From Mythologies to Algorithms: Reading AI with Roland Barthes

Introduction

Artificial intelligence has rapidly become surrounded by a remarkably stable vocabulary. We are told that AI is inevitable, that algorithms know what we want, that data never lies, that machines will replace human workers, and that the future belongs to those who adapt. These expressions recur in newspaper headlines, corporate presentations, political speeches, and everyday conversation with such regularity that they have acquired the character of common sense. Public debate usually concerns whether they are true. Are algorithms genuinely objective? Can AI understand us? Will automation inevitably transform every profession?

Roland Barthes approached questions of this kind differently. Rather than asking whether such statements were true or false, he asked why they appeared so self-evident in the first place. The shift may seem subtle, but it changes the object of inquiry. Instead of evaluating the factual accuracy of particular claims, Barthes directs our attention to the process through which historically contingent interpretations acquire the appearance of natural facts. How do certain ways of describing the world cease to appear as interpretations and come to be accepted as reality itself?

This question lies at the heart of Mythologies, the collection of essays Barthes published in 1957. Its subject was not ancient legend but the ordinary objects and practices of postwar France: wrestling matches, detergent advertisements, photographs, wine, toys, automobiles, tourism, and popular magazines. None of these fascinated Barthes because they were deceptive. On the contrary, they were perfectly ordinary. What interested him was the way they transformed historically specific meanings into facts that appeared timeless, universal, and beyond discussion.

The objects of everyday life have changed considerably since Barthes wrote Mythologies, but the mechanism he described remains strikingly familiar. We no longer inhabit a world structured primarily by magazines, cinema, and television. Increasingly, our experience is organized through recommendation systems, personalized feeds, predictive algorithms, generative models, and conversational AI. These technologies perform practical tasks, but they are also accompanied by narratives that shape how we understand them. They present particular historical developments as though they were the natural direction of technological progress, obscuring the political decisions, economic interests, and cultural assumptions from which they emerged.

Artificial intelligence itself is not a myth. It is a technological development grounded in advances in computing, statistics, and machine learning. The mythology arises elsewhere—in the language through which AI is described, the metaphors used to explain it, the expectations attached to it, and the assumptions that quietly accompany its adoption. As in Barthes' analyses, the object itself is less significant than the meanings that accumulate around it.

Seen from this perspective, the central question is no longer whether AI is intelligent. More revealing is what contemporary culture makes appear natural through AI. Which historical choices disappear from view? Which values become increasingly difficult to question? Which interpretations gradually acquire the authority of common sense? These are, fundamentally, Barthes' questions. More than half a century after Mythologies, they remain remarkably contemporary.

I. Myth Is Not Falsehood

To read artificial intelligence through Barthes requires setting aside one of the most widespread misunderstandings surrounding the word myth. In ordinary language, a myth is usually taken to mean a false belief or an invented story. To expose a myth, therefore, is assumed to be a matter of demonstrating that it is untrue.

Barthes uses the term in a very different sense. Myth is not defined by the truth or falsity of what it says but by the way it produces meaning. It is a particular mode of signification that takes something produced by history—a social practice, a political arrangement, an economic interest, or a cultural convention—and presents it as though it belonged to the natural order of things. History does not disappear completely; rather, it recedes into the background until what is contingent begins to appear necessary and what is culturally produced comes to seem self-evident.

For this reason, myth does not simply deceive. It naturalizes. It transforms historically specific values into apparently universal realities, making them seem timeless precisely by concealing the processes through which they were produced. What has a history comes to appear as though it had none.

Barthes illustrates this mechanism through a series of now famous examples. French wine functions not merely as an alcoholic beverage but as a symbol of national identity. Professional wrestling appears to be an athletic contest, although its deeper significance lies in staging moral conflict, justice, and retribution rather than genuine competition. Perhaps the best-known example is the cover of Paris Match depicting a young Black soldier saluting the French flag. Nothing in the image is fabricated. The photograph records a real event. Yet it quietly communicates a broader message: French colonialism appears harmonious, benevolent, and universally accepted. A complex political history is condensed into an image that presents imperial order as perfectly natural.

This distinction is crucial because it reveals that myth does not primarily replace reality with fiction. It reorganizes meaning. An ordinary sign—a photograph, a word, an object, or a gesture—already possesses a conventional meaning. Myth appropriates that existing sign and transforms it into the signifier of a second-order semiological system. The original meaning remains visible, but it now functions in the service of a broader cultural narrative that presents particular values and historical arrangements as though they belonged to nature itself. Myth, in other words, does not erase the first meaning; it empties it sufficiently for another meaning to inhabit it.

This account of myth is especially relevant to contemporary digital culture because today's myths rarely resemble propaganda in its traditional sense. They do not usually announce themselves through overt ideology or political slogans. Instead, they circulate through interfaces, technical vocabulary, marketing language, and the ordinary interactions that increasingly mediate everyday life. Recommendation systems, conversational agents, predictive models, and search algorithms perform practical functions, but they also become vehicles through which particular assumptions about knowledge, intelligence, creativity, efficiency, and even human agency are quietly naturalized.

This is where Barthes' analysis becomes unexpectedly contemporary. Artificial intelligence is not simply another technological innovation awaiting philosophical evaluation. It is also one of the most powerful contemporary producers of myth, not because the technology itself is fictitious, but because the language surrounding it continually transforms historical choices into natural necessities. The question is therefore not whether AI is real, because it plainly is, the question is what realities AI increasingly encourages us to take for granted.

II. AI Does Not Only Produce Answers. It Produces Myths

Most discussions of artificial intelligence revolve around familiar questions. Is AI accurate? Can it be trusted? Is it creative? Does it threaten employment? These are undoubtedly important issues, but they are not the questions that Barthes would have asked first. Before evaluating what artificial intelligence does, he would have examined how contemporary culture talks about it. Every new technology is accompanied by narratives that shape its public meaning long before its technical capacities are fully understood. AI is no exception. Machine learning models, recommendation systems, and large language models are genuine technological developments, yet they also function as powerful cultural signs. Around them has emerged a vocabulary that increasingly determines not only how these systems are perceived but also how social change itself is imagined.

From a Barthesian perspective, the task is therefore not to determine whether artificial intelligence is mythical. It plainly is not. The more interesting question is what becomes natural through artificial intelligence. Which historical decisions disappear behind apparently neutral descriptions? Which assumptions gradually cease to appear as assumptions? The mythology surrounding AI does not consist of a single grand narrative but of a series of interconnected ways of interpreting technological change. Taken together, they reshape our understanding of history, knowledge, identity, value, and creativity.

Myth One: AI Is Inevitable
History as Destiny

Perhaps the most pervasive mythology surrounding artificial intelligence is the belief that its expansion follows an irresistible historical logic. AI, we are told, will transform education, journalism, medicine, law, scientific research, and the arts. Whether these transformations are welcomed or feared appears almost secondary, since technological progress itself is presented as something that simply unfolds. Society can adapt to it or fall behind, but it cannot alter its direction.

What disappears from this account is precisely the historical dimension that Barthes regarded as essential. Technologies do not emerge independently of human action. They are shaped by political choices, corporate investment, legal frameworks, research priorities, environmental costs, labour relations, and cultural expectations. None of these factors is inevitable, yet the language surrounding AI frequently presents technological development as though it possessed an autonomous momentum. Human agency gradually fades from view, while history begins to resemble destiny.

This mythology does not deny that AI exists. Rather, it transforms one particular trajectory of technological development into the only imaginable future. Once history assumes the appearance of nature, alternative possibilities become increasingly difficult even to conceive. Questions concerning regulation, ownership, democratic oversight, or different models of technological development lose urgency because the future already appears to have been decided.

Myth Two: Algorithms Are Objective

Knowledge Without Interpretation

If the first mythology naturalizes history, the second naturalizes knowledge. Algorithms are frequently described as objective precisely because they calculate rather than judge. Unlike human beings, they are said to possess neither prejudice nor emotion. Their authority appears to derive from mathematics itself.

There is, of course, an important sense in which computational systems can reduce certain forms of human inconsistency. A statistical model does not become ideological simply because it performs calculations. The mythology emerges elsewhere. It begins when the countless human decisions embedded within computational systems disappear behind the apparent neutrality of their outputs. Every dataset reflects prior selections, every model embodies design choices, every optimization procedure privileges particular objectives over others, and every recommendation system presupposes judgments about what deserves to be measured, rewarded, or ignored.

The algorithm consequently comes to appear less as an interpretation of reality than as reality speaking through calculation. Mathematics itself is not being naturalized; a particular conception of mathematics is. What disappears is not human intervention but its visibility. Once again, historical decisions acquire the authority of nature.

Myth Three: The Algorithm Knows Me Better Than I Know Myself

The Computational Self

The mythology of objectivity prepares the ground for a more intimate transformation. If algorithms are assumed to represent reality more accurately than human judgment, it becomes increasingly plausible to imagine that they also represent ourselves more accurately than we do.

This assumption has become woven into the ordinary experience of digital life. Recommendation systems do not simply organize information; they promise to reveal preferences that users themselves have not yet recognized. The remarkable effectiveness of many recommendations encourages a broader conclusion: our tastes, interests, and desires appear increasingly accessible through patterns hidden within data. Identity becomes something to be inferred computationally rather than discovered through reflection, conversation, or lived experience.

Yet recommendation systems do not merely identify preferences; they also participate in their formation. Suggestions repeatedly encountered become increasingly familiar, familiar objects attract further attention, and repeated attention gradually shapes future choices. What appears to be the passive discovery of pre-existing preferences is simultaneously an active process through which preferences themselves are produced. Taste emerges not independently of algorithmic systems but through continuous interaction with them.

The mythology therefore concerns more than recommendation technology. It proposes a new image of the human subject: an individual whose identity appears fundamentally measurable, predictable, and transparent to computational analysis. The algorithm no longer merely knows our preferences; it increasingly defines what it means to have them.

Myth Four: Efficiency Is Always Good

When Optimization Becomes a Value

Once technological systems become accepted as objective interpreters of both reality and ourselves, it becomes increasingly difficult to question the values embedded within their operation. Among these values, none enjoys greater authority than efficiency.

Artificial intelligence is almost invariably introduced through the language of optimization. It promises faster decisions, lower costs, greater productivity, and more efficient forms of organization. Because these aims appear obviously desirable, they often escape philosophical scrutiny. Yet efficiency is never an end in itself. It is meaningful only in relation to the purposes it serves, and those purposes are always open to debate.

Education, for example, cannot be reduced to the efficient transmission of information, just as medicine cannot be exhausted by diagnostic speed, universities cannot be evaluated solely by productivity, and artistic creation cannot be measured only by the rapid generation of finished works. Efficiency becomes mythological when one criterion among many quietly assumes the status of a universal standard. A historically specific conception of progress gradually presents itself as progress itself.

As this conception becomes increasingly taken for granted, alternative values become more difficult to defend. Reflection, ambiguity, contemplation, care, democratic deliberation, and even creativity often appear inefficient precisely because they resist the forms of measurement through which optimization defines success. What disappears is not only alternative practices but alternative ways of imagining what human activities are for.

Myth Five: AI Is Creative

The Reorganization of Creativity

The debate surrounding AI and creativity reveals the cumulative force of these preceding mythologies. Once technological development appears inevitable, algorithms appear objective, human identity becomes computationally legible, and efficiency emerges as the dominant measure of value, the question of creativity is transformed almost without being noticed.

Public discussion usually asks whether artificial intelligence can genuinely create. Barthes would almost certainly have approached the issue differently. Before asking whether machines are creative, he would have asked how creativity itself has come to be understood. Why has a concept traditionally associated with exploration, uncertainty, imagination, experimentation, and expression become increasingly identified with the successful production of artefacts that appear original or aesthetically convincing?

From this perspective, the mythology does not lie primarily in AI's capacity to generate texts or images. It lies in the gradual redefinition of creativity itself. A historically rich and contested concept begins to contract until it appears largely synonymous with the production of outputs. Once this transformation becomes sufficiently familiar, it no longer appears as an interpretation but as common sense.

Whether one ultimately celebrates or resists generative AI is therefore not the deepest philosophical question. More fundamental is the process through which artificial intelligence reorganizes the meanings of concepts such as intelligence, learning, creativity, authorship, and originality. Like the myths Barthes analysed in the 1950s, these transformations do not operate through deception. They operate through naturalization. They quietly reshape the vocabulary through which contemporary society understands itself.

III. When Demystification Becomes Part of the Myth

At first sight, Barthes' project appears straightforward. Modern societies continually transform historical arrangements into natural realities, and the task of the mythologist is to reverse that process by restoring history where myth has made it disappear. To expose a myth is to reveal that what appears inevitable, universal, or self-evident is in fact the product of particular social, political, and cultural circumstances. Demystification thus seems to promise a form of liberation: once the mechanism becomes visible, myth loses its authority.

This was a plausible ambition in the context in which Mythologies was written. Many of the myths Barthes analysed derived part of their effectiveness from their invisibility. They succeeded because they passed unnoticed as common sense. Contemporary algorithmic culture presents a more complicated situation. Artificial intelligence is surrounded by constant discussion, criticism, and explanation. Its limitations are widely publicized. Researchers document bias, journalists investigate the environmental costs of large-scale computation, technology companies publish reports describing the weaknesses of their own systems, and even ordinary users quickly learn that conversational AI can fabricate information, recommendation systems privilege engagement over accuracy, and predictive models are shaped by the data on which they are trained. The mechanisms that sustain these technologies are often more visible than those surrounding many earlier forms of mass culture.

One might therefore expect mythology to weaken as understanding increases. If people know that algorithms are imperfect, if they recognize that datasets reflect human choices and that AI systems are neither neutral nor infallible, should their cultural authority not diminish accordingly? Yet the opposite often seems to occur. Public awareness of these limitations has grown alongside an unprecedented expansion of algorithmic systems into everyday life. We continue to rely on them to organize information, recommend books and films, assist with writing, guide travel, mediate social interaction, and increasingly shape educational and professional practices. Criticism has become widespread, but so has dependence.

This apparent paradox suggests that contemporary myth no longer operates exactly as Barthes described it. It does not necessarily depend on concealment. Instead, it increasingly depends on circulation. A recommendation system need not be perfectly accurate in order to become authoritative. A large language model need not possess genuine understanding in order to become a habitual source of explanation. A personalized feed need not represent reality objectively in order to influence what millions of people encounter every day. Their cultural power derives less from the unquestioning belief they inspire than from the ordinary place they occupy within everyday life.

The distinction is important because it shifts the location of myth. In Mythologies, naturalization often occurred by concealing the historical character of signs. In digital culture, naturalization increasingly occurs through repetition. Each interaction with an algorithm quietly reinforces the expectation that algorithms are the appropriate means through which information should be organized, choices should be guided, or preferences should be interpreted. Familiarity gradually replaces persuasion. What is repeated often enough ceases to demand justification.

The language surrounding artificial intelligence illustrates this process particularly well. Few people today believe that generative models literally think in the same sense that human beings think. Public discussion has become considerably more sophisticated than it was only a few years ago, and distinctions between prediction, reasoning, understanding, and consciousness are regularly debated. Nevertheless, expressions such as assistant, collaborator, creative partner, or even co-author have entered ordinary discourse with remarkable ease. Whether these descriptions are ultimately appropriate is less significant than the cultural work they perform. By circulating repeatedly through journalism, advertising, software interfaces, and everyday conversation, they gradually reorganize the vocabulary through which human and machine activity are understood. What once appeared metaphorical increasingly comes to seem literal.

A similar transformation can be observed in recommendation systems. Most users understand that digital platforms do not reveal timeless truths about music, literature, or cinema. They know that recommendations reflect previous behaviour, commercial objectives, and engagement metrics. Yet these systems increasingly function as cultural intermediaries, influencing which books become visible, which artists gain audiences, which films are discussed, and which voices remain largely unheard. Their authority does not depend upon being mistaken for impartial arbiters of taste. It depends upon becoming the ordinary infrastructure through which cultural experience is organized.

This marks an important difference between many of the myths Barthes analysed and those surrounding artificial intelligence. Contemporary myths often remain effective even after their mechanisms have been publicly exposed. Indeed, criticism can sometimes strengthen rather than weaken their legitimacy. Every controversy surrounding AI reinforces the impression that these technologies have become indispensable. The terms of public debate gradually shift from whether algorithmic systems should occupy such a central position in social life to how they ought to be governed, regulated, or used responsibly. The technology itself increasingly escapes question, while discussion concentrates on its proper management.

History once again begins to resemble nature, although by a different route from the one Barthes originally described. Myth no longer requires ignorance; it requires habituation. Its authority lies not in preventing criticism but in rendering criticism compatible with continued dependence. This may be the most significant transformation that Mythologies undergoes in the age of artificial intelligence. The contemporary myth is not the one that survives because it remains hidden. It is the one that continues to organize our experience even after we have learned to recognize it.

IV. Can We Still Demystify AI?

If contemporary myths increasingly survive their own demystification, what remains of Barthes' project? At first glance, the answer might appear discouraging. If exposing the historical character of artificial intelligence no longer weakens its authority, has criticism itself lost its force?

The question is worth asking because the conditions under which Barthes wrote Mythologies have changed profoundly. In the 1950s, the mythologist could imagine occupying a position outside myth, revealing the historical processes concealed beneath apparently natural realities. The critical gesture consisted in restoring contingency where ideology proclaimed necessity. Once history became visible again, the spell of myth seemed capable of being broken.

Algorithmic culture complicates that picture. Artificial intelligence is not surrounded by silence but by an extraordinary abundance of explanation. Technology companies publish technical papers describing the architecture and limitations of their models. Researchers analyse bias, hallucinations, and environmental costs. Governments debate regulation, journalists investigate corporate practices, and users themselves quickly learn that recommendation systems, search engines, and conversational AI are shaped by probabilities rather than certainty. Knowledge about AI now circulates almost as rapidly as AI itself.

Yet greater transparency has not brought about a corresponding decline in mythology. Understanding how a recommendation system operates does not prevent us from treating its suggestions as meaningful. Knowing that a language model predicts statistically probable sequences of words does not prevent us from speaking as though it possessed intentions, opinions, or understanding. The historical mechanisms become increasingly visible, while the mythology continues to function. Explanation and naturalization no longer exclude one another; they increasingly coexist.

This suggests that the task of criticism must itself be reconsidered. If demystification is understood simply as exposing hidden mechanisms, it will inevitably prove insufficient. Contemporary myths do not derive their authority solely from ignorance. They derive it from the way particular interpretations become woven into the ordinary practices through which we navigate everyday life. They are sustained not because people fail to understand the technology, but because the language surrounding the technology gradually becomes the language through which reality itself is interpreted.

The aim of critique, therefore, cannot be to eliminate myth once and for all. Barthes himself knew that myth was not an accidental distortion of culture but one of its recurring modes of signification. Every society generates narratives through which it renders its institutions, values, and practices intelligible. The mythology surrounding artificial intelligence is neither the first nor the last of these narratives. What distinguishes it is the speed and scale with which digital technologies now reorganize the categories through which we understand ourselves.

For that reason, the most important questions are no longer technological but semiological. When artificial intelligence is described as inevitable, what historical alternatives disappear from view? When algorithms are called objective, what human judgments become invisible? When machines are said to understand, create, or collaborate, how do these words themselves begin to change? The issue is not simply whether such statements are true or false. It is the work they perform within contemporary culture, the assumptions they stabilize, and the forms of life they quietly authorize.

This is why Barthes remains indispensable in the age of artificial intelligence. His enduring contribution lies less in providing a theory of technology than in offering a way of reading the world. He reminds us that language never merely reflects reality. It also organizes it. Every society develops expressions that eventually become so familiar they cease to appear as interpretations at all. The task of criticism is to restore their history—not because doing so will finally free us from myth, but because it preserves the possibility of recognizing that what appears most natural is often the result of human decisions that might, under different circumstances, have been otherwise.

Artificial intelligence will undoubtedly continue to transform education, scientific research, artistic production, communication, and work. Whether those transformations should be welcomed, regulated, or resisted will remain matters of political and ethical debate. Barthes directs our attention to a different, though no less important, question. Before we decide what artificial intelligence ought to become, we should ask how it has already begun to shape the language through which we imagine intelligence, creativity, knowledge, and even ourselves. Those words have a history. Remembering that history may be the first step toward ensuring that our technological future does not come to appear as though it had none.

References

Barthes, R. (1972). Mythologies (A. Lavers, Trans.). Hill and Wang. (Original work published 1957)

Culler, J. (1983). Barthes. Fontana Press.

Saussure, F. de. (2011). Course in General Linguistics (R. Harris, Trans.). Bloomsbury Academic. (Original work published 1916)