Language, language access and AI
If you've used Google Translate, you've used AI for translation. Early results were often frustrating, but we appear to be moving towards a world in which high-level translations are still best performed by human translators, while AI can provide an ever-expanding base. For some, not all languages.
As public sector employees, you may be tempted to just add a 'translate with Google or Microsoft!' dropdown for every language and call it done. But for essential and legal translations, and questions of equity, consider the following.
High vs low resource languages
- Artificial Intelligence and Language Translation in Scientific Publishing: "NMTs and other large language models (LLMs) have a significant imbalance in their coverage of languages, and these systems tend to perform better with high-resource languages such as English, Spanish, Chinese, and French."
- Use human translators for essential and legal work: "Human translation [..] offers a deeper understanding of the cultural context and nuances of the translated texts [..] legal terminology and concepts can vary between different jurisdictions, AI translation tools may not always be able to account for these differences. Currently, it is thus recommended that a skilled human translator with expertise in legal translation is used to ensure the highest levels of accuracy and quality when translating legal texts."
In the real world
- In a KALW radio broadcast about language access equity, journalist Zhe Wu with the San Francisco Public Press discusses attending meetings at San Francisco City Hall and being asked to translate for monolingual, usually elderly Cantonese speakers. She mentions that cities like San Jose and Millbrae use AI tools to provide these services while acknowledging these tools are not the most accurate. In the same broadcast, San Francisco officials say human translations are better quality. But the burden of translation often falls on community organizers, and this is not always available.
- San Jose is going all in on AI translations for public meetings: Sandra Hernandez, CEO of California Healthcare Foundation, says "This technology is a very powerful tool in the area of language access [..] In good hands, it has many opportunities to expand the translation capability to address inequities. [..] But the human interface is very important to make sure you get the accuracy and the cultural nuances reflected."
- California plans to rely on AI to translate healthcare information
- Some people believe that with AI, people no longer want to learn other languages.
- In the AI era, is translation already dead? This is a human translation of this Chinese language article written by a translator in China.
How is your city, county, state, federal, or national government thinking about AI and translation? Does it improve language access and equity? Will it prevent you from actually offering better services? I'll leave you with this thoughtful piece on the challenges of procuring and using AI, and how "if not approached carefully, AI may solve old problems while creating new ones."