Google DeepMind’s Morni AI to Cover 125 Indic Languages

Google DeepMind’s India unit is spearheading an ambitious AI project named Morni (Multimodal Representation for India), aiming to encompass 125 Indic languages and dialects. Speaking at the Global Fintech Fest, Manish Gupta, Director of Google DeepMind India, highlighted that among these languages, 73 had no existing digital corpus. Despite Hindi being spoken by nearly 10% of the global population, its online text representation is just 0.1%.

Project Vaani and Language Data Collection

To address the lack of digital data, Google has launched Project Vaani in collaboration with the Indian Institute of Science and ARTPARK (Artificial Intelligence & Robotics Technology Park). This initiative has successfully completed its first phase, creating an open-source database with over 14,000 hours of speech data from 58 languages, contributed by 80,000 speakers across 80 districts. Currently in its second phase, the project aims to collect 154,000 hours of anonymized speech data from all Indian districts.

Expansion of Google Translate

In addition to Morni, Google has recently expanded its language coverage in Google Translate by adding 110 new languages, including five Indian languages. This expansion, facilitated by the PaLM-2 transformer model, now supports over 1,500 global languages and reaches more than 600 million people.

Digital Agri-Stack Initiative

Google is also developing a digital agri-stack to enhance agricultural practices by facilitating loans to farmers, providing affordable crop insurance, and improving subsidy programs. This initiative aims to modernize agricultural data management and support government programs through a data-driven approach.