Council Post: Role of AI in creating inclusive products & solutions
Language is often the greatest barrier towards access to information, and hence, opportunities. Today language translation using AI is removing that barrier. For example, Microsoft’s Azure AI now empowers organizations to translate between 100 languages and dialects globally, making information in text and documents accessible to more than 5.6 billion people worldwide3. These include not only the world’s most spoken languages like English, Chinese, Hindi, Arabic and Spanish, but also dialects that are native or preferred by a smaller population.
There are close to 7,000 languages spoken around the world, but sadly, every two weeks a language dies with its last speaker. Recent advances in AI have enabled inclusion of low resource (i.e., having less training data), and often endangered, languages and dialects such as Tibetan, Assamese and Inuktitut (a dialect from the Eskimo-Aleut language family from Canada). A multilingual AI model called Z-code combines several languages from a language family such as Hindi, Marathi, and Gujarati. Leveraging the commonality across languages and using that shared transfer learning capability allows AI to improve the whole language family, reducing data requirements to achieve high-quality translations. Languages are a celebration of a people’s history; what a wonderful way to enable inclusion, as well as preserve culture and heritage!