Another Step for AI Translation and Its Implications for Low-Resource Languages
The rapid advancement of artificial intelligence is transforming many industries, including translation. Powerful new AI systems like chatbots and Large Language Models can now translate between languages with high accuracy. However, most of this progress has focused on high-resource languages like English, Spanish and Mandarin. Far less attention has been paid to lower-resource languages with fewer speakers and a lot less digital data as is the case for many of our Pacific Islands languages.
"rely on imperfect machine translation or switch to contents in “mainstream” languages"
Public attitudes on AI remain mixed. People recognize the benefits of AI translation, like increased access to information across languages. But they also worry about risks to privacy and lack of control over how their data is used to train language models. These concerns are even more pronounced for minority language communities like Polynesians, Micronesians and Melanesians.
As a translation company, Huri Translations is on the front seat to attest to how speakers of low-resource languages like Chuukese and Marshallese often face challenges in the digital world. Content translated into Chamorro or Palauan is contingent on the goodwill of organizations to invest in serving small communities. Content in their native tongue can be scarce, and they may need to rely on imperfect machine translation or switch to contents in “mainstream” languages. This compounds their sense of marginalization and underrepresentation.
At the same time, our island communities have an opportunity to shape the development of AI in a way that protects their interests while staying attuned to the outside world. Doing so will require engaging with both the tech industry and government regulators to craft a framework compatible worldwide.
Some solutions could include requiring informed consent for use of minority language data, putting stronger limits on how it can be applied, and giving speakers more oversight through ethics boards. Firms hoping to serve these communities must make extra effort to build trust and co-design systems that will ensure protection of their data privacy, a Taonga in the case of the Māori, and allow the tech sector to cater for them.
There are also technical solutions like federated learning, where user data stays on their devices. This allows training models for say, Fijian or Samoan, without centralizing sensitive personal information related to their health, employment history or judicial background.
Getting translation right for underserved populations will take a scalable, collective approach. But it has huge potential to connect ideas and technology across languages and empower users from the islands. With care, discernment, and cultural values such as Tikanga, AI translation can be harnessed to give voice to people of all backgrounds.
At the same time, our island communities have an opportunity to shape the development of AI in a way that protects their interests while staying attuned to the outside world. Doing so will require engaging with both the tech industry and government regulators to craft a framework compatible worldwide.
Some solutions could include requiring informed consent for use of minority language data, putting stronger limits on how it can be applied, and giving speakers more oversight through ethics boards. Firms hoping to serve these communities must make extra effort to build trust and co-design systems that will ensure protection of their data privacy, a Taonga in the case of the Māori, and allow the tech sector to cater for them.
There are also technical solutions like federated learning, where user data stays on their devices. This allows training models for say, Fijian or Samoan, without centralizing sensitive personal information related to their health, employment history or judicial background.
Getting translation right for underserved populations will take a scalable, collective approach. But it has huge potential to connect ideas and technology across languages and empower users from the islands. With care, discernment, and cultural values such as Tikanga, AI translation can be harnessed to give voice to people of all backgrounds.