We devote so much attention to discussing AI in the language of fear that we often overlook some more imaginative question: What if AI could help save something that matters rather than replace it? What if humanity's most sophisticated tools could help preserve the most fragile parts of our shared heritage; these languages that are, in effect, quietly vanishing at the edges of the world?

Every couple of weeks, another language disappears. Not with a flourish or any kind of ceremony, but with a quietness that seems almost to be a kind of cruelty. An elder dies.

A story is told for the last time. A lullaby is sung for the last time. And a worldview, a history, a cosmology, falls into silence. The extinction of a language is a type of climate crisis; a shrinkage of the human imagination.

It is ironic that the very technologies propelling global cultural flattening might also become the means to repair that loss.

Image Credit: AI Hierarchy on Wikimedia Commons

AI Has Caused Harm. That’s Why This Matters More.

Let us be truthful: AI has not gracefully entered the world.

It has reproduced biases endogenous to the data we feed it. It has privileged dominant languages to such a degree that English, Mandarin, and a few others drown out thousands of smaller languages. It has scraped Indigenous stories, songs, and words without consent, taking cultural heritage and producing it as training data for companies that will never return those stories, songs, or words to the communities to which they belong. It has automated misinformation and accelerated cultural homogenization, and has exacerbated anxieties, some legitimate, of erasure over preservation.

So, the skepticism that communities have of AI is not paranoia. It is memory. It comes from histories of extraction, misrepresentation, and harm. But this is why it also holds such important possibility.

Because if AI can cause harm at scale, then it can repair at scale, but only if we choose to use it differently. Only if we decide that technology should serve humanity, not only efficiency or profit.

Image Credit: Cash Macanaya on Unsplash

Take the Quiz: Religion, Schools, and Equality

Religion in Schools: Teaching Respect, Not Bias.

AI as an Unexpected Guardian of Cultural Memory

Languages that are endangered often suffer from what linguists refer to as “critical mass.” There are too few speakers, too few recordings, and too few places for the language to function in its usual, natural habitat. AI allows for scaling up of what is possible. For example, machine learning models can identify grammatical structures that would have previously taken years of fieldwork.

Neural networks trained on a few audio samples can yield approximations of pronunciation breaks and patterns. Speech recognition systems can be trained to accommodate rare dialects which infuse tones, clicks, or complex morphology, in ways that previous software simply could not contend with.

Artificial intelligence does not take the place of the work of community elders. Instead it multiplies it. It is speeding up the documentation of language prior to the disappearance of the last speaker of the language. It is creating digital files that would take most of our lifetimes in order to build.

But importantly: it requires participation, consent, and cultural humility, which tech hasn’t always taken into consideration.

Image Credit: Haseeb Jamil on Unsplash

When Communities Lead, Technology Follows

The most impactful preservation efforts occur when the community directs the mission and the technology listens.

Digital tools are enabling children in Hawai‘i to practice pronouncing a language that was on the brink of extinction. Inuit technologists in Canada are programming AI models to grapple with Inuktitut's sophisticated grammar. Sámi activists in Europe are digitalizing collections of oral histories that had been trapped in old cassette tapes and developing searchable documents that can exist as contextualized and, living archives.

These are not just simply technical innovations, they are acts of reclamation.

AI is meaningful not when the language is stored in the AI, but rather, when a child can speak, read, laugh, and dream in that language.

Image Credit: Hannah Busing on Unsplash

The Emotional Weight: What We Lose When a Language Dies

My fascination with endangered languages has never been separable. It comes from a quiet lament of seeing not just a language consisting of words, but a language containing a worldview. A language holds (often untranslatable) metaphors; it contains emotions that exist only in its specific grammar; it carries memories related to the land where the language was birthed.

Once a language is lost, we have lost a way of making sense. AI cannot replicate this soul. It can, however, hold its outline long enough for a community to resurrect it.

Sometime it is the difference between a language becoming an artifact and a language becoming the future.

Image Credit: Getty Images on Unsplash

A New Narrative for Technology

We speak about AI as if it is a force of nature, something that happens to us, rather than something we are building. But technology is merely a tool and tools take shape based on the hands that wield them. AI has caused harm.

It has exacerbated inequity. It has homogenized cultures. It has sucked up data without asking. Those truths don't cease to be true simply because we can envision it differently. But maybe the point isn't to deny the harm AI does; it's to believe the power of AI can be steered toward good.

If we will design AI with care and responsibility, AI could become one of the most powerful allies endangered languages have ever had. AI can document. AI can analyze.

AI can teach. AI can offer support for communities combating centuries of erasure. AI will not save all languages. But it can save some.

And perhaps true advancement is not simply about innovating something new, but how willingly we choose to protect the voices that shaped us and how urgently we would deploy tools to honor that which should not be displaced.