The Rate of Loss

Languages disappear at a rate that most people find, when they first encounter the data, simply implausible. Linguists estimate that of the roughly seven thousand languages spoken today, somewhere between half and ninety percent will be gone by the end of this century. Many will leave behind no recordings, no grammars, no dictionaries. The knowledge encoded in them — the specific ways they categorized experience, the concepts they made available that no other language has, the stories they carried — will be gone with them. Cultural practices are disappearing alongside language but are even harder to measure. A ceremony requires not just knowledge of its procedures but the social conditions that give it meaning: the community that understands it, the relationships it enacts, the seasonal and ecological context it responds to. When those conditions erode — through urbanization, forced relocation, the death of elders, the economic pressure to assimilate into dominant culture — the ceremony does not immediately stop being performed. It slowly becomes something else: a performance for outsiders, a simplified version, a memory of what it used to be, then nothing.

The Archive Is Not the Tradition

There is a long history of well-intentioned archiving projects that treated the problem of cultural loss as a documentation problem. Record everything, the thinking went, and you will preserve it. What this approach misses is that documentation and transmission are not the same thing. A video recording of a dance ceremony documents the visual form of the movements. It does not transmit the years of embodied learning that a dancer brings to those movements, the relationships within which the ceremony functions, or the seasonal knowledge that determines when the ceremony is appropriate. The archive preserves the form. The tradition lives in the practice. When the practice stops, the archive becomes an artifact of something that no longer exists, not a continuation of it. This is not an argument against archiving — archiving is valuable and important. It is an argument for not mistaking archiving for preservation.

Where AI Can Add Value the Archive Cannot

The specific contribution of AI to cultural preservation is not documentation — it is potential reanimation. A conversational AI trained on the recorded knowledge of a tradition can engage with that knowledge in ways that a static archive cannot: responding to questions, generating new instances of traditional narrative forms, situating specific elements within their broader cultural context. This is not the tradition itself. It is something new that the tradition's recorded knowledge has made possible. The distinction matters. A language model trained on recordings of an endangered language and grammatical documentation of that language can generate new text in the language, respond to questions in the language, and help remaining speakers access the language in interactive ways. It cannot restore the language to the social life that made it alive. But it may serve as a bridge: a way for members of the community to maintain connection with the language in the absence of a critical mass of living speakers, and a way for learners to develop facility that might otherwise be inaccessible. A UNESCO-affiliated research project working with communities in Papua New Guinea — one of the most linguistically diverse regions on earth, with over eight hundred languages, many of them critically endangered — has been testing AI language tools built on documentation collected over several decades. Early results suggest that community members, particularly younger people, engage with AI language tools at significantly higher rates than with traditional textual resources, and that the conversational format allows for types of practice that passive consumption of recordings does not.

The Sovereignty Question

Any discussion of AI-assisted cultural preservation must engage directly with the question of who controls the technology and who controls the knowledge it encodes. Communities whose cultural traditions are being documented and used to train AI systems are not always — or even usually — the same communities that develop, own, or govern those systems. The history of extraction is long: ethnobotanical knowledge used by pharmaceutical companies without benefit-sharing, sacred imagery reproduced commercially, ceremonial knowledge published without community consent. A tangent that sharpens the issue: the concept of indigenous data sovereignty — the right of indigenous peoples to govern the collection, ownership, and application of data about their communities — has been developing as a framework precisely in response to this history. The Global Indigenous Data Alliance and similar bodies have articulated principles that would, if applied to AI development, require genuine community governance of any AI system trained on indigenous cultural knowledge. The gap between those principles and current practice in the AI industry is significant.

What Living Transmission Still Requires

Researchers at the University of Auckland studying Maori language revitalization have found that technology tools — including apps, AI tutors, and online resources — are most effective when they are embedded in programs that maintain human community and face-to-face transmission. Technology as a supplement to living practice produces measurably better outcomes than technology as a substitute for it. This finding points toward a realistic framing of what AI can and cannot do. Cultural traditions are not files. They are living systems. AI can help create conditions for those systems to persist and for separated community members to maintain connection with their heritage. Whether the tradition lives depends on whether the community does — and that is a question of political will, economic conditions, and human solidarity that technology cannot answer.