Does Artificial Intelligence Dream of Electric Sheep?
A critical look at AI's role in art, the myth of originality and exploring the tension between creative remixing and cultural uniformity.
Bayraktar, Kerem Ozan. "Yapay zekâ elektrikli koyun düşler mi?" [Does Artificial Intelligence Dream of Electric Sheep?]. In Argonotlar Almanak 2023, edited by Kültigin Kağan Akbulut and Seçil Epik, 261-67. Istanbul: Argonotlar, 2023.
— Cahit Arf, Can a Machine Think and How Can It Think?, 1959
Artificial intelligence, already a staple in fields from healthcare to the military, is now rapidly being adopted in the worlds of art and design. As with the printing press, photography, and digital imaging before them, machines are once again driving major shifts in our cultural codes and aesthetic sensibilities. One reason AI development has focused so keenly on the arts is the inherent difficulty of mechanizing them. In our current, human-centered discourse, we love to measure machines against ourselves. Since creativity is considered a uniquely human trait, algorithms used in art and design are constantly being put to the test, a dynamic that has fueled their exponential development in recent years.
Machines have always been part of artistic production, but the learning and adaptation skills of today’s AI algorithms are far more advanced than those of their predecessors. Capable of processing vast datasets to generate specific outputs, they are now taking on tasks traditionally performed by humans. This utility, as a partner, an assistant, or, some might argue, a slave, has only increased their visibility and use, making human-machine interaction a central topic of debate. This new reality revives age-old questions about originality and the creative process, while also sparking new controversies over the value of art and the traceability of the artist’s contribution. What separates the developer who builds the algorithm from the artist who uses it? How much influence does the algorithm truly wield, and can we scrutinize the designer's decision-making process? How are our very definitions of creativity and originality being reshaped? And what does this mode of mass production reveal about our contemporary culture?
The fact that AI-generated works are the product of complex algorithms and technological processes makes their aesthetic evaluation a formidable challenge. It's therefore essential to grasp the role of technology, the nature of training data, and the dynamic partnership between the artist and the AI system.
The Myth of the Artist and Originality
Technological leaps often force us to confront slippery concepts like intelligence, consciousness, and creativity. Without a shared consensus on these terms, we risk asking the wrong questions about the wrong things. In the dialogue between AI and art, the way we define the artist and the concept of originality deeply influences how we perceive the work they create.
First, we must remember that the figure of the artist as we know it is a product of a relatively brief period of human history. While creative practices have always existed, the modern conception of the artist has not. This figure—the lone genius—was a historical consequence of the Romantic era's emphasis on individualism and expression. The myth carried over into modernism and, though somewhat tarnished, persists today, often for commercial purposes. This belief in innate, individual genius obscures the collaborative and interdisciplinary nature of art, at times erasing it completely. In truth, art has always been a collective endeavor. Renaissance workshops were hubs of artists working under a single master. Artistic movements like Dada, Surrealism, and Fluxus were founded on principles of collective creation. This collaboration extends beyond people to include machines, materials, techniques, and events.
Viewed from this wider perspective, no artist creates in a vacuum. Production is always a mix of different elements: learned knowledge, memories, emotions, physical reactions, and the influence of social and historical contexts. Yet, while collective, every creation is unique because it is filtered through the specific experience of its contributors. Each artistic process acquires its own character, shaped by the distinct combination of knowledge, memory, and feeling that flows through it. The link between this uniqueness and originality should not be understood as the mark of singular genius, but as a novel rearrangement of existing elements. Originality is not the creation of something from nothing. It is the act of combining familiar components in an unfamiliar way. This principle operates throughout the natural world: the formation of crystals, the growth of plants, and the mechanics of the cosmos are all processes where existing matter and forces combine to create new structures. Every crystal is unique, yet its uniqueness is simply another name for a different combination.
AI operates on this same principle. It learns from human-created datasets and generates new works by rearranging those existing elements. The algorithm's method of processing and organizing this trove of collective human knowledge directly shapes the final design. The artist's degree of control over this process varies, depending on the algorithm's architecture and the specific conditions of their work.
Remix
All creative work can be seen as a form of remix, where borrowing, quoting, adapting, and reinterpreting are essential acts. While the term is most associated with music, the practice is universal. The reconfiguration of existing ideas, materials, and images is a core component of creativity. Viewing cultural history through this lens reveals remixing across eras and media, seen in Ottoman miniatures influenced by Persian and Chinese traditions, Warhol’s soup cans, Tarantino's cinematic homages, the Renaissance’s reinvention of antiquity, Japanese prints shaped by Western perspective, and Bob Dylan’s reinterpretation of folk ballads. No creative field is free from this blending.
Naturally, inspiring, referencing, quoting, appropriating, and reinterpreting are unique actions, each driven by different intentions and leading to various results. In the 20th century, modernists consciously used techniques like pastiche and quotation to deconstruct the myth of the artist and question the very nature of authorship. The history of art includes pieces that are deliberately dull, unattractive, or lacking creativity; creations made by machines or unknown collaborators; and works that modify, disrupt, or even erase the art of others.
In the digital age, remixing has become effortless, migrating from the art world into the daily lives of millions. Anyone with a smartphone engages in constant remix: taking screenshots, sharing quotes, creating memes, editing podcasts, or producing TikTok videos. Remix is a defining feature of our culture.
AI tools with user-friendly interfaces have supercharged this process. By writing a few lines of text, anyone can generate images, sounds, or even code drawn from a vast sea of sources. Yet this is a unique kind of remix. While theoretically similar to collage, it is far more complex. Interacting with a program is completely distinct from working with physical materials. Though interfaces often simulate analog actions (like the "cut" tool in Photoshop), the metaphor only goes so far. AI engine works with language. The user submits a prompt, either a string of text or an image converted into text, and the program creates a remix based on it. The user can only guide the process to a certain degree, because the program possesses its own autonomy. This is the heart of the current debate.
Another key difference is that tracing the sources of a generative remix is nearly impossible (and, for most users, completely irrelevant). It’s less like collage and more like cooking a meal with ingredients from unknown origins. On platforms like Midjourney, users freely share their text prompts, which function like recipes in the generative art community. There is no anxiety about originality, so recipes are exchanged without hesitation. This culture has even given rise to the term "prompt engineering," a field dedicated to mastering the language of AI. Users share detailed notes on which words yield which results, the effects of punctuation, and which artists’ styles combine most effectively. In their quest for novel blends, many people with no prior interest in art history are now discovering artists and movements.
AI acts as a remix machine, learning and reproducing the styles, patterns, and compositional logics of the data it's trained on. But looking at an AI-generated image, it is incredibly difficult to identify its precise sources. Just as the name Rembrandt evokes not just figures and subjects but a whole atmosphere of light, shadow, and composition, AI-generated images have a similar atmospheric quality. The AI learns the atmosphere. You might feel a hint of Rembrandt in an image, but unless the prompt was explicit, pinpointing the reference is often impossible.
Then there is the issue of total hybridization. When you mix Rembrandt with a dash of Blade Runner's noir aesthetic, a touch of Hokusai's elegance, and the visual texture of a TikTok video, the original references dissolve into a blur of faint resemblances. The traces of origin are scattered across the pixels, like musical notes dispersed in a new composition. The moment one user's creation gains popularity, hundreds of variations are instantly produced by others. In this environment, individual authorship becomes almost meaningless. We are all entangled with a vast, collective machine.
The Risk of Homogenization
While every original work is a remix, not every remix is original. Each crystal may be unique, but at a certain scale, they appear identical. Culture tends to confer originality on broader shifts, not micro-level variations. As a result, many AI-generated images look strikingly similar, failing to become more than variations on existing genres. This tendency toward homogenization is driven by the structure of the algorithms, the biases in their training data, and the echo-chamber effects of online communities.
AI models are typically trained using a deep learning architecture known as a convolutional neural network (CNN), a mathematical model that mimics the neural networks of the human brain. These systems are trained to recognize and classify objects. When you type "cup," the AI generates images of a cup because it has learned to associate the word with a vast dataset of labeled images. Just as we learn to map our visual world with language from birth, AI is taught how to map and define the world. The training data is manually collected and labeled, and the CNN processes it in layers, learning to identify basic features like edges and corners before moving on to more complex patterns.
The content of these datasets directly determines what the model learns and, consequently, its biases. If the training data is incomplete or prejudiced, the AI will inherit and amplify those flaws. This is not a technical limitation but a result of human design. It is a mistake to think of these programs as objective; they are encoded with the ideological biases of the cultures that created them. More importantly, this learning model bypasses a fundamental aspect of art. Art is deliberately ambiguous in its pairing of words and images; it loves to play with and break the very maps AI relies on. This is why Magritte's “pipe is not a pipe”, and why Kosuth's "One and Three Chairs" are all different chairs.
Another factor driving similarity, particularly on platforms like Midjourney, is the feedback loop created by user engagement. When users upvote certain images and discard those they deem ugly, abstract, or nonsensical, the platform learns to favor popular aesthetic styles. While developers may try to balance these inputs, the trend is toward cleaner, clearer, and more conventionally appealing visuals. To what extent does eliminating errors and ambiguity inhibit a machine's creative potential? A program that understands words only in their most common usage and generates images in their most familiar forms would be useful, but would it not also be severely limited in its capacity for interpretation? We will likely have answers to these questions soon.
Of course, the tendency toward similarity is not unique to AI. Human culture is full of it. Popular films, music, and television shows often follow predictable formulas. Throughout history, different eras and regions have produced works with shared characteristics, giving rise to distinct styles. Impressionist paintings, for example, share a common visual language. The similarities in AI-generated work reflect the fact that the models are trained on the same pool of shared cultural values. Depending on your timescale, this can be seen as a form of homogenization.
Creative Remixes
Can new and original genres truly emerge from AI? Let's consider this through a biological metaphor. If a population of rabbits on an isolated island continually inbreeds, its genetic diversity will plummet. The rabbits will become increasingly similar and, eventually, highly susceptible to disease. This is a parallel to cultural stagnation. A system that is not open to outside influence will eventually collapse. To create different kinds of rabbits, you must enrich the gene pool. Similarly, the birth of new artistic genres will come from reinterpreting and transforming remixes across different cultural and disciplinary contexts.
While traditional notions of creativity focus on a clean break from the past, fostering new genres through remix requires a more dynamic and inclusive view. We must look beyond the intrinsic qualities of a single work and consider its interaction with different disciplines, cultures, and forms. It is also crucial to remember that evaluations of creativity are tied to context, scale, and personal experience. A washing machine may be a boring, repetitive object on its own, but in a work of art, it can acquire entirely new meaning. A city planner and a poet see the same sidewalk through different eyes.
Randomness can also enrich a system. When any system—biological or mechanical—encounters unforeseen data, it will interpret and respond based on its capacity for adaptation. Many AI programs allow for some control over randomness, enabling outputs that deviate from a prompt's literal interpretation. The key question is the breadth of the system's possibility space. How many faces does the die have? If an algorithm is programmed to optimize for specific parameters such as harmony or rhythm in music, its outputs will be guided toward a particular style, limiting its potential. All systems have such constraints.
Finally, we must distinguish between the machine’s technical ability to imitate and the artistic merit of what it creates. We may be impressed by an AI-generated poem, but is our admiration for the poem itself or for the fact that a machine wrote it? Often, the praise is implicitly directed at the ingenuity of the engineers who built the algorithm. Technological innovation and artistic creativity are related, but they are not the same. It is always possible to make original art with old technology, and it is a fallacy to assume new technology automatically yields new ideas. That said, the most creative work in this space often comes from artists who subvert these tools, transforming their functions or pushing them to their limits. In short, our fascination with technology itself should not be confused with art.
Autonomy
The degree of an AI's autonomy directly shapes the user's experience and creative influence. A highly autonomous system, by allowing for uncontrolled processes, might increase the potential for truly novel outputs. But a completely uncontrolled system risks producing chaos. Complex systems in art often strike a balance, maintaining overall control while allowing for a degree of freedom. Abstract Expressionism is a classic example. The process aims for a state of controlled chaos, where the artist guides the material but also allows it to assert its own properties.
How much control should a designer have? Who is the author of a work generated by an uncontrollable system? The answer often lies in the program's interface and the degree to which it allows for external intervention. A designer's technical skill, their fluency in programming languages and ability to fine-tune parameters, can also expand their level of control. This, however, does not mean that art requires technical mastery. Artists have always collaborated with specialists and machines. But a general understanding of how these systems work provides a crucial perspective on the opportunities and limitations they present.
Ultimately, the meaning of a work is shaped by how an artist chooses to engage with these tools, embracing their autonomy or asserting control. It is impossible to deduce this from a single image. A holistic view, considering an artist's entire body of work and process, is required.
Boundaries and Differences
I have tried to outline the general dynamics of AI as a machine for mixing, both controlled and uncontrolled. But when we speak of art, a purely technical or sensory evaluation is insufficient. The meaning of an artwork is layered, extending far beyond the pixels on a screen. An artist's intentions, their presentation methods, the work's dialogue with history and other art, and the entire universe of discourse surrounding it; all of these determine its value.
Achieving this kind of holistic view is becoming increasingly difficult in our culture of intensive remix. Context becomes fluid, and a sense of time and place begins to fade. Artistic movements that arose from specific historical conditions are reduced to aesthetic styles, stripped of their context and deployed as mere visual effects. We now see combinations that would be impossible in the physical world. Everything is constantly blending. AI generates photographs of prehistoric eras and paints Boeing jets with the brushstrokes of Van Gogh. When any piece of culture is digitized and fed into this system, it is abstracted from its original network of relationships. The boundaries between media (painting, cinema, games) and subjects (politics, comedy, landscape) dissolve, and the definitions of "artist" and "designer" become blurred. It is, as the movie title suggests, “everything everywhere all at once”.
This state of constant flux has itself become a dominant aesthetic. The imagery we so often see in New Media art, amorphous, perpetually melting and transforming, glitching and shimmering, resembling everything but being nothing, is an icon of our time.
To foster diversity and escape the pull of homogenization, we must emphasize difference. When all differences are erased, everything dissolves into sameness, like particles disappearing into water. This state, where criticism and even description become impossible, is akin to a thermodynamic death—a final equilibrium where all meaningful motion has ceased.
1 Arf, Cahit. "Makine Düşünebilir Mi ve Nasıl Düşünebilir?" In Atatürk Üniversitesi – Üniversite Çalışmalarını Muhite Yayma ve Halk Eğitimi Yayınları Konferanslar Serisi No: 1, 91-103. Erzurum, 1959.