Mon, 11 Aug 2025

Mon, 11 Aug 2025 How to get AI to work in 22 languages

India tackles the problem of making AI translate between its many languages and dialects

* Vineet Sawant, a delivery driver in Mumbai, found it difficult to navigate instructions on the Zepto app due to language barriers. However, after the introduction of an AI translation service by Zepto and Reverie Language Technologies, he can now receive instructions in Marathi.
* The Indian government's Bhashini project has made significant progress in developing high-quality datasets needed to train AIs for 22 official languages and hundreds of dialects.
* Experts warn that without tech that understands and speaks these languages, millions are excluded from the digital revolution, especially in education, governance, healthcare, and banking.
* The use of generative AI systems like ChatGPT has made the task more urgent, as vast amounts of data are required to train an AI, but high-quality datasets for many Indian languages do not exist or are not digitized.
* Reverie Language Technologies is deploying its AI-driven translation technology for various Indian companies, while Professor Bhattacharyya's work on Bhashini aims to address the issue by creating a government project that develops high-quality datasets and language models.
  >>