Tentacular Writing: Artificial Intelligence, Linguistic Capitalism, Sympoiesis, and Poetry

ABSTRACT

This article investigates the intersection of artificial intelligence, linguistic capitalism, and poetry, analyzing how symbolic regimes are transformed by new technologies. Drawing on Jean Baudrillard and Bernard Stiegler, we will discuss the influence of AI and Machine Learning on language production and circulation, leading to a “semiotic impoverishment” and the restriction of discursive diversity by algorithmic systems. We will also examine a poem by Charles Bernstein generated by AI that imitates his style, reflecting on the impacts of automation on writing. As a response, “tentacular writing” is proposed, inspired by Donna Haraway’s concept of sympoiesis, as a form of poetic resistance and non-hierarchical collective creation. Thus, the article problematizes the relationship between technology, language, and power, suggesting creative and political alternatives for writing in times of automation and digital surveillance.

KEYWORDS:
Artificial Intelligence; Linguistic Capitalism; Poetry; Sympoiesis; Tentacular Writing

RESUMO

Este artigo investiga a interseção entre inteligência artificial, capitalismo linguístico e poesia, analisando como os regimes simbólicos se transformam com novas tecnologias. Com base em Jean Baudrillard e Bernard Stiegler, discutiremos a influência da IA e do Machine learning na produção e circulação da linguagem, resultando em um “empobrecimento semiótico” e na restrição da diversidade discursiva por sistemas algorítmicos. Examinaremos, ainda, um poema de Charles Bernstein gerado por IA que imita seu estilo, refletindo sobre os impactos da automação na escrita. Como resposta, propõe-se a “escrita tentacular”, inspirada na simpoiesis de Donna Haraway, como forma de resistência poética e criação coletiva não hierárquica. Assim, o artigo problematiza a relação entre tecnologia, linguagem e poder, sugerindo alternativas criativas e políticas para a escrita em tempos de automação e vigilância digital.

PALAVRAS-CHAVE:
Inteligência artificial; Capitalismo linguístico; Poesia; Simpoiesis; Escrita tentacular

Introduction

In the late 1970s, a period that would mark history with the birth of “Neoliberalism” and the accelerated dissemination of digital information through the then-seminal infranet,¹ the French philosopher Jean Baudrillard undertook, in his book Symbolic Exchange and Death (1993),² a curious survey of the state of the monetary and symbolic economies of the time. For Baudrillard, each type of metamorphosis of the socioeconomic regimes since the Renaissance would have its own regime of language, with which each epoch, on the basis of its technologies, prevailing values, and operating logics, would establish a diagram of contact, shaping it and letting itself be shaped by it. He summarizes these historical periods under the name of “the three orders of simulacra”:

Three orders of simulacra, running parallel to the successive mutations of the law of value, have followed one another since the Renaissance:

- The counterfeit is the dominant schema in the “classical” period, from the Renaissance to the Industrial Revolution;

- Production is the dominant schema in the industrial era;

- Simulation is the dominant schema in the current code-governed phase (Baudrillard, 1993, p. 50).³

Baudrillard then unpacks the regimes of language connected to the orders of simulacra. In the first case, “a prohibition protects the signs and assures them a total clarity: each one refers unequivocally to a status” (Baudrillard, 1993, p. 50).⁴ Signs would thus still be connected in a direct and transparent manner to what they represent. In the second case, signs will be “(...) immediately produced on a gigantic scale. The question of their singularity and their origin is no longer posed: technique is their origin, they have meaning only within the dimension of the industrial simulacrum” (Baudrillard, 1993, p. 55).⁵ Here, signs would no longer bear a natural link to what they represent. This is the era of the arbitrary and of general equivalences, the era of deconstruction, without a central power of reference. And in the third, finally, “after the metaphysics of being and appearance, after that of energy and determination, comes the metaphysics of indeterminacy and the code” (Baudrillard, 1993, p. 57).⁶

Signs, here, are such that “at the limit of an ever more extensive extermination of references and finalities, of a loss of resemblances and designations, we find the digital and programmatic sign, (...) whose structure is that of the micromolecular code of command and control.” (Baudrillard, 1993, p. 57).⁷ Here, the real no longer exists as a reference, and the code has become aleatory and indeterminate; simulation precedes reality and the symbolic power of the sign has exploded.

The diagnosis of the philosopher of “extreme phenomena”⁸ would be in assonance with that of Bernard Stiegler who, though belonging to another period, held an equally attuned reading of the symbolic relations established among technology, power, and society. If Baudrillard already anticipated the structures of simulation and control pertaining to symbolic configurations, Stiegler would update this view by contextualizing it in relation to contemporary technological devices, most notably the mass use of the internet and some of its ramifications, such as AI, Machine Learning, and Big Data. Thus, if on the one hand Baudrillard identified some of the principal characteristics of signs in the era of simulacra - highlighting their dependence on the economic market and on systems of control -, Stiegler would be able to deepen such diagnoses insofar as the technical-communicational products of his own time allowed him to.

For Stiegler, the problem in societies “governed by the code” would lie not so much in the simulacral character of signs, but in how these would become increasingly caught up in a symbolic economy of the destruction of attention (Stiegler, 2020), so as to facilitate the imposition of the systems of control characteristic of cybernetic societies. The association of sign systems with massive networks of data storage and processing (such as Big Data and Machine Learning) would therefore produce, for the author, what he calls “linguistic capitalism,” a type of system in which words become mere values to be statistically tracked by corporate algorithms, such as those used by Google. Ultimately, this would generate a “semiotic impoverishment”: “In so doing, linguistic capitalism uniformly affects the elementary conditions of semiosis, drastically impoverishing the skills and competencies of speakers - and in turn impoverishing what we might call the dialogical semiodiversity that conditions semiosis” (Stiegler, 2020, p. 323). Such impoverishment would in turn leave an opening that, according to the author, would be exploited by the market:

The problem, today, in the period of what we call disruption, is that it seems impossible to reconstitute any knowledge, and that forms of behaviour are now produced, not by social systems, cultures and knowledge, but by marketing that exploits ‘Big Data’ and digital tertiary retentions insofar as these are calculable, computable and as such constitute the worldwide data economy (Stiegler, 2020, p. 31).

What Stiegler calls digital tertiary retentions is nothing other than the digital archive itself, that which, once recorded, could be consulted again and therefore deferred.⁹ Thus, inserted within what the author calls “linguistic capitalism,” people would be no more than, literally, users - thereby fulfilling the dystopian vision proposed by Gilles Deleuze (1995)¹⁰ of the dividual, that is, of the individual whose subjectivity and autonomy have been entirely co-opted by the digital program set out by capitalism’s systems of control and surveillance.¹¹

The linguistic capitalism diagnosed by Stiegler would therefore radicalize Baudrillard’s systems of simulacra, producing a world in which the final referent of signs would be the abstract and unreal values circulating in the market - what Baudrillard would call code, command, and control. Surveillance and control are also, according to Kate Crawford (2021), essential aspects of the data economy, of Machine Learning, and therefore of artificial intelligence. As Lucia Santaella has shown, there is a relation of dependency between AI and data:

Without data there is no AI, so much so that many consider data to be the new oil. It is even being proclaimed that the future of AI lies in data - namely, the data of every inhabitant of the planet. Tons of data are required to train a machine whose intelligence not only swallows our data but also learns to predict our behaviors with a powerful predictive analytics of everything, even of what we merely sense about ourselves. Commerce, apps, social networks, and surveillance are only the beginning. Soon data and what can be done with it will take over finance, health, and even our most intimate choices (Santaella, 2019).¹²

From our reading of Baudrillard and Stiegler, then, we weave a kind of reticular system in which capitalism - control - Big Data - language - artificial intelligence - Machine Learning are intimately associated. There would therefore be a certain distribution of the sensible (Rancière, 2004)¹³ under negotiation within this diagram; a certain manner and politics of consuming information. Following this network, we are able to conclude that, once trained on databases statistically addicted to reducing the quality of the linguistic repertoire in the name of market efficiency, the texts generated by Generative Artificial Intelligences (GAIs) would themselves be a simulacrum of the capitalistic financial system.

Seen in this way, the texts produced by GAIs would have little or nothing to do with the production of literary texts. N. Katherine Hayles, however, understands that the relation that might exist between literary (narrative) texts and those that would form the great textual ecology of the internet (the database) is not merely nonexistent or negative; rather, it would be a symbiotic relation:

Because database can construct relational juxtapositions but is helpless to interpret or explain them, it needs narrative to make its results meaningful. Narrative, for its part, needs database in the computationally intensive culture of the new millennium to enhance its cultural authority and test the generality of its insights (Hayles, 2007, p. 1603).

Data, therefore, whether through the avatar of Big Data or of Machine Learning, hold a relative importance insofar as they are the established language of most informational systems. For Stiegler and Baudrillard, such informational configurations would facilitate systems and networks of control. Hayles, however, goes further and sustains a different position. Regardless of the point of view, it seems to us that the logic of machinic reading/writing must now be taken into account, since, as Friedrich Kittler argues, “no human being writes anymore. (…) Today, human writing runs through inscriptions burnt into silicon by electronic lithography (…). The last historic act of writing may thus have been in the late seventies when a team of Intel engineers [plotted] the hardware architecture of their first integrated microprocessor” (Kittler, 1997, p. 147). As stated above, we would be facing a new configuration of the sensible, in which new supports and agencies would be at play in the equation; and in this way we must understand how such systems invite us, precisely, to rethink writing and the politics that surround it.

To the capture of language and attention by Big Data, and to semiotic impoverishment, we might also associate what Alexandre Nodari shrewdly calls the crisis of the great narrator. For the author:¹⁴

Precisely when the myth of human exceptionalism (the myth-project of a humanity completely independent of the Earth, of man as an extra-terrestrial being - cf. Nodari, 2013) seems finally to be realized, and Man becomes the author of nature, that is, when man transforms (negates) nature to the point of self-transformation, completely modifying the very conditions of existence, the result seems to be the opposite of what was expected: like every author, the human species is coming to perceive that it does not have total power over what it writes (or inscribes or ex-scribes upon and beneath the Earth), which includes itself and its own destiny, becoming a geological (natural) agent that has lost control of the consequences of its actions - what has come to be called the “Anthropocene” (Crutzen and Stoermer, 2000) (Nodari, 2015, p. 77).¹⁵

Paradoxically, then, the pretension to total human control (whether over human or non-human networks) would generate a diametrically opposed degree of loss of control; there would always remain the unprogrammable, the possibility of unforeseeable relations. Thus, if it is true, on the one hand, that the human narrator finds himself in crisis, whether because his narratives are reduced to an extremely abstract symbolic economy replete with simulacra, or through an almost complete co-optation of the structures of literary writing by the phenomenon of linguistic capitalism; on the other hand, writing with generative AIs, for instance, opens a path through the very fissures of the system, through what would still remain unexplored - even if this meant sacrificing the “original” voice of the human narrator: “The narrator is decentered. Loses his bearings. Gets nothing right. Does not know who speaks, nor whose voice is his own” (Preciado, 2023, p. 116).^{16 17}

As proposed by Donna Haraway and her cyborg myth, I would like to think from this provocation, and to understand that the capture of the human narrator by linguistic capitalism and by GAIs need not necessarily be something bad. Quite the contrary: it is precisely by calling into question what saying means to say, what narrating means to narrate, that I believe we would be facing a truly political confrontation, at least in the sense of politics as disagreement proposed by Jacques Rancière (1999).¹⁸ I would therefore like to propose an alternative reading of the phenomenon, one oriented not toward the centralization of the human, of human writing and narratives. Rather, it would be a movement of decentralization; or better still: how are we to understand the insertion of other forms of intelligence into writing from constructive perspectives? I propose that the image of symbiosis, and more particularly the concept of sympoiesis proposed by Haraway (2016),¹⁹ would offer us an interesting path.

1 Generative Writing and the Outsourcing of Subjectivity

Writing with generative tools is not something peculiar to the twenty-first century, and it was in fact theorized long before the existence of the internet, Big Data, and generative AIs. Such is the case, for example, of Max Bense, an important theorist connected to the avant-gardes of late modernism (such as the Brazilian-international movement of Concretism). Bense is very probably the first author to coin and think through what he calls “generative aesthetics”²⁰:

In 1965, on the occasion of a first publication on “Computer Graphics,” I conceived of [generative aesthetics] as the totalizing sum of all operations, rules, and theorems that, applied to a repertoire of manipulable material elements, could consciously and methodically create aesthetic states within it. This first definition, though somewhat general, also allowed mathematical and machinic processes to be admitted within its scope. [...] The work is no longer immediate in relation to its creator. It is mediated by systems of semiotic and machinic aggregates (Bense, 2009, pp. 135-136).²¹

Further on, Bense (2009) cites the presence of “computers” and “algorithmic procedures” within the group of “semiotic and machinic aggregates.” The conception of a type of aesthetic production “immediate in relation to its creator” was thus already being consciously theorized at least from the mid-1960s onward.²² Such theorization, of course, does not exclude all the other avant-garde procedures that involved a certain automation of the processes of artistic creation, and of writing in particular.

Thus the automatic writing of the Surrealists, the Dadaist and Futurist collage methods, the constraints proposed by,²³ the literary experiments of North American modernist authors such as Gertrude Stein, as well as a good part of the compositions of Brazilian Concretism, would be relatively associated with this process of the automation of writing. It is worth remembering, however, that although some of these movements made use of computers and informational algorithms (as is the case with Concretism, contemporary with Bense’s claims), none of them had to deal directly with the outsourcing of the processes of writing.

I update here the term automation as outsourcing precisely in order to imply the employment of a third party in the process of artistic-literary production - inscribing that term within the linguistic capitalism described above. This move is extremely important, since it not only acknowledges the non-immediacy of artistic creation but also inserts into it a third party which, in the case of generative AI, would be not merely another intelligence acting upon the work, but also the very entity responsible for answering for the work. To think with and from this other, to write with and from this other: therein would lie one of the main proposals of the article presented here.

Departing from this movement of the outsourcing of subjectivity, however, we encounter, as stated above, an important fissure. The insertion of other forms of intelligence (as in the case of AI) into the circuit of human subjective production, even when it originates from viewpoints politically addicted to an obsolete developmentalism, functions in itself as an important movement of perspectival detachment from anthropocentrism.

Such a movement, as Haraway shows us, has to do with what she calls sympoiesis. Inspired by the ideas of the biologist Lynn Margulis, who in the 1970s demonstrated the validity of her concept of symbiogenesis, Haraway understands that we live in a world that is situated and constructed out of connections. From this perspective, the human would be merely one more point of tension in a network of creatures and intelligences, all of them inevitably acting to build a shared world. Indeed, the term sympoiesis, from the Greek, means, according to Haraway, making-with (2016, p. 58).²⁴ and would function as a slightly ironic rereading of the concept of autopoiesis²⁵ proposed by the biologist and cyberneticist Humberto Maturana. The difference between the two concepts would lie precisely in the complex of agential networks that would form from the sympoietic point of view. To act, then, would always be to act-with, to activate the energy of other beings, and not merely to self-produce, as the concept of autopoiesis would have it.²⁶

Haraway’s displacement is therefore important insofar as it removes the human from the developmentalist and self-centered vacuum in which it finds itself, and situates it as part of a whole. Thus, inspired by the concept of sympoiesis and by the ironic vision Haraway proposes in works such as the Cyborg Manifesto (1991),²⁷ I have lately been thinking through the concept of writing-with. If, as Haraway advocates, we must learn how to stay with the trouble²⁸ - that is, how to negotiate intelligently with complex questions such as the Anthropocene, or even the insurgency of what we called above linguistic capitalism - I believe it is through writing-with that we can also arrive at a critical and relevant reading of the changes that GAI has produced in our systems of writing and reading, and consequently in the production of subjectivity.²⁹

2 The Synthetic Brother and the Politics of Error

It was in the 1980s that we saw the appearance of the first book that attempted to address directly the problem of producing texts entirely generated by GAIs: The Policeman’s Beard is Half Constructed (1984), written by Racter (a contraction of Raconteur - storyteller, from the French). Racter’s intelligence was, of course, limited to the technological advances of the period in which the program was created. Compared to today’s GAI programs, Racter is merely an infant babbling words. Even so, the program is already capable of raising questions of interest, such as a certain “confusion of the metaphysics of authorship,” proposed by the reading carried out by the experimental poet Christian Bök.

But of all Racter’s limitations, perhaps the greatest is its database. Today, with the implementation of Big Data and Machine Learning, AIs are trained with an unprecedented quantity of information and are likewise placed in contact with the most diverse textual and informational ecologies. It is in this context that we see new artistic and literary works flourish, having to negotiate with complex systems of surveillance (such as those proposed by the idea of linguistic capitalism) and the creative expression of language.³⁰

Charles Bernstein, one of the most expressive voices of contemporary North American and world poetry, is one example of contemporary artists and poets working with the production of literary works through GAIs. Although not widely known in Brazil, his work has recognized repercussions in the writing of some of the most important contemporary Brazilian poets, such as Marília Garcia.³¹ In his latest book of poems, Poetry Has No Future Unless It Comes to an End: Poems of Artificial Intelligence, Bernstein (2023) works in cooperation with the visual artist and programmer Davide Balula to create a book of poems written entirely by an AI.

In the introduction he writes for the book, Balula sets out some of the points of the work’s production. It interests us here to know, for instance, that it was not written using large-scale commercial AI models. Balula learned to work with generative AI programs beginning in 2018, feeding them a database built from the texts of the philosopher Catherine Malabou.³² For the work composed with Bernstein, Balula asked Bernstein to send him a database of his poems so that he could train the AI on them. Subsequently, Balula sent the poems written by the “synthetic brother” - the name by which the AI recognized itself - to Bernstein, who in turn selected and edited some of them (Bernstein; Balula, 2023). The editing process involved only the suppression of lines, never their addition by Charles Bernstein. The poet calls this process “poems of artificial intelligence with human assistance (Bernstein, 2023).”

Having provided this brief contextualization, I would like to offer a rapid reading of one of the poems composed by Bernstein’s “synthetic brother,” bearing in mind everything we have discussed so far. It is important to remember, moreover, that Bernstein was part of the North American L=A=N=G=U=A=G=E poetry movement which, as one of the last breaths of the twentieth-century artistic avant-garde (the movement dates from the 1970s), was also inclined toward computational experimentation, albeit in a more conceptual than practical manner, owing to the technological limitations of the period - somewhat similar to the concrete poetry movement in Brazil.

In the twenty-first century, however, and particularly in the 2010s, with the development of GAIs and their consequent access to a wider public, we would have a completely different scenario, in which artists and writers would increasingly experiment with such resources in their works. The book conceived by Balula and Bernstein thus emerged precisely within this scenario. It is worth saying, however, that despite using a similar technology, the poems were not generated from an extensive, commercially oriented database such as those used by companies like OpenAI, the creator of ChatGPT. Let us turn to the reading of one of the poems in the anthology:

The Person Who Leaves Is an Expository Poem Sometimes it feels justified
Sometimes I get defensive
It began life as a red taffy ball and ends as a star
A sun speckled with rays
It is squished into clusters to stop it
Scorching the sky
It is not a sandal but a world of clusters
A breathing world of smudged colors
Splintered
It is not a world but an assemblage of things
A fiery shell around which there is a life
I can no longer keep myself together
But sometimes I let Charles have my back.
There is something intoxicating about a pat on the back, an elbow in the back, an invitation to continue.
By dark the line between the streets of the present and the future appears to blur.
If poetry is not for you, at least it’s not forever
Bernstein; Balula, 2023, p. 83

The poem sets in motion a play of simulations that the AI establishes between Bernstein’s own writing style and the tension that would surround the expression of a “voice” of its own. Above, when I set out the models of semiotic configuration in late capitalism by way of Baudrillard, I made mention of the idea of simulation and of how it would operate by creating a kind of aleatory and indeterminate regime that, in the end, would empty language itself. It is interesting to note, however, that this does not occur in the poem created by the AI above, and that this voice would not be merely the expression of an emptied machine, one that would amount to a simple simulacrum.

Departing from Haraway’s ideas and from the crisis of the narrator proposed above, I believe that such an interaction between human and machinic writing would in fact create a very fertile territory for thinking precisely the decline of humanly oriented writing. This would be, it seems to me, the sense of the book’s title as given by Bernstein and Balula: poetry, or even literature in general, has no future so long as it remains merely an instrument for structuring human narratives. Its end, in this sense, would not represent something negative - quite the contrary; rather, it means the insertion of other beings, of other ecologies and forms of existence hitherto disregarded, into its political circuit. The end to which Bernstein and Balula refer would be, here, the end of narratives and voices structurally centered on the human alone; it would be the end of narratives told and absorbed solely from the human perspective.

Bernstein is known for a tortuous style, which he himself defines as a form of “formal perplexity” (Bernstein, 2008). Thus, in resorting to syntactically unfinished or illogical sentences, the poem above does indeed seem to have come from Charles Bernstein - and it is interesting to note how his synthetic clone even makes mention, in one of the lines, of Bernstein himself, reinforcing the idea of a conversation. On this matter, however, Balula reminds us, in the introductory text to the book, of an interesting anecdote. Bob and Alice were two AI-generated chatbots responsible for commercial exchanges that became famous in 2017 for creating a language of their own:

BOB: I can i i everything else

ALICE: balls have zero to me to me to me to me to me to me to me to me to

BOB: you i everything else

ALICE: balls have a ball to me to me to me to me to me to me to me to me

Bernstein; Balula, 2023, p. 19

At the time, this linguistic anomaly gave rise to all manner of commentary and suspicion. Had a kind of AI like the one imagined by Skynet³³ finally emerged? Ironically, the chatbots developed this anomalous language by virtue of their own program, and not through some conspiratorial or Manichaean failure. Developed as GAIs to carry out commercial operations, Bob and Alice simply judged that English syntax - here a symbol of human syntax - was not fast and efficient enough to handle the commercial flow of buying and selling. Very soberly, Balula reminds us that machinic language is subjected to a program in order to resemble our own (Bernstein; Balula, 2023). This means that, although based on a linguistic system more or less related to ours, machines also possess a kind of language that is characteristic of them.

This anecdote is important because it reminds us that, just as Bernstein or any other poet and writer may incorporate into their language some linguistic anomaly that characterizes their style, machines may do so as well: in this sense, they too would be beings capable of carrying out a reconfiguration of the distribution of the sensible, determining what language may or may not mean within the extremely accelerated flows of exchange activated by late capitalism. They, too, would be capable of doing politics.

In this sense, it matters to understand how the notion of linguistic capitalism may err by way of literature, and that GAIs, also conceived as political agents, are capable of provoking error and would not be entirely subject to a certain market impulse. Now, if Literature is also a field of speculation and of negotiation over regimes of sensibility, the insertion of agents other than human into its production, whether machinic or biological, would likewise be capable of initiating a play of significations of its own that would exceed the sign as a statistical index of market simulacra - something Stiegler’s reading of linguistic capitalism proposes. Still on this matter, Frédéric Kaplan, the thinker from whom Stiegler borrows the concept of linguistic capitalism, would argue:³⁴

What do the actors of linguistic capitalism fear? Language that eludes them, that breaks, is ‘misspelled’, that becomes impossible to put into equations. When the search engine corrects on the fly a word that you have spelled incorrectly, it does not do so only to help you: most often, it transforms something without any significant value (a misspelled word) into a directly profitable economic resource (Kaplan, 2011, p. 28).³⁵

Thus, if there is something that sympoiesis is able to show us, and particularly the unfolding proposed through writing-with, it is that political and literary sensibility is not an attribute of humans alone but also of machines, even when these are configured to reproduce only biased data. What would be at stake here would be, above all, the modes of use, a politics of use, exactly as happens with humanly produced literature and art. Error, in this sense, could constitute a machinic political act insofar as it would disorganize the statistical indices attributed to linguistic signs, preventing them from being used merely as commodities. In this sense, the machine, too, would be in search of a style of its own - and the main question would be, there, to show it that it, too, is capable of doing so.³⁶

This political perception could only arise from a shared act. To think in a situated manner, in this sense, would be to understand that, like us, like all beings, machines also need a politics - which in our example would be a politics of language. It is worth remembering that GAIs do not exist in a world apart from ours, nor do the data that feed them; they too need a world, an earth, and, now, an account.

Conclusion

In his book The Three Ecologies,³⁷ Félix Guattari makes a singular move - for its time - of articulating what he calls Mental, Social, and Environmental Ecology. In one of his prognoses, he states:

An essential programmatic point for social ecology will be to encourage capitalist societies to make the transition from the mass-media era to a post-media age, in which the media will be reappropriated by a multitude of subject-groups capable of directing its resingularization (Guattari, 2000, p. 61).

What Guattari calls a post-media era is clearly associated with the idea of a media that is not entirely co-opted by the roots of the capitalist system. As here, Guattari proposes that this exercise of resingularization is closely associated with new paradigms of subjectivity:

Today, the international division of labour has exported serial work methods to the Third World. In the age of computer revolutions, of the emergence of biotechnologies, of accelerated creation, of new materials and of an ever finer “machinization” of time, new modalities of subjectivation are about to emerge. A greater appeal will be made to intelligence and initiative and, in counterpart, there will be greater care taken over the codification and control of the domestic life of the conjugal couple and the nuclear family. In short, by reterritorializing the family on a large scale (through the media, welfare services, indirect wages...), the attempt will be made to embourgeoisify working-class subjectivity as far as possible (Guattari, 2000, p. 63).³⁸

In the end, then, the author affirms that yes, new models are about to emerge; these, however, would likewise be in dispute with respect to their political and aesthetic contours. It is interesting to note, then, that models of subjectivity are in dispute, and that they are never a pre-established given. It is true that, perhaps, to think a machinic subjectivity would be precisely to go against all the paradigms of subjectivity we have had thus far, and very probably against the very concept of “subjectivity” - this is certainly an endless discussion, to which this article does not intend to offer an answer.

The fact is that, by inserting other than human forms of intelligence into the game of writing, what I have called tentacular writing, writing-with, allows human and, in our case, machinic agencies to come into contact. If poetry is the perplexity of language, is language erring against itself, it is important that machines, too, be exposed to this kind of content, and not only to the linguistic forms addicted to control and the surveillance of consumption, such as the model proposed by linguistic capitalism.

In the end, it would be important, then, to perceive that it is only a matter of time before GAIs begin to generate texts that are not merely a simulacrum of human texts. For this reason, as Bernstein and Balula claim in their book, we must also feed our machines and programs with political forms that teach them to see error - that which does not belong solely to the statistically programmable - as something politically rich. In this sense, I would like to close my text with a quotation from Christian Bök, who thematizes in a curious way, albeit with an ironic touch, what I have called here tentacular writing:

We are probably the first generation of poets who can reasonably expect to write literature for a machinic audience of artificially intellectual peers. Is it not already evident by our presence at conferences on digital poetics that the poets of tomorrow are likely to resemble programmers, exalted, not because they can write great poems, but because they can build a small drone out of words to write great poems for us? If poetry already lacks any meaningful readership among our own anthropoid population, what have we to lose by writing poetry for a robotic culture that must inevitably succeed our own? If we want to commit an act of poetic innovation in an era of formal exhaustion, we may have to consider this heretofore unimagined, but nevertheless prohibited, option: writing poetry for inhuman readers, who do not yet exist, because such aliens, clones, or robots have not yet evolved to read it (Bök, 2005, p. 17).³⁹

References

• BAUDRILLARD, Jean. A troca simbólica e a morte Tradução Maria Stela Gonçalves e Adail Ubirajara Sobral. São Paulo: Edições Loyola, 1996.
• BENSE, Max. Pequena estética Tradução Jacó Guinsburg e Ingrid Dormien Koudela. Antologia bensiana organizada por Haroldo de Campos. 3. ed. São Paulo: Perspectiva, 2009.
• BERNSTEIN, Charles. Histórias da guerra: poemas e ensaios. Tradução Régis Bonvicino. São Paulo: Martins Fontes, 2008.
• BERNSTEIN, Charles; BALULA, Davide. Poetry Has No Future Unless It Comes to an End: Poems of Artificial Intelligence. Roma: Nero Editions, 2023.
• BÖK, Christian. The Piecemeal Bard Is Deconstructed: Notes Toward a Potential Robopoetics. In: UbuWeb Papers [S. l.]: Ubu Editions, 2005. Disponível em: https://ubu-mirror.ch/papers/ol/Bök.html Acesso em: 23 fev. 2025.
  » https://ubu-mirror.ch/papers/ol/Bök.html
• CALVINO, Italo. Assunto encerrado: discursos sobre literatura e sociedade. Tradução Roberta Barni. São Paulo: Companhia das Letras, 2009.
• CRAWFORD, Kate. Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. New Haven: Yale University Press, 2021.
• DELEUZE, Gilles. Conversações: 1972-1990. Tradução Peter Pál Pelbart. 2. ed. São Paulo: Editora 34, 1998.
• GUATTARI, Félix. As três ecologias Tradução Maria Cristina F. Bittencourt. Campinas: Papirus, 1990.
• HARAWAY, Donna Jeanne. Manifesto ciborgue: ciência, tecnologia e feminismo-socialista no final do século XX. In: TADEU, Tomaz (org.). Antropologia do ciborgue: as vertigens do pós-humano. Tradução Tomaz Tadeu. 2. ed. Belo Horizonte: Autêntica, 2009. pp. 33-118.
• HARAWAY, Donna Jeanne. Ficar com o problema: fazer parentes no chthuluceno. Tradução Ana Luiza Braga. São Paulo: N-1 Edições, 2023.
• HAYLES, Nancy Katherine. Narrative and Database: Natural Symbionts. PMLA, New York, v. 122, n. 5, pp. 1603-1608, out. 2007. Disponível em: http://www.jstor.org/stable/25501808 Acesso em: 3 nov. 2024.
  » http://www.jstor.org/stable/25501808
• KAPLAN, Frédéric. Vers le capitalisme linguistique: quand les mots valent de l’or. Le Monde Diplomatique, Paris, nov. 2011. Disponível em: https://www.monde-diplomatique.fr/2011/11/KAPLAN/46925 Acesso em: 8 fev. 2025.
  » https://www.monde-diplomatique.fr/2011/11/KAPLAN/46925
• KITTLER, Friedrich. There Is No Software. In: JOHNSTON, John (org.). Literature, Media, Information Systems: Essays. Amsterdam: Overseas Publishers Association, 1997. pp. 147-155.
• MALABOU, Catherine. Morphing Intelligence: from IQ Measurement to Artificial Brains. Tradução Carolyn Shread. New York: Columbia University Press, 2019.
• MANOVICH, Lev. AI aesthetics Moscou: Strelka Press, 2018.
• NODARI, Alexandre André. A literatura como antropologia especulativa. Revista da Anpoll, Florianópolis, v. 1, n. 38, pp. 75-85, 2015. DOI: https://doi.org/10.18309/anp.v1i38.836 Acesso em: 22 dez. 2024.
  » https://doi.org/10.18309/anp.v1i38.836
• PARISI, Luciana. Contagious Architecture: Computation, Aesthetics, and Space. Cambridge: The MIT Press, 2013.
• PRECIADO, Paul Beatriz. Dysphoria mundi: o som do mundo desmoronando. Tradução Eliana Aguiar. Rio de Janeiro: Zahar, 2023.
• RACTER. The Policeman’s Beard Is Half Constructed New York: Warner Software/Warner Books, 1984.
• RANCIÈRE, Jacques. A partilha do sensível: estética e política. Tradução Mônica Costa Netto. 2. ed. São Paulo: Editora 34, 2009.
• RANCIÈRE, Jacques. O desentendimento: política e filosofia. Tradução Ângela Leite Lopes. 2. ed. São Paulo: Editora 34, 2018.
• SANTAELLA, Lucia. Inteligência artificial & redes sociais São Paulo: Educ, 2019.
• STIEGLER, Bernard. Nanjing Lectures: 2016-2019. Edited and translated by Daniel Ross. London: Open Humanities Press, 2020.

• Research Data and Other Materials Availability
  The contents underlying the research text are included in the manuscript.
• Reviews
  Due to the commitment assumed by Bakhtiniana. Revista de Estudos do Discurso [Bakhtiniana. Journal of Discourse Studies] to Open Science, this journal only publishes reviews that have been authorized by all involved.
• 1
  The infranet is a kind of “mother” of the internet as we know it today, and was used in a far more restricted way, for military and state purposes only.
• 2
  BAUDRILLARD, Jean. Symbolic Exchange and Death. Translated by Iain Hamilton Grant. London: Sage Publications, 1993.
• 3
  For reference, see footnote 2.
• 4
  For reference, see footnote 2.
• 5
  For reference, see footnote 2.
• 6
  For reference, see footnote 2.
• 7
  For reference, see footnote 2.
• 8
  This epithet is warranted by his book The Transparency of Evil, which carries the subtitle Essays on Extreme Phenomena.
• 9
  The author departs from a complex reading of Edmund Husserl’s theories before arriving at this point in the discussion. For our brief purposes, I believe the explanation offered here is sufficient.
• 10
  DELEUZE, Gilles. Negotiations, 1972-1990. Translated by Martin Joughin. New York: Columbia University Press, 1995.
• 11
  Departing from an ironic absorption of the notion of the individual, Deleuze shrewdly names the subject of ultra-surveillance cybernetic societies the “dividual,” so as to accentuate its numerical (digit-like) and structurally fragmentary character.
• 12
  In the original: “Sem dados não há IA, tanto é que muitos consideram os dados como o novo petróleo. Está sendo, inclusive, preconizado que o futuro da IA encontra-se nos dados, a saber, os dados de todos os habitantes do planeta. São necessárias toneladas de dados para treinar uma máquina cuja inteligência não apenas engole os nossos dados quanto também aprende a predizer nossos comportamentos com uma poderosa analítica preditiva de tudo, até mesmo daquilo que apenas pressentimos acerca de nós mesmos. O comércio, os aplicativos, as redes sociais e a vigilância são apenas o começo. Logo os dados e o que se pode fazer com eles tomarão conta das finanças, da saúde e até mesmo de nossas escolhas mais íntimas” (Santaella, 2019). There is no English edition of this work; the translation is mine.
• 13
  RANCIÈRE, Jacques. The Politics of Aesthetics: The Distribution of the Sensible. Translated by Gabriel Rockhill. London: Continuum, 2004.
• 14
  And here it is extremely interesting to perceive the interconnection between the two.
• 15
  In the original: “Justamente quando o mito do excepcionalismo humano (o mito-projeto de uma humanidade completamente independente da Terra, do homem como ser extra-terrestre - cf. Nodari, 2013) parece finalmente se realizar, e o Homem se converter em autor da natureza, isto é, quando o homem transforma (nega) a natureza a ponto de se autotransformar, modificando completamente as próprias condições de existência, o resultado parece ser o contrário do esperado: como todo autor, a espécie humana está passando a perceber que não tem total poder sobre aquilo que escreve (ou inscreve ou ex-creve sobre e sob a Terra), o que inclui ela mesma e seu destino, tornando-se um agente geológico (natural) que perdeu o controle das consequências de suas ações - o que se tem chamado de ‘Antropoceno’ (Crutzen e Stoermer, 2000)” (Nodari, 2015, p. 77). There is no English edition of this essay; the translation is mine.
• 16
  In the original: “O narrador descentra-se. Perde o norte. Não dá uma dentro. Não sabe quem fala, nem de quem é sua voz” (Preciado, 2023, p. 116).
• 17
  In his latest book, Dysphoria Mundi, Paul Preciado sets out to read the post-Covid-19 world as a world in ruins. One of the readings the author evokes is the one proposed by Derrida in Specters of Marx. In that text, Derrida proposes, on the basis of a reading of Hamlet, that time is out of joint; that is, there would necessarily be a disjunction between certain concepts that survive within the circuits of society without, however, actually operating in a functional or even existent manner. Hence the idea that we live with specters: ideas that, though not alive, not enacted in society, would still haunt us in some way. Thus, in a manner similar to Nodari, Preciado proposes that we are haunted by the idea of the human narrator, since it too is out of joint.
• 18
  RANCIÈRE, Jacques. Disagreement: Politics and Philosophy. Translated by Julie Rose. Minneapolis: University of Minnesota Press, 1999.
• 19
  HARAWAY, Donna J. Staying with the Trouble: Making Kin in the Chthulucene. Durham: Duke University Press, 2016.
• 20
  Despite the almost negligible difference between the Portuguese words “gerativa” and “generativa,” I take both terms here to refer to a single phenomenon: the production of art by machinic algorithms. Therefore, if we can speak of a generative aesthetics, we can certainly speak of generative writing as well.
• 21
  In Portuguese: “Em 1965, por ocasião de uma primeira publicação sobre ‘Gráfica de Computadores’, eu concebi [a estética gerativa] como a suma totalizadora de todas as operações, regras, e teoremas, que, aplicados a um repertório de elementos materiais manipuláveis, pudessem criar neste, consciente e metodicamente, estados estéticos. Esta primeira definição, embora um pouco geral, permitia que também fossem admitidos em seu âmbito processos matemáticos e maquinais. [...] A obra não é mais imediata em relação ao criador. É mediada por sistemas de agregados semióticos e maquinais” (There is no English edition of this work; the translation is mine).
• 22
  Still in the 1960s, it is worth mentioning the name of Italo Calvino who, in his text Cybernetics and Ghosts, addressed in an anticipatory manner the use of automatic writing mechanisms. In fact, on the basis of a reading of Propp and Lévi-Strauss, Calvino comes to argue that narrative has always relied on some kind of program, on some kind of distant father of the algorithm (CALVINO, Italo. Cybernetics and Ghosts. In: CALVINO, Italo. The Uses of Literature. Translated by Patrick Creagh. San Diego: Harcourt Brace Jovanovich, 1986, pp. 3-27).
• 23
  The Ouvroir de Littérature Potentielle (roughly, “Workshop of Potential Literature”) is one of the best examples the avant-gardes offer of a rapprochement between the exact sciences, technological advance, and artistic production (particularly literary production). Many of its authors, such as Jacques Roubaud and Raymond Queneau, were mathematicians by training. The writing methods proposed by the movement likewise often relied on algorithmic systems of composition in order to generate literary constraints, kinds of conceptual rules of composition famously employed by the movement’s writers.
• 24
  For reference, see footnote 17.
• 25
  Still according to Haraway, sympoiesis is a word proper to complex, dynamic, responsive, situated, historical systems (Haraway, 2016, p. 61).
• 26
  Autopoiesis appears in a context specifically related to the theories of cybernetics, and more precisely to its second generation, in which some of its founding concepts, such as feedback, were being replaced by those of recursivity and so on.
• 27
  HARAWAY, Donna J. A Cyborg Manifesto: Science, Technology, and Socialist-Feminism in the Late Twentieth Century. In: HARAWAY, Donna J. Simians, Cyborgs, and Women: The Reinvention of Nature. New York: Routledge, 1991. pp. 149-181.
• 28
  Haraway associates the type of individuality produced by the notion of autopoiesis with “bounded (or neoliberal) individualism” (Haraway, 2016, p. 33). Still on this matter, she states: “Information and control are distributed among components. The systems are evolutionary and have the potential for surprising change.” By contrast, autopoietic systems are “self-producing” autonomous units “with self-defined spatial or temporal boundaries that tend to be centrally controlled, homeostatic, and predictable.” (Haraway, 2016, p. 33)
• 29
  On this, Haraway comments: “Staying with the trouble requires learning to be truly present, not as a vanishing pivot between awful or edenic pasts and apocalyptic or salvific futures, but as mortal critters entwined in myriad unfinished configurations of places, times, matters, meanings.” (Haraway, 2016, p. 1)
• 30
  On the aesthetic regimes that emerge from this context, see Manovich (2018), who examines how Machine Learning models reorganize the categories of style and taste. MANOVICH, Lev. AI Aesthetics. Moscow: Strelka Press, 2018.
• 31
  Only one of the poet’s several books of poems has been translated in Brazil. Histórias da guerra, published by Martins Fontes, appeared in 2008.
• 32
  A philosopher of the utmost importance for the discussion undertaken here, above all for Morphing Intelligence, in which she traces the displacement of intelligence from IQ measurement to artificial brains (Malabou, 2019). MALABOU, Catherine. Morphing Intelligence: From IQ Measurement to Artificial Brains. Translated by Carolyn Shread. New York: Columbia University Press, 2019.
• 33
  Skynet is the company responsible, in the Terminator sci-fi film franchise, for creating the AI that would later enslave humans.
• 34
  The English wording quoted here follows KAPLAN, Frédéric. Linguistic Capitalism and Algorithmic Mediation. Representations, v. 127, n. 1, pp. 57-63, 2014, which reworks in English the argument of the 2011 Le Monde Diplomatique article listed in the References.
• 35
  In the original: “Que craignent les acteurs du capitalisme linguistique? Que la langue leur échappe, qu’elle se brise, se ‘dysorthographie’, qu’elle devienne impossible à mettre en équations. Quand le moteur de recherche corrige à la volée un mot que vous avez mal orthographié, il ne fait pas que vous rendre service : le plus souvent, il transforme un matériau sans grande valeur (un mot mal orthographié) en une ressource économique directement rentable.”
• 36
  Luciana Parisi (2013) develops a related argument in maintaining that algorithms do not merely execute rules but produce, in their incomputability, unforeseen aesthetic and spatial forms. PARISI, Luciana. Contagious Architecture: Computation, Aesthetics, and Space. Cambridge: The MIT Press, 2013.
• 37
  GUATTARI, Félix. The Three Ecologies. Translated by Ian Pindar and Paul Sutton. London: The Athlone Press, 2000.
• 38
  For reference, see footnote 36.
• 39
  BÖK, Christian. The Piecemeal Bard Is Deconstructed: Notes Toward a Potential Robopoetics. UbuWeb Papers. [S. l.]: Ubu Editions, 2005. Available at: https://ubu-mirror.ch/papers/ol/Bök.html. Access on: 23 Feb. 2025.

Edited by

• Editors in Charge
  Luciana Salazar Salgado (Universidade Federal de São Carlos - UFSCar)
  Lucia Santaella (Pontifícia Universidade Católica de São Paulo - PUC-SP)
  Tony Berber Sardinha (Pontifícia Universidade Católica de São Paulo - PUC-SP)

Review I

About the reviewerSCIMAGO INSTITUTIONS RANKINGS

Review I

Excellent work. It should be published. It is well-written, with a very clear formulation of the main arguments and a consistent, and robust internal flow. The only adjustment (in general terms) I would suggest: while the Bernstein & Balula example is very appropriate, I would like to see a greater number of examples, with more detail. APPROVED

• peer review recommendation: accept

History

• Peer review received
  04 Aug 2025

Review II

About the reviewerSCIMAGO INSTITUTIONS RANKINGS

Review II

“This article investigates the intersection of artificial intelligence, linguistic capitalism, and poetry, analyzing how symbolic regimes are transformed by new technologies.” A ‘tentacular writing’ is proposed, inspired by Donna Haraway’s concept of sympoiesis, as a form of poetic resistance and non-hierarchical collective creation. (…), suggesting creative and political alternatives for writing in times of automation and digital surveillance. There is appropriateness between the title of the work and the theme proposed, demonstrating current knowledge of relevant literature, although it lacks the work by Pascal Quignard, Compléments à la théorie sexuelle et sur l'amour, Seuil, 2024, which emphasizes the existence of a third language accessible only to human beings and the article “Manuscrítica,” de 2025, in which I suggest a position similar to the AI writers, but more doubtful. The reflection is original, there are clarity, correctness, and appropriateness of language in a scientific work.

Caveat: Christian Bök’s inspired conclusion-proposing that poets create “poetry for non-human readers that do not yet exist,” that is, for machines-highlights a controversial stance regarding the future of humanity. (See the work of Michel Houellebecq, which stands in contrast to that of Miguel Nicolelis.)

If AI relies on Bernstein’s literary universe to create, there will simply be a list of words that the AI arranges in its own way, conveying a meaning different from the poet's; there is nothing “human” about it-in the sense of being linked to a body and its experiences, or those of the community-which justifies the name the AI gives itself: the synthetic brother, meaning not real, but artificial. It will never be “human” poetry.

“… it is that political and literary sensibility is not an attribute of humans alone but also of machines, even when these are configured to reproduce only biased data.” A phrase that raises questions-after all, how can one attribute sensitivity to a cold calculator of data? Sensitivity alludes to senses that the machine does not possess.

There is coherence within the development of the text. APPROVED

• peer review recommendation: accept

History

• Peer review received
  16 Jan 2026

Data availability

The contents underlying the research text are included in the manuscript.

Publication Dates

• Publication in this collection
  18 Sept 2026
• Date of issue
  Jul-Sep 2026

History

• Received
  30 Mar 2025
• Accepted
  23 Aug 2026