AI linguistic diversity aims to make artificial intelligence (AI) accessible for all 7,000+ languages globally, addressing significant gaps where traditional AI training predominantly focuses on English and a few other languages. This exclusion creates biases, threatens digital inclusion, and undermines cultural heritage for speakers of low-resource languages, particularly in Africa, where 1,500 to 3,000 languages exist but lack sufficient written resources. For instance, while English boasts over 7 million Wikipedia articles, Tigrinya has only 335. Currently, only 42 African languages are supported in AI models, limiting opportunities for many. Vukosi Marivate and Chinasa T. Okolo advocate for enhanced data collection and model development to bridge this divide. They emphasize that lack of investment in underrepresented languages perpetuates inequity, potentially locking millions out of essential opportunities. To cultivate AI that serves everyone, innovative approaches to language data creation and model development are crucial, as seen in initiatives like the African Next Voices project.
Source link
