Introduction

In this paper, one explores the role of the translator in the digital age with a particular focus on the posthumanist tradition and how the latter can inform the discussion regarding the translator’s identity. Posthumanism is used as a broad theoretical framework to form the basis for the discussion that follows in the case study. The reasons for this are presented in the next sections, where it becomes clear that the breadth of the posthumanist theory makes it difficult to present it as a strict theoretical model. A case study follows in which it is attempted to compare ChatGPT’s translation outcomes with those of a human translator in literary texts. Kavafy’s poems are used for this analysis, as demonstrated below. The choice of these poems aims to reveal the difficulties that ChatGPT translation faces when it comes to imitating the style of the poet. What is more, a comparison between the style of Kavafy and Walt Whitman is undertaken via the ChatGPT translation to see how it can imitate the style of the poet and implement it in a different situation. Finally, the conclusions that can be drawn from this comparison are presented along with recommendations for future studies. For instance, a comparison between ChatGPT’s translations of Kavafy’s poems using Whitman’s style and those of a native speaker could further inform the findings of the present study.   

Literature review

Posthumanism and Transhumanism

Various studies have revealed the difficulties regarding a simple definition of the term posthumanism (e.g. Bignall & Braidotti, 2019; Braidotti, 2016). As Rosi Braidotti puts it, the “posthuman predicament… …implies both a chronological and a conceptual aspect” (Braidotti, 2010, p. 69). The chronological aspect is evident from the first compound of the term, “post”, which implies that this theory comes after and goes beyond humanism. A way towards a simple understanding of the posthuman turn lies actually in its distancing from putting the human being in the center of the debate, as was the case for the humanist movement. However, this is way too simplistic. Humanism also presupposes a certain social structure, a cult, and an admiration for antiquity (KRISTELLER, 1978), among others, whereas posthumanism tries to not only distance itself from such things but to deconstruct them. This is the “conceptual aspect” that Braidotti talks about (Braidotti, 2010, p. 69). Posthumanism is a critical theory that draws upon the neo-materialism of Deleuze, the thought of Derrida, Haraway’s manifesto, feminist epistemologies and more (Bignall & Braidotti, 2019; Badmington, 2010; Hobden, 2013). It attempts to deconstruct the Eurocentric, masculine perceptions of traditional philosophy and in the process create new modes of thinking for an epistemology of life where one can think through new epistemological frames, such as the intraspecies frame, the techno-mediated frame, and more non anthropocentric frames (Bignall & Braidotti, 2019; Ferrando, 2013; Herbrechter, 2013; Nayar, 2018; Wolfe, 2010). Posthumanism tries to explore the boundaries between the human and the world, i.e., non-human animals, nature, technology, and remove those boundaries towards a relational ontology where the center of attention is not the human being but there are many equally important centers (Braidotti, 2016).  

            Of particular interest for this paper is posthumanism’s techno-mediated epistemological frame. As Francesca Ferrando states, if there is a common interest in technology that posthumanism and transhumanism share “the ways in which they reflect upon this notion is structurally different” (Ferrando, 2013). Posthumanism does not turn technology into its main focus, falling into techno-reductionism, but tries to put into perspective the historical and ontological dimensions of it. To understand technology from a posthuman epistemological frame, Donna Haraway’s and Martin Heidegger’s modes of thinking on the matter should be taken into account. The former, in her A Cyborg Manifesto, attempts to deconstruct the dualistic frameworks and eliminate the strict boundaries that separate mind and body, the organism and technology (Haraway, 2010). The latter views technology as a way of revealing, a way to reestablish the lens through which human beings see the world (Heidegger, 1977). Technology, therefore, reshapes the relationship between the human being and its milieu, its environment, the world. One common denominator between posthumanism and transhumanism is “the notion of technogenesis”, as stated by Ferrando (2013). She continues by saying that for both theories technology appears to be “a trait of the human outfit” (Ferrando, 2013). In both cases, it is more than a functional tool, it is an ontological aspect of the human experience. Posthumanism views technology through the mediation of feminist epistemological frames, among others, erasing the boundaries between technology and self. Adding to this, Foucault’s technologies of self should be considered, as the philosopher tries to create a relational ontology by dismantling the separation between the self and technology (Foucault et al., 1988). In short, posthumanism is a praxis, where the separation between the imagined futures and their enactments does not exist (Ferrando, 2013). In transhumanism, on the other hand, such a separation between the imagined futures and the current situation is evident, as the mandates to use technology in order to transcend the present human experience, to enhance our physical bodies and to extend our capabilities comprise some of the basic tenets of this theory (Fukuyama, 2004).   

The evolving role of the translator 

Nowadays, the role of the translator seems to have evolved from the traditional translator to a techno-savant translator (Massey et al., 2022). In this simple schema, the traditional translator had to be an expert on both languages, the source and the target language, and be well-versed in the translation process. The modern or perhaps the post-modern version of the translator needs to be an expert on CAT tools and, generally, on the technological tools that need to be implemented in their work. The post-modern version may even try to deconstruct the concepts related to the traditional translator, such as the attainment of expert level in both languages. This is where advancements in machine translation come into play with the increased machine learning capabilities of the systems. In the age of ChatGPT 4.0 it seems that the human translator can be reduced not only to a post-editor but to a mere prompting programmer. Transhumanist mandates to use technology as a way to transcend the human condition and liberate us from the constraints of biology (Fukuyama, 2004; Cronin, 2020; O’Thomas, 2017) could inform this tech-enthusiastic, albeit grim, image of the new translator. In its extreme, this is the transhuman translator that uses technological tools to liberate themself from the constraints of their own biology, the lack of language expertise for instance. This extreme techno-enthusiasm can spur from the impressive capabilities of new technologies, such as LLMs like ChatGPT 4.0, but these cannot replace the human translator as shall be demonstrated in the case study of this paper. Despite the fact that a recent study shows that ChatGPT translations outperform manual human translation, especially in terms of word variety (Deng, 2024), there are still issues related to the performance of ChatGPT in translating literature and, of course, with the question of who produces the data the ChatGPT uses to learn, that is the human agent behind the machine.

               The posthumanism framework, on the other hand, can provide a more realistic approach to the evolution of the human translator in the age of AI. It calls for a change in the translator’s identity, presenting a relational identity that signifies the interconnectedness of the human agent and the technological objects at hand (Lee, 2023; Fenoulhet, 2020). It would go even beyond that to assign agency to non-human agents, such as ChatGPT, so that the decentralization of the point of interest is complete. However dubious this might seem; it is a notion that should not be a priori dismissed. Nevertheless, it is important to state that the posthumanism framework works just fine in the case of the translator without reaching the point of distributed agency. In other words, the “ego” of the translator needs not be dispersed, as Lee (2023) points out. In the relational ontology that posthumanism presents, the will of the translator arises from various centers of interest. For instance, this “will” can be revealed through the relationship between the human agent that wants to perform the translation, ChatGPT 4.0, and the other human agents that inform the workings of the LLM, by providing the data for example. It is a coproduction. One might say that this is coproduction of various agents, which is actually accurate, or a dispersal of human agency, but, still, these are anthropocentric views that one implements. Instead, the role of the translator in the new technological landscape leaves them situated in a technological milieu where the various interactions between them and the machine shape their new relational ontological identity.

            The posthuman translator interacts with ChatGPT in ways that create a relationship between them and the machine. The translator is not merely a prompting programmer or a post-editor. There is a relationship that is produced via the various interactions between them and, of course, via trial and error. A relationship that was only possible to occur among humans up to now. However, this new relationship should not be viewed as a manifestation of the human agency in different forms, rather as a revealing of the new ontological identity of the translator. This revealing is only possible through the extension of the human ontology via technology, an extension that is capable of revealing a new relational identity, the posthuman translator.

Methodology

Following the brief presentation of the theoretical framework of this study that is presented above, the focus turns to literary translation and the challenges that AI faces regarding the translation of this type of linguistic corpora. In the case study, the performance of ChatGPT 4.0 in literary translation is measured and analyzed, compared to human translations. Even though further studies could inform this measurement further using specific metrics, in this case a qualitative analysis of the findings is performed. The findings are informed by the theoretical framework of posthumanism that was presented in the relevant sections of this paper to create an image of the persona that is called the posthuman translator and how this posthuman translator works in the field of literary translation.   

Challenges of AI in Literary Translation

Despite the rapid advancements in machine translation, particularly with the development of sophisticated models like ChatGPT 4.0, AI still struggles with several key aspects unique to literary texts. One of the most significant challenges lies in maintaining style and tone. The idiosyncrasies of an author’s voice, which often carry the emotional depth, irony, or subtle shifts in mood that define literary works, are difficult for AI to replicate accurately. AI-generated translations tend to flatten these nuances, resulting in a loss of the original’s rich tonal variations (Venuti, 1995). This limitation is especially pronounced when translating complex authors like Cavafy, whose work often hinges on understated irony and emotional restraint.

Another persistent issue is preserving cultural context. Literary texts are not just linguistic entities; they are embedded with cultural references that extend beyond the literal meaning of words. Translating these references requires an understanding of the historical and cultural connotations tied to specific phrases or allusions, something that current AI models like ChatGPT-4 are not fully equipped to handle (Leppihalme, 1997). For instance, in Cavafy’s poetry, historical events are often used metaphorically to express personal or collective emotions, which an AI might interpret too literally, losing the cultural and emotional resonance.

In addition to this, AI faces difficulties in handling ambiguity and subtext, both of which are crucial elements in poetry. Ambiguity, whether intentional or not, adds depth to literary works by allowing multiple interpretations. However, AI models tend to oversimplify ambiguous or layered meanings, defaulting to a single, more literal interpretation that can strip the original of its intended complexity (Munday, 2009). In Cavafy’s case, his use of historical events as metaphors for personal longing may go unrecognized by the AI model, leading to translations that, while grammatically correct, fail to convey the intended emotional or symbolic significance.

Furthermore, the challenge of recreating poetic form is another area where AI-generated translations often fall short. Poetry relies heavily on structure—rhyme, meter, line breaks, and other formal elements all contribute to the overall effect of the poem. However, AI models do not consistently preserve these structures, which can significantly alter the reader’s experience of the text (Köbis & Mossink, 2021). The translation of Cavafy’s poetry, which is rich in its use of form to convey meaning, often suffers when these elements are not adequately retained.

Lastly, rendering emotional and figurative language presents a formidable challenge for AI translation. Emotive and figurative expressions—metaphors, similes, and imagery—are deeply intertwined with the emotional subtext of a poem. Translating these requires more than lexical accuracy; it demands an understanding of the emotional and symbolic layers underlying the text (Shuttleworth, 2014). ChatGPT-4, while capable of generating grammatically sound translations, often fails to grasp these deeper meanings, resulting in translations that may be technically correct but emotionally shallow.

In sum, while AI has made remarkable progress in fields such as technical translation, it still falls short in the realm of literary translation, where style, tone, cultural context, ambiguity, poetic form, and emotional depth are paramount. These limitations underscore the need for human intervention in literary translation, where the translator's role is not merely to convert words from one language to another but to navigate the intricate web of meaning that makes literary texts so rich and multi-layered.

Methodology of the case study

This qualitative case study utilizes a comparative analysis approach to examine six poems generated by ChatGPT-4, alongside the prompts used for generation and the human translations of Cavafy’s poetry by Keeley and Sherrard (1992). The objective is to discern how well AI-generated works capture themes, stylistic choices, and literary devices present in the original texts. The study is also designed to test the AI’s capacity to emulate the poetic styles of established poets, such as Walt Whitman, further probing the machine’s potential to perform creative tasks typically reserved for human translators.

Sample and Prompting

The study uses a purposive sampling method (Patton, 1990) to select six poems from Cavafy, focusing on those that display thematic complexity—particularly in Cavafy's use of irony, historical references, and emotional restraint. Each selected poem was translated into English using custom prompts designed to guide ChatGPT-4 toward capturing Cavafy’s unique voice. In parallel, prompts were crafted to instruct the model to emulate Whitman’s poetic style, using three well-known poems of Whitman, ‘When I heard the Learn’d Astronomer’, ‘Time to Come’ and ‘O Me! O Life’, and applying a free-verse structure and more expansive imagery to Cavafy's translations.

Research Procedure

The research is divided into three stages:

• Poem Generation: Using ChatGPT-4, six poems by Cavafy were translated into English based on three sets of carefully designed prompts that aimed to evoke the stylistic and thematic complexity of the original texts.
• Comparative Analysis: The generated poems were compared to the human translations by Keeley and Sherrard (1992). The comparative analysis focuses on several aspects, including thematic coherence, stylistic fidelity, and the preservation of cultural references. Both literal and non-literal translation choices are examined in detail.
• Style and Theme Emulation Testing: ChatGPT-4 was also tasked with emulating the style of Walt Whitman, to evaluate the AI’s ability to adapt to different poetic voices and thematic structures. The goal here was to determine whether AI could transcend simple translation and actively engage with the literary essence of different poetic traditions.

Results & Analysis

‘Prompting’ Before ‘Acting’

Before proceeding with the analysis of the poems, we carefully crafted and inserted a series of prompts into ChatGPT-4, with the aim of generating translations that aligned with the thematic, stylistic, and linguistic features of both Cavafy and Walt Whitman. Initially, we prompted the model with the following instruction: “I will give you poems from the Greek poet Cavafy that are written in Greek. Please consider the register, style and thematic of the poems and provide a translation of the poems from Greek to English, okay?” This was done to ensure that the system maintained an awareness of the original poetic nuances. As we assessed the initial translations, another prompt was used to address the lexical choices made by the model: “The translations you have generated indicate a more realistic and logical approach to the vocabulary used, e.g. you use verb and/or noun clauses like 'built walls' and 'shut me out' in the poem 'Walls,' and in some cases, they may be more formal like 'place them' in the poem 'When they awaken.' In the reference translations, we can find in the same instances that the lexical choices are 'built tall, strong walls,' 'close me off,' and 'put them' respectively. Is there any specific reason for your choice?” This allowed for reflection on how the model handled word choice and encouraged the AI to consider alternative phrasing. Finally, to test the model's ability to emulate another poet’s style, we provided the following direction: “Now I am going to provide you with some poems of Walt Whitman and I want you to proceed to the translations of Cavafy's poems from Greek into English, emulating Whitman's thematic, rhyme and poetic style. Okay?” This step was crucial in determining whether the AI could successfully adapt the thematic depth, free-verse structure, and expansive tone characteristic of Whitman’s poetry, while preserving the essence of Cavafy's original works.

Cavafy’s Poetic and Language Choices

Examining the nuances in Cavafy’s poetry, however, requires an understanding of his unique poetic and linguistic choices, which are marked by refined subtlety and a distinct, often melancholic tone. Cavafy’s poetry is celebrated for its refined subtlety, marked by a restrained yet intense emotional tone that often borders on the melancholic. His language is characteristically spare and understated, which stands in contrast to the elaborate and verbose tendencies of many of his contemporaries. Cavafy’s poems often feature themes of historical reflection and personal longing, using historical figures and events as vehicles for his introspective commentary on modern life (Keeley & Sherrard, 1992). This stylistic approach creates a distinct atmosphere, where the weight of the past mingles with the immediacy of personal experience, leaving much for the reader to infer from what is unsaid (Bowersock, 1970).

The act of translating Cavafy’s work demands careful navigation of these subtleties and layers of meaning. Keeley and Sherrard’s translation (1992) exemplifies this, as it retains the ironic tone and nuanced metaphors inherent in Cavafy’s style—elements that AI-generated translations often risk flattening in favour of directness. Cavafy’s particular use of historical allusion, which allows him to address universal themes such as desire, existential uncertainty, and the transience of power and beauty, calls for a translator who can capture both his linguistic precision and the depth of his emotional expression (Mendelsohn, 2009). Thus, translating Cavafy requires a delicate balance: respecting his sparse stylistic choices while preserving the depth and irony interwoven in his words.

Chat-GPT 4 Poems’ Generation

And now follows the analysis of the poems selected for our research:

"Walls"

Keeley and Sherrard’s translation of "Walls" focuses on the stark imagery of entrapment and isolation, using phrases such as "thick and high" to describe the metaphorical barriers surrounding the speaker. ChatGPT-4's translation, while similar, opts for the phrasing "tall, strong walls," which gives the impression of more physical, rather than psychological, confinement. Both versions succeed in conveying the central theme of isolation, though Keeley and Sherrard’s tone is more passive and resigned, while ChatGPT-4's translation feels slightly more direct and active.

"When They Come Alive"

In the poem "When They Come Alive," the key themes of sensuality, secrecy, and longing are central to both translations. Keeley and Sherrard capture the subtle eroticism of Cavafy’s lines, with phrases like "half-hidden in your lines" suggesting an undercurrent of suppressed desire. ChatGPT-4 retains these themes but uses more straightforward phrasing, such as "visions of your erotic longing," which diminishes the poem’s subtlety. Both translations maintain the introspective tone, but Keeley and Sherrard’s use of understatement adds a layer of sensual ambiguity that is slightly lost in the AI-generated version.

"Aimilianos Monae, Alexandrian, A.D. 628-655"

Both translations of this poem are remarkably similar, with Keeley and Sherrard opting for more polished and refined phrasing. ChatGPT-4 retains the sombre irony of Aimilianos’ reflection on his own youthful bravado but employs slightly more formal language, such as "out of talk, appearance, and manners" instead of "with words, with appearance, and with manner." Despite these differences, both versions effectively convey the tragedy of youthful idealism and the futility of Aimilianos' efforts to protect himself from the harshness of reality.

"Young Men of Sidon (400 A.D.) "

In "Young Men of Sidon (400 A.D.)," Keeley and Sherrard demonstrate a refined command of language that aligns closely with Cavafy’s original style. The description of "a delicate odor of flowers mingled with the scent of the five perfumed young Sidonians" exemplifies their ability to evoke a sophisticated, almost ethereal atmosphere, indicative of Cavafy's nuanced approach. ChatGPT-4, while accurate, adopts a more straightforward expression with phrases like "five scented young men of Sidon." This choice, while conveying the core meaning, lacks the subtle elegance that characterizes Keeley and Sherrard’s version, which encapsulates the quiet opulence of the scene. In the young protagonist’s critique of Aeschylus’s heroism, Keeley and Sherrard’s rendition of "sentiments of that kind seem somehow weak" subtly underscores Cavafy’s disdain for performative bravery. ChatGPT-4, in contrast, renders this line more directly as "Expressions like these seem somewhat like cowardice," a phrasing that, while clear, sacrifices some of the layered irony that Keeley and Sherrard’s choice retains.

"A Byzantine Nobleman in Exile Composing Verses"

Similarly, in "A Byzantine Nobleman in Exile Composing Verses," Keeley and Sherrard exhibit a finely tuned sensitivity to tone. Their translation of the nobleman’s sardonic remark, "may she be cursed, that viper Irini Doukaina," precisely captures the blend of bitterness and wit that defines Cavafy's protagonist. ChatGPT-4’s rendition retains the essential sentiment but lacks the same degree of linguistic flair, resulting in a more literal interpretation. The difference is further evident in the nobleman’s self-assured declaration of his literary skill; where Keeley and Sherrard use "the intellectuals of Constantinople don’t know how to compose," ChatGPT-4 opts for a more direct "I insist that no one knows better." This divergence highlights a recurring pattern: Keeley and Sherrard’s translation choices tend towards evoking a sense of nobility and irony, whereas ChatGPT-4’s interpretations, while accurate, often read more literally, resulting in a slightly flattened portrayal.

"Alexandrian Kings"

In "Alexandrian Kings," Keeley and Sherrard's diction vividly conveys the opulence and theatricality of the scene. Expressions such as "the Alexandrian Gymnasium a complete artistic triumph" and "rose-colored pearls" capture both the grandeur of the setting and the underlying irony of its emptiness. ChatGPT-4's translation, though similar in structure, is marked by a more straightforward language, exemplified in phrases like "pink silk" and "white ribbons embroidered with pink pearls." This shift towards literal language tempers the richness of Keeley and Sherrard's translation, which more adeptly balances the visual splendor of Cleopatra's court with the spectators' understanding of its hollowness.

Whitman’s Poetic and Language Choices

While Cavafy’s style is defined by its restraint and layered subtleties, Walt Whitman’s poetic approach presents a striking contrast. Embracing an expansive and overtly celebratory tone, Whitman’s work exudes inclusivity and grandiosity, often reflecting his democratic ideals through an open and direct style (Bloom, 1999). Known for his use of free verse, long flowing lines, and a preference for anaphora and parallelism, Whitman’s poetry embodies a rhythmic and accumulative vision of humanity that aligns with his broad view of American identity (Miller, 1997). His language is direct and exuberant, frequently engaging the reader through inclusive pronouns like “I” and “we,” fostering a sense of shared experience (Aspiz, 1980).

This contrast in style and tone between Cavafy and Whitman posed a unique challenge for ChatGPT-4, which was tasked with emulating Whitman’s stylistic features while translating Cavafy’s reserved and introspective poetry. The following section, “Chat-GPT 4 Walt Whitman’s Emulation,” examines how effectively the AI integrated these stylistic elements, moving between the restrained structure of Cavafy’s originals and Whitman’s expansive voice in poems like "Walls," "Desires," and "Windows."

Chat-GPT 4 Walt Whitman’s Emulation

Between “Walls”, ‘Windows” and “Desires”

The professional translations of "Walls," "Desires," and "Windows" follow Cavafy’s original style with precision and restraint, maintaining his characteristic compactness and emotional subtlety. In "Walls," the translation uses simple, direct language—"thick and high"—to evoke the physical and psychological barriers that imprison the speaker, emphasizing a passive realization of isolation. The tone remains subdued, with the speaker reflecting on how these walls closed them off from the world without their noticing. Similarly, in "Desires," the professional translation employs a delicate, yet vivid, metaphor of unfulfilled desires as "beautiful bodies of the dead" preserved in a "magnificent mausoleum." This controlled language underscores the quiet sorrow of missed opportunities without overt emotional display. Lastly, in "Windows," the professional translation retains the introspective tone, focusing on the speaker’s struggle to find relief from the darkness. The precise phrasing of "Perhaps the light will prove another tyranny" reflects Cavafy’s restrained ambiguity, where the longing for liberation is tinged with fear of what it may bring.

In contrast, the Whitman-inspired translations of these poems adopt a more expansive, philosophical tone, reflecting Whitman’s tendency toward introspective exploration and grander themes. In "Walls," the lines are lengthened, echoing Whitman’s free-verse style, with phrases like "Oh, the walls, the walls they built around me" adding a sense of active contemplation, as if the speaker is engaging in a dialogue with their own fate. The repetition and rhythm build a more dynamic confrontation with the theme of isolation. In "Desires," the Whitman-inspired version broadens the metaphor of preserved desires, portraying them as "bodies untouched by time," transforming Cavafy’s quiet resignation into a universal meditation on the passage of time and unrealized potential. The expansive imagery aligns with Whitman’s focus on the human experience as part of a larger, ongoing reflection on life and loss. In "Windows," the Whitman-esque translation shifts from passive observation to active engagement with the theme of seeking liberation. The phrasing, "wandering back and forth," imbues the speaker’s search with a more existential tone, in line with Whitman’s characteristic blend of wonder and reflection.

The primary difference between the professional and Whitman-inspired translations lies in their treatment of tone, style, and thematic emphasis. The professional translations stay true to Cavafy’s economical use of language and emotional restraint, focusing on the internal struggles of the speaker without overt embellishment. The quiet resignation and controlled imagery in the professional versions create an introspective atmosphere, where emotions are implied rather than explicitly stated. On the other hand, the Whitman-inspired translations expand on these themes with longer lines, broader imagery, and philosophical depth. By adopting Whitman’s free-verse structure and his cosmic introspection, these translations shift the focus from Cavafy’s individual emotional experience to a universal reflection on human isolation, desire, and the search for meaning. This contrast highlights two distinct interpretative paths: one grounded in subtle emotional complexity, and the other exploring the grand scale of human existence through expansive reflection.

Limitations of the Study

While this study provides valuable insights into the capabilities and limitations of AI in literary translation, several constraints must be acknowledged to contextualize the findings. One significant issue is AI-induced bias. ChatGPT-4, like other machine learning models, is trained on vast amounts of pre-existing textual data, which inevitably includes various biases inherent in the source material (Bender et al., 2021). These biases may surface in the AI-generated translations, subtly influencing how certain themes, characters, or cultural references are rendered. For instance, culturally specific allusions or marginalized perspectives may be misinterpreted or diluted, as the AI tends to reflect dominant narratives present in its training data (Caliskan et al., 2017).

Moreover, the scope of generalization presents a notable limitation. The findings of this study are specific to the performance of ChatGPT-4 and cannot necessarily be extended to other AI models. Different models, trained on alternative datasets or designed with different underlying architectures, may yield varying results (Floridi & Chiriatti, 2020). Additionally, the study relies on specific prompts used to generate the translations, meaning that even slight changes in the prompts could produce different outcomes. This lack of uniformity in AI-generated translations highlights the variability and unpredictability of AI in handling complex literary tasks.

Another important factor to consider is the volatility of AI technologies. Given the rapid pace at which AI systems evolve, the capabilities of current models like ChatGPT-4 are likely to be surpassed in the near future. Improvements in natural language processing and machine learning techniques may lead to more refined and accurate translations, potentially rendering the conclusions of this study outdated (Halevy et al., 2009). As such, while the present findings contribute to the understanding of AI’s role in literary translation, they must be interpreted with caution, particularly regarding their applicability to future iterations of AI models.

Conclusion

This study has explored the transformative impact of artificial intelligence on literary translation through the theoretical lens of posthumanism. By examining ChatGPT-4’s performance in translating and stylistically adapting the poetry of Cavafy and Whitman, it is evident that while AI exhibits remarkable capabilities in generating coherent and contextually relevant translations, it still falls short in capturing the nuanced stylistic, cultural, and emotional dimensions that are essential to literary texts.

The main challenges identified in AI-driven literary translation include the preservation of an author’s unique voice, the accurate rendering of cultural context, and the handling of ambiguity and subtext. These aspects, intrinsic to the literary experience, remain elusive to AI systems, which tend to prioritize linguistic correctness and structural fidelity over deeper thematic and stylistic fidelity. The inability to navigate complex poetic forms, historical allusions, and layered meanings underscores the continuing need for human involvement in the translation process.

Posthumanism offers a conceptual framework that reframes the translator’s role in this evolving technological landscape. Rather than viewing AI as a mere tool or a replacement, the posthuman translator emerges as a relational entity that collaborates symbiotically with AI systems. This paradigm shifts the translator’s identity from being a solitary expert to a co-creator within a network of human and non-human agents, reshaping how translation is performed and perceived.

However, this partnership is not without its complexities. The translator must navigate the limitations of AI, refining its outputs and addressing the biases embedded in its training data. In this sense, the posthuman translator acts as a mediator, combining the computational strengths of AI with the interpretative depth of human expertise to produce translations that honor the intricacies of literary texts. Viewing the process of translation as a co-production stemming from the dynamic and ever expanding relationship between the human agent and AI, the frame is set for understanding translation in the age of AI without resorting to either overstating the sanctity of the human translator or overemphasizing the capabilities of AI translation systems.