Challenges for Low-Resource Languages

Low-resource languages, such as Tigrinya, face significant challenges with large language models (LLMs) like ChatGPT, primarily due to the lack of high-quality training data. Despite its numerous speakers, Tigrinya receives inadequate attention and resources, resulting in a poor performance of AI-driven tools for its speakers. For example, when tested with various prompts, the AI generated nonsensical responses, highlighting the limitations of its understanding and utility for Tigrinya speakers. This situation underlines a broader issue in the AI landscape, where languages with less representation inevitably result in subpar technological support.