The Aesthetic Question: What Does AI Art Look Like

There is a distinct aesthetic that has emerged in the first years of generative AI art. It is characterised by a certain kind of visual excess — too-perfect surfaces, too-smooth transitions, too-many details that resolve into a kind of generic prettiness. The images are often striking on first viewing and slightly hollow on closer inspection. They have what some critics have called the “AI look” — a quality of having been produced by a system that has learned the surface regularities of images without understanding the underlying structures that give images their meaning.

This aesthetic is a function of the technical limitations of current generative models. The models learn statistical regularities in the training data, and the statistical regularities tend to produce outputs that are typical of the training data — which means, in practice, that the outputs tend toward the average. The more the model is pushed toward specific outputs by detailed prompts, the more it tends toward the generic. The result is an aesthetic of competence without personality — images that look like images, music that sounds like music, text that reads like text, but that lacks the specific formal decisions that give a work its identity.

This is not a permanent limitation. The models are improving, and the aesthetic is evolving. Some artists are developing techniques for using the models in ways that produce more distinctive outputs — through careful prompt engineering, through the use of fine-tuned models trained on specific datasets, through post-processing and combination with traditional techniques. The Mexico City painter in the opening vignette is one example; her hybrid practice of generating images and then painting over them produces work that has both the formal strangeness of the AI output and the material presence of painting.

But the question of whether AI-assisted work can develop a distinctive aesthetic — an aesthetic that is not just a reproduction of the training data’s regularities but a genuine formal contribution — is open. The history of new tools in the creative disciplines suggests that it can. The camera produced a distinctive aesthetic that was not available to painting — the instantaneous capture of a moment, the optical properties of specific lenses, the grain of specific films. The synthesiser produced a distinctive aesthetic that was not available to acoustic instruments — the sustained pure tone, the filter sweep, the precise modulation. Generative AI may produce its own distinctive aesthetic — but it has not yet done so in a way that is widely recognised as such.

The deeper aesthetic question is whether the formal contributions of generative AI will be of a kind that the creative culture values. The camera’s contribution was valued because it expanded what could be represented and how. The synthesiser’s contribution was valued because it expanded what could be sounded and how. Generative AI’s contribution may be the expansion of what can be combined and how — the ability to merge styles, to generate variations at scale, to explore formal possibilities that would be prohibitively time-consuming to explore manually. Whether this contribution is valued will depend on whether the creative work that uses it is good enough to justify the valuation.

The Collaborative Model: Toward a New Division of Creative Labour

The most productive way to think about the role of generative AI in creative practice may be as a collaborator rather than a substitute. The model is not the creator; the model is a participant in the creative process, contributing outputs that the human creator then selects, edits, develops, and integrates into a finished work. The creative labour is redistributed between the human and the machine, but it is not eliminated — it is reorganised.

This collaborative model has precedents. The film director collaborates with cinematographers, editors, actors, composers, production designers — each of whom contributes to the final work, but none of whom is the sole author. The architect collaborates with structural engineers, contractors, clients, and the builders who execute the design. The composer collaborates with performers, conductors, and the acoustic properties of the performance space. Creative work has always been, in many disciplines, a collaborative activity — and the concept of authorship has evolved to account for this collaboration.

The difference with generative AI is that the collaborator is not a human being with their own creative intentions and aesthetic judgment — it is a system that generates outputs in response to prompts. The collaboration is asymmetric: the human creator directs, selects, and develops; the AI generates and suggests. The creative intentions and aesthetic judgment remain with the human — but the execution is shared.

This model has implications for how creative work is valued and credited. If the work is the product of a collaboration between a human creator and an AI tool, the credit for the work should reflect that collaboration. The human creator deserves credit for the direction, selection, and development — for the decisions that shaped the work into its final form. The AI tool does not deserve credit — it is a tool, not an agent — but the use of the tool should be acknowledged, in the way that a painter might acknowledge the use of specific materials or a composer might acknowledge the use of specific instruments.

The collaborative model also has implications for how creative work is taught and learned. If the skill of selection and development is central to the new creative practice, then education should focus on developing that skill — and on developing the aesthetic judgment that allows the creator to recognise which outputs deserve development. This is a different skill from the skill of execution, and it requires a different kind of education. The traditional atelier model, in which students learn by copying the work of masters, may be relevant here — the student uses the AI to generate many variations, and then studies them carefully to develop the judgment of which variations work and why.

The collaborative model is not the only model. Some creators will choose to work entirely without AI, on the grounds that the value of their work depends on the unmediated quality of human creation. Some creators will choose to work entirely with AI, on the grounds that the new possibilities the tools open up are worth the loss of unmediated creation. Most creators will probably choose something in between — using AI tools for some aspects of their work and traditional techniques for others, in a mix that reflects their own aesthetic, their own skills, and the specific demands of each project.

The Cultural Question: What AI Does to the Culture of Art

The deepest question raised by generative AI in the creative disciplines is not about individual creators or individual works — it is about the culture of art itself. The culture of art is the network of institutions, practices, values, and conversations that constitutes the creative life of a society. It includes the galleries and museums that exhibit work, the publishers and record labels that distribute work, the critics and journals that evaluate work, the schools and universities that teach work, the audiences that consume work, and the broader public conversation about what creative work matters and why.

This culture is being affected by generative AI in several ways. The galleries and museums are grappling with whether to exhibit AI-assisted work, and if so how to contextualise it. The publishers and record labels are grappling with whether to publish AI-assisted work, and if so how to compensate the creators whose work trained the AI. The critics and journals are grappling with how to evaluate AI-assisted work — whether to apply the same standards as for human-created work, or to develop new standards that account for the new medium. The schools and universities are grappling with how to teach creative practice in the AI era, as discussed above. The audiences are grappling with whether they care whether a work was created with AI assistance, and if so how that affects their valuation of the work.

The most significant cultural shift may be in the relationship between creative work and the broader public conversation. The traditional model of this relationship — in which creative work is produced by a small number of creators, evaluated by a small number of critics, and consumed by a larger audience — is being disrupted by the democratisation of creative production that generative AI enables. When anyone can produce competent creative work with AI assistance, the role of the professional creator changes. The professional creator is no longer the person who can produce the work — anyone can do that. The professional creator is the person whose work is worth paying attention to, for reasons that have to do with the specific quality of their judgment, their direction, their development of the work, rather than with the mere fact of having produced it.

This shift may, in the long run, be healthy for the creative culture. It may force a more honest reckoning with what makes creative work worth attending to — a question that the culture has often been able to avoid by relying on the mere fact of professional production as a proxy for quality. But the transition is painful for the creators whose professional standing depends on the old model, and the new model has not yet stabilised into a form that can support the creative culture as robustly as the old one did.

The deeper cultural question is what the creative culture is for — what role it plays in the broader life of a society, and what kind of creative culture best serves that role. The traditional answer, articulated in various ways by philosophers and critics from Aristotle to the present, is that the creative culture is a primary vehicle for the exploration of human experience — for the articulation of what it feels like to be alive, for the working out of values and meanings, for the transmission of cultural memory and the imagination of cultural possibility. This role is not necessarily dependent on the specific practices of professional creation that have characterised the modern era. It is dependent on the existence of a vibrant creative culture in which work of genuine human significance is produced, evaluated, and shared. Whether such a culture can be sustained in the AI era — and what it would look like — is the deepest open question.

The Long View: Creativity After the Tools Changed

The introduction of generative AI into the creative disciplines is a transition of the kind that has happened before in the history of the arts, and it will produce changes of the kind that previous transitions have produced. Some practices will disappear. Some practices will be transformed. Some new practices will emerge. The creative culture will be different in ways that are difficult to predict in advance but that will, in retrospect, seem inevitable.

The long view is useful here because it reminds us that the creative culture has always been in transition. The practices that we now think of as traditional — the novel, the symphony orchestra, the oil painting, the feature film — are themselves the products of specific historical transitions, and they have not always existed in the forms we now know. The novel as a literary form is a product of the eighteenth century; the symphony orchestra in its modern form is a product of the nineteenth century; the feature film is a product of the twentieth century. Each of these forms emerged from a transition that disrupted the practices that preceded them, and each was, in its time, the subject of anxiety about whether the new form would destroy the creative culture that preceded it.

The anxiety was not always misplaced. The transition to the novel did destroy the market for certain kinds of narrative verse. The transition to the symphony orchestra did destroy the market for certain kinds of chamber music. The transition to the feature film did destroy the market for certain kinds of live entertainment. The creative culture was changed by each of these transitions, and the change was not uniformly positive — there were real losses, real practices that disappeared, real practitioners whose careers were disrupted.

But the creative culture as a whole survived each transition, and in some ways flourished. The novel did not destroy literature; it transformed it. The symphony orchestra did not destroy music; it transformed it. The feature film did not destroy theatre; it transformed it. The creative culture adapted to the new tools, found new ways to use them, and continued to produce work of genuine human significance.

The transition to generative AI is likely to follow a similar pattern. Some practices will be lost. Some practitioners will be disrupted. But the creative culture as a whole is likely to survive and, in some ways, flourish. New forms will emerge that use the new tools in ways that are now difficult to imagine. New aesthetics will develop that exploit the specific capabilities of the new medium. New relationships between creators and audiences will form that are enabled by the new tools. The creative culture will be different — but it will not be dead.

The deepest question is not whether the creative culture will survive the transition — it will — but what kind of creative culture will emerge on the other side. This is a question that the creative community itself will answer, through the work it produces, the practices it develops, the institutions it builds, and the values it chooses to affirm. The answer is not determined by the technology; it is determined by the choices that creators, audiences, and institutions make in response to the technology. Those choices are being made now, in studios and classrooms and galleries and publishers and legislatures and conversations, and the answer that emerges will shape the creative culture for generations.

What is clear is that the transition is real, that it is significant, and that it is not going to be reversed. The tools are here, they are being used, and they are being improved. The question is not whether to engage with them — engagement is happening, whether one wants it to or not — but how to engage with them in a way that preserves what is most valuable about the creative culture while being open to the new possibilities the tools enable. This is the task of the present generation of creators, critics, educators, and audiences. It is a difficult task, but it is not an unprecedented one. The creative culture has navigated transitions before. It will navigate this one.

The Future of Creativity: Art, Music, and Writing in the AI Age

The Historical Context: Creativity and Its Tools

The New Tools: What Generative AI Actually Does

The Question of Authorship: Who Made This?

The Economic Disruption: What Happens to Creative Labour

The Legal Question: Copyright, Training Data, and the Right to Train

The Educational Question: Can You Learn to Be Creative When the Machine Can Do It

The Aesthetic Question: What Does AI Art Look Like

The Collaborative Model: Toward a New Division of Creative Labour

The Cultural Question: What AI Does to the Culture of Art

The Long View: Creativity After the Tools Changed

Further Reading

Comments

The Historical Context: Creativity and Its Tools

The New Tools: What Generative AI Actually Does

The Question of Authorship: Who Made This?

The Economic Disruption: What Happens to Creative Labour

The Legal Question: Copyright, Training Data, and the Right to Train

The Educational Question: Can You Learn to Be Creative When the Machine Can Do It

The Aesthetic Question: What Does AI Art Look Like

The Collaborative Model: Toward a New Division of Creative Labour

The Cultural Question: What AI Does to the Culture of Art

The Long View: Creativity After the Tools Changed

Further Reading

Comments

Subscribe