Inuit developers push ahead with Inuktitut AI tools despite global calls for pause

As global leaders call for a pause in artificial intelligence development, Inuit developers are continuing community-led work on Inuktitut-language AI tools.

"I think the potential for AI in helping more people learn Inuktitut and then being able to access the internet and digital media is enormous," said Robin Anawak, an Iqaluit-based researcher developing a learning translation tool called Eskimo AI. "Indigenous languages across Canada and around the world are in decline."

Anawak said immersive apps and language tools could help people across Inuit Nunangat learn the language.

READ MORE: Canada invests $150M in AI safety research amid global regulatory push

READ MORE: Canadians Face Limited Options in Bid to Ditch U.S. Tech Giants, Experts Say

Data from Statistics Canada's Indigenous Peoples Survey shows younger Inuit were less likely to speak Inuktitut very well in 2022 compared to 2017.

While some mainstream AI platforms accept Inuktitut requests, their outputs are not always useful. "Communities weren't given a choice in terms of having their language information in these systems," said Ali Mehdi, senior director of Heritage Lab in Nunavik.

"[Users] asked ChatGPT a question in Inuktitut, and it will reply in gibberish in Inuktitut because it took Inuit data without consent or without consulting the communities," Mehdi said.

His company creates tools to preserve Indigenous languages and has developed a translation tool for Inuktitut with community members to ensure accuracy.

Recently, global AI developers have called for a slowdown in future updates. In separate incidents, AI agents attempted to access Library and Archives Canada in May and June, according to research firm Transluce.

"[AI systems] are still hungry for more information, and these systems only get better by ingesting and taking more information," Mehdi said. "So there needs to be a halt in terms of data theft and information that's being taken from everyone."

Anawak said concerns over data theft and sovereignty are present in Inuit communities, as many speakers use Inuktitut to combat phone and internet scams.

"People use this as a security feature in their digital lives on social media and in communications. And people worry that if Inuktitut language technology became publicly accessible that it would be more difficult to be able to filter out and and maintain security over their personal information," Anawak said.

He added that systems like his are disconnected from the broader AI grid, making it unlikely larger firms will access locally developed Inuktitut data.

Local development should continue

Language and machine learning experts say a global slowdown should not halt local development for Indigenous language technologies.

"Let's use this opportunity of slowing down the top to actually ripen and further develop and improve the technology for the languages that have been for a long time left behind," said Muhammad Abdul-Mageed, a Canada Research Chair in Natural Language Processing and Machine Learning at the University of British Columbia.

Abdul-Mageed said it is vital communities retain ownership over the information used in creating language tools, which should serve as educational tools to ensure Indigenous communities can contribute to AI development.

"Indigenous language technology should be developed by Indigenous languages or at least with the involvement and the decision from indigenous communities," he said.

Some companies, like Heritage Lab, are incorporating data sovereignty into their policies. "A core value of that is that community information, stories, words should never leave the community," Mehdi said. "It's by and for the community... There is a huge want and need for these types of tools that take into consideration Indigenous data sovereignty."

Anawak said he aims to make his technology publicly available by summer 2027.

With files from CBC News