Core 12: AI and Creativity
Narrative Engineering: The Core Basics - Part Twelve
Every significant technological shift in the history of creative industries has been accompanied by two things. The first is a wave of anxiety about what the technology will destroy: livelihoods, art forms, the conditions that made a particular kind of creative work possible. The second is a more important but considerably less visible question about who will own and control the new infrastructure the technology creates, and whose interests that infrastructure will be designed to serve.
The printing press generated anxiety about the authority of hand-copied manuscripts and the scribes whose livelihoods depended on them. Photography generated anxiety about the economic model of painted portraiture and the future of painters who had built careers on precisely the kind of representational accuracy that photography made instantaneous and cheap. Recording technology generated anxiety about the economic model of live performance, and the musicians who had previously been the only means through which audiences could access music suddenly had to reckon with recordings that could be reproduced and distributed without their presence.
In each case the anxiety was partially justified and partially misdirected. The technology did change the creative landscape significantly and produced genuine casualties among practitioners whose value proposition was too closely tied to the specific conditions the technology had displaced. But the most consequential effects were not the ones that generated the most anxiety. They were the structural effects: the shifts in who owned the infrastructure through which creative work reached audiences, who controlled the distribution that determined which creative work reached which audiences on what terms, and whose interests the new systems had been designed to serve. The practitioners who navigated each technological shift most successfully were not the ones who resisted the technology or the ones who adopted it most enthusiastically. They were the ones who understood its structural implications clearly enough to position themselves on the right side of the new infrastructure that it created.
Artificial intelligence is no different in structural terms, though it is different in the scale of its application and the speed at which it is reshaping the conditions of creative work. The anxiety about what AI will destroy in creative industries is real and in some respects legitimate. But the most important questions about AI and creativity are not about replacement. They are about ownership, authorship, infrastructure, training data consent, and the structural conditions that will determine whether AI becomes a tool that expands the capabilities of creative practitioners or a mechanism through which the economic value of human creativity is further concentrated in the hands of a small number of technology companies.
What AI Is Actually Doing to Creative Industries
To understand AI's relationship to creative industries clearly, it is necessary to separate three distinct phenomena that are frequently conflated in public discourse and that require different analytical frameworks.
AI as a production tool is the most straightforward of the three and in some respects the least consequential structurally. Tools including Adobe Firefly, Midjourney, DALL-E, Runway, and Sora allow creative practitioners to generate images, video, and other visual content from text descriptions, to automate time-consuming mechanical production tasks, and to iterate through creative options at a speed and scale that was previously impossible without significant additional resource. For practitioners who understand the tools and who have the creative judgment to direct them toward specific outcomes, these capabilities represent a genuine enhancement of creative productivity. The question of whether the outputs of these tools constitute creative work in any meaningful sense, and whose creative work they constitute, is a separate and more contested question.
AI as a distribution mechanism is less often discussed in conversations about AI and creativity but is equally significant for understanding its structural impact. The algorithmic recommendation systems that determine which creative content reaches which audiences on streaming platforms, social media platforms, and content discovery services are AI systems, and they have been shaping the distribution of creative work for more than a decade before generative AI attracted significant public attention. These systems were trained on engagement data reflecting historical patterns of attention, which means they tend to surface content that resembles what has already been successful with the audiences whose consumption patterns dominate the training data. The structural consequence is that genuinely innovative creative work, and creative work from practitioners outside the communities whose consumption patterns shaped the training data, is systematically disadvantaged by the AI systems that determine discovery. The recommendation engine is not a neutral conduit. It is an active shaper of what creative work reaches audiences, and it shapes that distribution in ways that reflect the historical patterns embedded in its training data rather than the intrinsic quality or cultural significance of the work it is or is not surfacing.
AI as a competitive product is the category that generates the most anxiety, and the anxiety is in some cases legitimate. The practitioners whose value proposition was primarily the efficient production of competent content at specified volumes and to specified standards face genuine competitive pressure from AI systems that can produce adequate content faster and at lower cost. Stock photography was the first creative category to experience this competitive pressure at scale, but it will not be the last, and the practitioners most exposed are those whose work is most accurately described as content production rather than as the expression of a distinctive creative perspective that no AI system trained on aggregated human output can replicate.
The distinction that matters here is between creative work whose primary value is in its efficient production at a specified standard, and creative work whose primary value is in the distinctive human judgment, developed aesthetic sensibility, and culturally specific perspective that constitute its most durable and most irreplaceable qualities. AI has reduced the cost of adequate content. It has not reduced the value of exceptional content, and the gap between adequate and exceptional has arguably become more commercially and culturally significant as adequate content becomes abundant rather than less significant as the narrative of AI replacement tends to suggest.
The Training Data Question
The most structurally significant AI issue for creative practitioners is not competitive displacement. It is the training data question, and it connects directly to the intellectual property framework examined in Core 10, because it represents the most consequential and most contested instance of the pattern this series has been documenting throughout: creative practitioners generating value that is captured by the infrastructure owners rather than by the creators.
The major generative AI systems were trained on datasets that included enormous quantities of creative work produced by living practitioners, scraped from the internet without their consent and without compensation. The images that trained Midjourney, Stable Diffusion, and DALL-E included the portfolio work of millions of photographers, illustrators, and visual artists whose aesthetic choices, technical approaches, and distinctive stylistic decisions became the raw material from which AI systems learned to generate visual content. The text that trained large language models included the published writing of millions of authors, journalists, essayists, and creative writers whose voice, style, and creative judgment now inform the outputs of systems that compete with their ability to earn a living from writing.
The legal status of this training data use remains actively contested. Lawsuits filed by visual artists against Stability AI, Midjourney, and DeviantArt, by the Authors Guild against OpenAI, by the New York Times against OpenAI and Microsoft, and by numerous other rights holders are making their way through courts in multiple jurisdictions, and the outcomes of these cases will shape the legal framework governing AI training data for decades. But the structural situation is already clear in its practical consequences, regardless of how the legal questions are eventually resolved. The creative output of a generation of practitioners was used, in most cases without their knowledge, consent, or compensation, to build commercial products that are now generating billions of dollars in revenue for the technology companies that built them and that compete directly with those practitioners' ability to earn a living from the specific creative skills and aesthetic capabilities the training data represented.
This is structurally identical to the pattern that has characterised the relationship between creative practitioners and infrastructure owners throughout the history of creative industries. The creativity originates with the practitioners. The economic value is captured by the infrastructure owners. The mechanism changes from era to era but the structural logic remains consistent, and the training data question is its most recent and most technologically sophisticated expression.
The specific case of visual artists is worth examining in detail because it illustrates the structural stakes most clearly. Illustrators and digital artists who had built professional practices over years of developing distinctive styles found their specific aesthetic approaches reproduced by AI image generators prompted with their names. Users could request images in the style of a living artist and receive outputs that bore unmistakable aesthetic similarity to that artist's work. The artist received nothing from the process and had no mechanism to prevent it. The AI company received the commercial benefit of having incorporated that artist's creative development into a product that competed with the artist's ability to charge for their work. The Getty Images lawsuit against Stability AI documented this with particular precision: Stability AI's training data included Getty's licensed image library, and the outputs of the AI system trained on that data sometimes reproduced Getty's watermarks, demonstrating not only that the training data had been used without licence but that it had been incorporated into the model in ways that reproduced identifiable features of the original images.
The Governance Question
The most important question about AI in creative industries is not what AI can do, which is a question whose answer changes rapidly and that has generated enormous attention. It is who governs AI development, on whose terms, and in whose interests, which is a question whose answer will be determined over the next several years and will shape the creative economy for decades.
The current situation, in which the governance of AI development is largely self-regulated by the technology companies building and deploying it, is structurally analogous to the situation that prevailed in the early decades of the music and film industries, when the companies that controlled distribution also controlled the terms on which creative practitioners engaged with those industries, set the royalty rates that determined practitioners' income, and designed the contractual frameworks that governed IP ownership and transfer. The outcomes of that self-regulation are extensively documented across this series: creative practitioners generated the cultural value, infrastructure owners captured the economic returns, and the gap between those two parties' interests was managed through contractual mechanisms that systematically favoured the infrastructure owners.
The regulatory frameworks now being developed in the European Union, the United Kingdom, and the United States will shape the AI landscape for decades, and the creative industry's engagement with those regulatory processes matters enormously. The EU's AI Act, which includes provisions relevant to training data transparency, establishes a precedent for AI regulation that other jurisdictions will reference. The specific questions of training data consent and compensation, of transparency requirements for AI systems that generate content commercially, of the liability framework for AI-generated content that reproduces or closely resembles existing creative work, and of the moral rights implications of AI systems that replicate individual creative styles without attribution or compensation, are not technical questions best left to technologists to resolve among themselves. They are questions about power, ownership, authorship, and whose interests the infrastructure of the creative economy is designed to serve. They require the active engagement of creative practitioners, creative industry institutions, and the policymakers who represent the interests of creative communities in regulatory processes that will otherwise be shaped primarily by the interests of the technology companies that stand to benefit most from minimal regulation.
The SAG-AFTRA and Writers Guild strikes of 2023, which secured meaningful contractual protections around AI use of actors' likenesses and AI's role in writing room processes, demonstrated that collective action by creative practitioners can produce structural protections in the AI context just as it has produced protections in every previous context where creative practitioners faced structural disadvantages relative to the institutions they negotiate with. The protections secured were not comprehensive, and the AI landscape is evolving faster than any collective bargaining agreement can fully anticipate. But they established the principle that AI's use of creative practitioners' work and likenesses requires negotiation and compensation, not simply the imposition of terms by the companies deploying the technology.
Cultural Authenticity and the AI Question
Beyond the economic and legal questions, AI raises cultural questions about creativity that deserve serious engagement rather than either dismissal or uncritical enthusiasm.
Creative work, at its most significant, is not the production of aesthetically acceptable or technically competent content. It is the expression of a distinctive human perspective shaped by specific experience, specific cultural context, and specific ways of understanding and inhabiting the world. The creative work that endures, that shifts culture rather than simply reflecting it, that creates genuine new aesthetic possibilities rather than recombining existing ones, is rooted in human specificity in ways that AI systems trained on aggregated human output cannot replicate because they have access to the patterns of human creative work but not to the human experience from which those patterns emerge.
For African creative practitioners specifically, this distinction carries particular weight. One of the most compelling and most durable arguments for the global significance of African creative work is precisely its cultural specificity: the aesthetic traditions, the conceptual frameworks, the ways of understanding beauty, narrative, and human experience that are rooted in specific African cultural contexts and that represent genuine creative alternatives to the aesthetic traditions that have historically dominated global creative industries. The cultural authenticity that makes Afrobeats compelling to global audiences is the product of specific human experience, specific cultural inheritance, and specific creative choices made by specific people who inhabit a specific world. Afrobeats cannot be generated by AI systems trained predominantly on Western creative output, not because the technology is insufficiently sophisticated, but because the cultural specificity that makes the music what it is is the product of human life in ways that training data cannot fully encode.
This is the most important thing that AI cannot do: be specific to a particular human life, cultural inheritance, and way of being in the world in the ways that constitute the most durable form of creative value. Understanding this clearly, and building creative practices around the qualities that most fully express what AI cannot replicate, is the strategic orientation that this moment in the development of AI requires.
What This Means for Creative Practice
Understanding AI's relationship to creative industries clearly leads to a set of practical orientations that are more useful than either reflexive resistance to the technology or uncritical adoption of it as a substitute for the creative judgment that constitutes the core of professional creative practice.
It means using AI as a tool where it genuinely enhances creative practice, in the generation of visual references, the exploration of compositional possibilities, the automation of mechanical production tasks, the rapid prototyping of creative directions, without allowing the ease of AI-assisted production to displace the development and exercise of the distinctive creative judgment that constitutes durable creative value. The creative practitioners who navigated the introduction of digital photography most successfully were not those who refused to engage with digital tools and not those who allowed digital tools to substitute for the photographic eye that had made their work distinctive. They were those who used the new capabilities of digital tools to serve the creative judgment they had developed rather than allowing those tools to define the limits of their creative thinking.
It means engaging seriously with the IP questions that AI raises: understanding what rights are implicated when creative work is incorporated into AI training datasets, what protections are legally available and developing, what contractual provisions to seek in agreements with companies that may use creative work in AI contexts, and how to structure creative practice in ways that maintain ownership and creative authorship in an environment where both concepts are under active pressure.
It means engaging with the governance questions at every level available: in collective bargaining through professional associations and unions, in consultation processes attached to regulatory development, in public discourse about what AI governance should look like and whose interests it should serve. The practitioners who had the most influence over the regulatory and contractual frameworks that shaped the music, film, and publishing industries were those who organised collectively and engaged with governance processes rather than treating those processes as too distant or too technical to be worth their attention.
The most important creative work of the next decade is not the work that uses AI most effectively, though effective use of AI as a tool will be a relevant professional capability across most creative disciplines. It is the work that most fully expresses what AI cannot: the irreplaceable human perspective, the cultural specificity rooted in lived experience, the distinctive creative authorship that no system trained on aggregated human output can generate from scratch. That work has always been the most valuable. In an environment of AI-generated adequate content abundance, it will be more valuable still.
This is the final piece in the Narrative Engineering: The Core Basics series. The next layer of the framework continues in the Builder tier, beginning with The Corporate Extraction of Creative Ideas, which examines the specific mechanisms through which the structural failures this series has mapped are actively perpetuated.