A conference proposed by the AISV/IAVS https://aisviavs.wordpress.com/, in collaboration with ECA-USP.

Over the years, both continental semiotics and American Peircean semiotics have developed a certain expertise in reflecting on what Jakobson called intersemiotic translation, that is, the translation between semiotic systems of different modalities (verbal, visual, audiovisual, musical, gestural). The most significant advances in the semiotic approach to intersemiotic translation are those that have focused on how different ways of narrating and of enunciating are translated across different substances on the plane of expression. In particular, a great deal of work carried out during the 2000s was dedicated to the various ways of saying or seeing, focusing on the media-specific features of each semiotic system in translation (support structure, linearity, tabularity, rhythms, trait density, etc.).

This approach has also made it possible to explore the translation of the material specificities of each language and to observe, for example, how the novel can, through its linguistic rhythm, imitate music, or be translated, in painting or photography, through blurring, or how music can convey the texture of paintings, and so on.

The emergence of Generative Artificial Intelligence (GAI) is radically transforming the landscape of intersemiotic translation and is proving to be a valuable object of study from both the perspective of semiotics and the language sciences. In fact, diffusion models such as Midjourney, Stable Diffusion, DALL-E, and others, use LLM technologies. This allows for the generation of images from linguistic prompts and, vice versa, the production of texts from images (more or less photographic, imitating more or less recognizable styles). It is also becoming possible to produce short videos from prompts, as the technologies continue to be refined in order to better capture the mood, atmospheres, and styles targeted by users. The success and subtlety of these intersemiotic translations today largely depend on the type of training the databases underwent for each experiment, the latent spaces available, and the objectives pursued in each instance of text generation (artistic, scientific, political, advertising, etc.).

In response to the emergence of Generative AI models and the new and old issues it raises, as a provocation this conference takes up the theme of inter-semiotic translation. One of the main axes of the conference is thus the comparison between computer-assisted translation and the forms of translation practiced before the emergence of generative artificial intelligence. It is certain that in order to give useful instructions to AIs and then evaluate the results (from an aesthetic, moral or scientific standpoint, for instance), one must already have an understanding of non-GAI-assisted intersemiotic translations. Conversely, the creativity of traditional translations remains a crucial field for comparison that allows us to grasp the limitations of AI. Other areas of reflection, however, are also welcome, including comparisons between the styles and functionalities of different diffusion models (Midjourney, Stable Diffusion, etc.), as well as discussions on the practices (artistic, political, advertising, etc.) in which these different kinds of intersemiotic translations are produced and through which they circulate within various domains of society.