Minnesota's Enterprise Translation Office (ETO) breaks down language barriers with ChatGPT

Minnesota's Enterprise Translation Office (ETO) has adopted ChatGPT as a core solution to break down language barriers and provide equitable public services to all residents. The previous decentralized translation approach had major limitations in cost and time, with inconsistent quality and request processing that could take up to a month.

ETO worked with TAIGA (Transparent Artificial Intelligence Governance Alliance) to redesign its workflow while adhering to safe and ethical AI principles. By using ChatGPT for the initial translation stage and having translation experts review and revise the output, they built a 'human-centered workflow' that maximizes efficiency.

For languages such as Hmong and Somali in particular, spreadsheets compiling culturally appropriate terminology were used to train Custom GPTs, strengthening cultural context. By feeding expert feedback back into the model, they shorten the quality assurance (QA) stage and manage the process so the same errors don't recur.

ETO has now applied ChatGPT across its translation work as a whole, and is running a pilot project using ChatGPT's Voice capabilities for real-time interpretation. If this project succeeds, it is expected to further improve access to public services for non-English-speaking citizens.