The success of LLMs require us to decide upon the question of what language is with a new and historically specific set of categories: vectors, contextual word embeddings, dependency parsing, number, data sets and resource richness, and so on. This paper discusses the related topics of language ontologies and language difference in light of these categories. It builds on pre-AI accounts of language and meaning that nonetheless use strikingly similar terms and ideas (I. A. Richards on vectors and contextual embedding; Alain Badiou on transitivity), in order to develop an understanding of language difference which in turn sharpens our idea of what poetic language is and what happens when it is translated. Language difference, in this account, makes possible a correlation of recent advances in decolonial linguistics with a provisional theory of poetic language and translation. It also brings into focus what is at stake (ethically and analytically) in comparison's 'transactions between contexts'.

One of the promised benefits of the internet and AI tools is the democratization of information: they seem to make the world’s knowledge available to anyone with a web browser and suggest that anyone can become a translator by relying on the extensive resources they offer. But as the limitations and dangers of LLMs have become more apparent, it is increasingly clear that users, especially college students, need careful guidance in using these tools in ethical and effective ways. As Wharton professors Ethan and Lilach Mollick have argued, teachers can help students use LLMs to learn evaluative skills and become more attentive readers and writers. José Antonio Bowen and C. Edward Watson repeatedly emphasize in their book Teaching with AI that AI tools are most effective when coupled with thoughtful reflection and expert mentoring. This is as true for translation as for other skills, especially given the ways that LLM-assisted translations can either challenge or perpetuate biases and existing power dynamics. This paper outlines specific methods for helping college students learn how to create, evaluate, revise, and reflect on AI-supported translations that balance fidelity to language and meaning with awareness of the ethical concerns that such translations can and should raise. I will share the experiences of my students (at a socioeconomically diverse, large public American university in a conservative Western state) with LLM-assisted translation as they moved through a sequence of assignments that builds from comparing existing translations of a text, then engaging with the original source (using AI translation as necessary), evaluating LLM-assisted translation results, revising prompts for AI-based translations, evaluating new results, and reflecting on the process throughout. Mentoring students through this sequence can help them become not only more effective translators but also more ethical and self-aware technology consumers inside and outside of the academic setting.

Arabic and Chinese are two large languages. Each plays an important role in the development of digital humanities, translation studies, and AI. Is there a direct line between these two large languages in the make-up of translation generating AI? Is translation to and from these two languages mediated through English? Are AI translation tools educated in culture, ethos and visualising capacity that are inherent in language? This contribution reflects on these issues from the prism of Arabic and Chinese wine poems. Arabic and Chinese have in common a culture and tradition of wine drinking and poetry, but each tradition is grounded in a unique ethos. Would it be possible to 'teach' AI translation tools to be sensitive to difference in culture and ethos?

The emergence of Large Language Models (LLMs) has dramatically transformed the landscape of translation, particularly in the realm of literary and poetic works.This paper explores the potential of Chat-GPT and Claude as a conversational collaborative tools in the translation of the work by Argentinian performative poet Marico Carmona. This research investigates how a conversational approach to prompting can enhance translation processes and focuses on the dynamic interaction between human translator and AI, highlighting how iterative dialogue and targeted prompting can reveal nuanced linguistic, literary and cultural interpretations. Key elements in this research are the AI's capacity to consider performative elements in the translation output, its ability to generate multiple translation alternatives and its potential to effectively serve the needs of alternative voices and ways of expression.