Since the government here decided to develop Indian LLMs, they will focus on voice-based models, rather than the text-based models elsewhere. It will provide access to a wider audience in local languages. Global models do have voice-capability in their AI assistants. However, their models are optimized for English text and a handful of other languages.
Population here is more comfortable in speaking to the model than typing its prompt. The model will use voice-based internet. The services will be accessed through voice commands. The concept will be a game changer. Voice will become a significant enabler.
Our internet using population is 500 million. Still India has 900 million people who remain offline. Many of these are from agricultural community, and non-tech savvy individuals. Voice-based models will serve them better. They will cater to the country’s linguistic diversity.
Some international models support multiple languages. Still, they use limited Indian datasets while training them. Indian AI models will use Indian datasets in training and will take care of linguistic and cultural nuances.
India has received 67 proposals for developing AI models — 22 proposals for LLMs and 45 for SLMs. The committee is examining these proposals.
Leave a Reply