Indigenous languages and artificial intelligence

Artificial intelligence offers both promising opportunities for language revitalization and serious challenges when it comes to Indigenous languages. Unfortunately, we have already seen detrimental use of these tools, such as AI-generated grammar books for Indigenous languages (Bettens, 2024). Since many Indigenous languages in Canada have limited digital data, AI systems trained on such small datasets without community oversight run the risk of providing inaccurate outputs and causing further damage to language revitalization efforts by communities. Many AI-created resources are riddled with errors. Nuanced and context-dependent languages are especially vulnerable to these errors. Without clear governance and Indigenous guidance, the extraction of linguistic data can echo earlier forms of cultural appropriation. Some Indigenous Peoples are concerned that using AI tools disconnects the language from its cultural and ceremonial ties, while others worry that the use of this technology could further weaken intergenerational transmission of knowledge by overshadowing human-led teaching.

Other Indigenous Peoples see AI tools as being useful for supercharging under-resourced language revitalization and help encourage language access for urban and younger members of the community. For those without access to a fluent speaker, respectfully created AI tools may support auditory and oral learning with interactive experiences. In an archival setting, AI tools could aid in the transcription of existing recordings, empowering communities to better access and find the voices of the previous generations. When guided by community priorities, AI can become a powerful ally rather than a new form of colonization, but that requires building Indigenous-led AI expertise, skills, and development. Indigenous-led AI language projects are proliferating with careful attention to digital sovereignty. Projects such as Danielle Boyer’s SkoBot, a language revitalization robot, directly serve Indigenous youth (SkoBots Language Learning, n.d).  In other examples, the Government of Nunavut has recently collaborated with Microsoft’s Azure AI to make Inuktitut part of their text-to-speech features (Government of Nunavut, 2024).