Deepshikha Behera

In this project I study the interplay of orality and digitality in the context of generative AI for low resource languages, especially ones that do not have a significant presence in the lettered world. In the process of understanding decoloniality in creative translation and other forms of  language use in Generative AI, this project provides perspectives on the interplay between creative genres, such as overlapping of poetry with music, storytelling and art, a phenomenon intrinsic to decolonial language-cultures. In doing so, it creates new strategies of undoing standardisation within AI models, against the backdrop of the linguistic, cultural and genre differences that shape the life-worlds of these communities. In the act of rethinking what we mean by ‘low-resource’ languages, this project proposes an alternative: a focus on the knowledge that is produced and disseminated through alternate mediums that are different from writing. 

By understanding ways in which communities that have survived on oral-aural traditions and narratives, embodied acts such as performative traditions and language-use, I devise ways in which these multimodal resources could be integrated into training AI models that are receptive to language difference. Subsequently, I engage with the shifting contours of ‘literariness’ and alternate forms of literary expression in the digital platforms among marginalised communities. My project, thus lays the framework for discussions on ethical considerations of having people and their lived experiences as ‘resources’ for AI model training and foregrounds the possibility for an ethical Human-AI collaboration.