Can our language data be used to train artificial intelligence?

Short Answer #

Yes, language data can potentially be used to train artificial intelligence systems. The more important question is who decides, under what conditions, and for whose benefit.

Expanded Answer #

Artificial intelligence systems learn from examples. For language technologies, those examples may include:

• Recordings
• Transcripts
• Dictionaries
• Stories
• Educational materials
• Translations

As Indigenous language technology advances, communities are increasingly asking important questions about how language resources may be used. The issue is not simply whether AI training is possible. The issue is governance. Questions communities may wish to consider include:

• Who controls the language resources?
• Who approves AI use?
• What are the intended outcomes?
• Who benefits?
• What protections exist?
• Can permissions change over time?

These questions do not have universal answers. Different communities may reach different conclusions. The most important principle is that communities should have meaningful authority over decisions involving their language resources.