Artificial intelligence is reshaping the search for meaning in ancient, undeciphered scripts, but it is not a silver bullet. Researchers testing AI on Linear A – the Bronze‑Age writing system of the Minoan civilization – found the software could quickly map which symbols tend to follow one another and which clusters of signs appear together. That speed, however, does not translate into fluent, accurate readings.
Linear A’s surviving corpus totals roughly 7,500 characters, a handful of tablets that fit on a single screen. With such a thin data set, any statistical model can produce a string of matches that appear plausible. The result is a flood of hypotheses that lack the comparative anchors scholars rely on to confirm meaning, such as bilingual inscriptions or known linguistic relatives.
Limits of AI in decipherment
Experts stress that AI cannot hand a human a finished translation because fluency and meaning are not interchangeable. A model may learn that a particular sign often follows another, but it does not know what either sign denotes. Verifying an AI‑suggested interpretation normally involves checking against native speakers, related texts, or a scholarly consensus built over decades – resources that simply do not exist for a language like Linear A.
Without a “Rosetta Stone” for these scripts, there is no way to test AI output beyond expert scrutiny. The process therefore leans heavily on peer review and independent analysis rather than on confidence scores generated by the algorithm.
AI as an accelerant, not a replacement
Despite the hurdles, AI offers a powerful accelerant for the field. What once required months of painstaking cross‑referencing can now be compressed into minutes, allowing a broader community of scholars and enthusiasts to explore the data. The technology democratizes the initial pattern‑finding stage, freeing human researchers to focus on the interpretive work that machines cannot perform.
Jane Adkins, a PhD candidate at Dublin City University, notes that AI’s role is analogous to a fast assistant in a centuries‑old puzzle. It can highlight promising avenues, but the final breakthrough still depends on a genuine comparative anchor and rigorous human review to separate a real discovery from an appealing coincidence.
The consensus among linguists and archaeologists is clear: AI will remain a tool, not a translator. Until a bilingual inscription or a solid linguistic link emerges for Linear A, Etruscan, or other undeciphered scripts, the human element stays indispensable.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory APP - actualités agentiques pour booster votre SEO et AEO.