AI and art: rethinking the boundaries

The author holds that the sudden irruption of Artificial Intelligence into the realm of art compels us to urgently revisit and revise many theoretical assumptions we had, until recently, taken for granted.

The comfortable, widely shared prejudice that allowed us to draw a firm and impassable line between what a machine is capable of producing and what an artist creates now risks crumbling — confronted as we are with machines able to generate meaningful output through operations of selection and combination of natural linguistic codes, drawing on a virtually boundless repository of cultural memory.

We are no longer dealing with a process of mere deterministic execution of finite, unambiguous instructions driven by human agency. What we face instead is a system of advanced repertoire processing whose rapid succession, selection, and recombination yields unpredictable results — results that those who rely on such a system can only evaluate after the fact, never before. The veto remains their sole prerogative. It is, for instance, widely acknowledged that Artificial Intelligence is prone to what are commonly termed hallucinations.

The hallucination of an artificial intelligence is comparable to a lie. And semiotics teaches us that anything used to deceive constitutes a language — it belongs to the domain of meaning, of signification, of the human. A human presence persists behind these semiotic processes, upstream or downstream, yet never in full command of them. This metalanguage, precisely by virtue of its intrinsic ambiguity, cannot be excluded from serious consideration in artistic production. Nor does the fact that such ambiguity arises from a stochastic process diminish its fascination or its force. To master the competencies required to navigate that ambiguity is, ultimately, the defining ambition of the digital artist.

A poem generated by an AI bot is often comparable to the work of an unskilled poet in their early attempts: both deploy, clumsily, a limited repertoire of commonplaces, yielding mediocre results. But if an AI can draw on millions of repertoires — and if one knows how to use it — the outcome can shift in register, depth, and intrinsic value.

More than that: through self-improving learning, AI systems can be trained to generate meaning in accordance with the particular idiolect (1) of their user. The artistic idiolect continuously violates its own rules in order to produce surprise. It functions as a genuine algorithm that learns and modifies itself with each new work. That is precisely what AI makes it possible to govern.

Art, historically, begins at the moment it renounces the representation of reality as it is, in order to become an individual aesthetic algorithm: reality as rendered by the artist. This is precisely the operation that AI systems are learning to manipulate with remarkable fluency. Beginning with the decoding of the specific idiolects of individual authors, art criticism rises to the status of a discipline capable of interpreting mechanisms that the author often applies unconsciously, automatically — much as artificial intelligence systems do.

We have moved swiftly past a period in which the opposition between human and artificial retained any coherent meaning.

For semiotics, the idiolect stands as evidence that the general code is not rigid, but is continuously resignified and personalised by the individual — or by the work of art itself.

At this point, a distinction is worth drawing:

The computational algorithm tends toward standardisation and repeatability: given the same input and the same rules, it yields the same output. The artist's personal algorithm — their idiolect — is, by contrast, oriented toward ambiguity and self-reflexivity. The artistic idiolect continuously violates its own rules in order to produce surprise. It is an algorithm that learns and modifies itself with each new work, introducing what semioticians call productive noise, or excess of meaning.

The digital artist, working with AI, wages a struggle against banality, predictability, and the complacency of a product designed to flatter the tastes of its audience. It is an extraordinarily difficult undertaking, because artificial intelligence generates responses calibrated to be accepted by the majority of users — a marketer's paradise, for such is the logic of the market.

To instruct an artificial intelligence to produce meaning without lapsing into the banal or the predictable is a genuinely complex task, and one that I consider to be precisely the mission of the new digital artist.

From the eighteenth century onward, art — literature, painting, music — has waged a social conflict between the patron or commissioner, who deferred to public taste, and the artist, who chafed against the social conventions to which they were expected to submit. The mass media era sharpened this conflict further: producers enlisted artists to sell products tailored to the appetites of the broad public, rewarding them in turn with extravagant remuneration, fame, and success.

Digital society shifts the control of patronage away from publishing houses, record labels, and galleries, transferring it to the owners of the great distribution platforms — Amazon, Spotify, ArtStation — which extract profit from direct sales conducted through their own channels. The cardinal role of opinion-makers, art critics, gallerists, and talent scouts is bypassed, giving rise to a new landscape in which the author, for all its difficulties, confronts the judgement of the public directly, without mediation.

Yet this new landscape carries certain advantages: it emancipates the artist as a self-producer. Their vision of art can travel through the network, reach its audience at accessible price points, and yield a degree of personal economic autonomy. Who determined that art must necessarily be a commodity reserved for wealthy connoisseurs?

The last barrier to fall in the new era of digital art is intimately bound up with the artisanal dimension of the artistic object. The artist who seeks to express their worldview through digital tools relinquishes the role of mere manual craftsperson. They become conceptual. They evolve into the role of cultural operator — and frequently require no critical endorsement to gain recognition.

The artwork, irreproducible at scale, contributes to the inflation of its own value — a value derived from scarcity and uniqueness, the benefits of which accrue only marginally to the author, serving instead to sustain the industry of legal rights, generally to the advantage of the network of producers and distributors.

The renunciation of scarcity value — of the uniqueness of the singular object — entails a reappraisal of the intrinsic value of the work as such. Digital art can be meaningful only if it abandons any pretension to court public taste. It is art when it expresses resistance, opposition, and an alternative to the received sense of the beautiful and the pleasant.

Digital art will earn that name if it proves capable of expressing balance, harmony, and conceptual depth through ambiguous, dissonant messages — messages remote from the common sense of the agreeable. It may be surprising, shocking, unsettling, politically incorrect.

And the artist who manages to achieve this by employing systems that are, by their very nature, calibrated to accommodate the preferences of the broad public — as AI indeed is — will secure a place of genuine standing in the cultural landscape of the years ahead.

──────────────────────────────────────────

(1) THE IDIOLECT IN UMBERTO ECO'S SEMIOTICS

In semiotics, the aesthetic idiolect is a rule governing the manipulation of existing codes. In computational terms, it might be defined as a meta-algorithm or a personalised function. The artist does not invent a language from scratch, but writes a personal script that selects, combines, and deforms the rules of the common language in a systematic and recognisable way — what we call style. As Umberto Eco argued, the aesthetic idiolect does not simply use the code: it establishes a new, provisional code operative within the singular work.