IIT Bombay's AI Speaks Your Language—Literally скачать в хорошем качестве

IIT Bombay's AI Speaks Your Language—Literally

IIT-Bombay has just given artificial intelligence a cultural upgrade by digitising 30 ancient Indian texts across astronomy, medicine, mathematics, and linguistics. With over 2.18 lakh sentences and 1.5 million words, some dating back 18 centuries, these works are now available as machine-readable datasets on AIKosh—India's official AI data portal. But this is more than just preservation. It’s the activation of Indian knowledge systems into a form AI can learn from. Backed by the Government of India and powered by the BharatGen initiative, IIT-Bombay’s contribution includes culturally significant datasets, Sanskrit translation corpora, Indian language speech tools, mathematical reasoning benchmarks, and visual understanding models. Trained entirely from scratch, this stack reflects a sovereign AI vision—one rooted in Indian languages, scripts, and values. With Param 1, a 2.9 billion parameter bilingual LLM, and over 20 TTS and ASR models in 19 Indian languages, this initiative is laying the foundation for a truly indigenous AI future. This episode of Front Page by AIM TV unpacks how India is not just using AI—it’s teaching it. From Sanskrit to Python, this is AI built for Bharat, by Bharat. Join us at Cypher 2025—India’s largest AI summit and expo. This year’s theme: Make AI in India.