The Role of AI in the Decoding of Lost/Ancient Texts

The automated decipherment of lost languages has emerged as a critical frontier in digital humanities, yet the epistemological and technical limitations of AI models remain underexplored and misrepresented. This paper proposes to study ‘To what extent can AI language models, such as NLPs trained in low-resource settings, help with the deciphering of lost languages, and what are the actual limitations relative to traditional methods used by linguists?’. It argues that, despite general fear of AI replacement, while AI models accelerate decipherment workflows, they cannot operate independently, as they require expert validation and interpretation to generate reliable results, via a thorough analysis of traditional methods compared to AI-assisted models, backed up by successful case studies such as Linear B and Ugaritic. The results demonstrate how these models offer valuable advantages when working in low-data environments. The research deduces that AI is a tool with immense potential, which can be reached only through a human-machine collaboration that regards the ethical concerns of these processes.