Google hosted its Google I/O Connect Bengaluru 2024 event on Wednesday unveiling several new AI models, tools, and initiatives specifically designed for the Indian AI community.
Among the key announcements was the launch of IndicGenBench, a benchmarking tool covering 29 regional Indian languages.
Google introduced the Gemma 2 AI model, accessible to all developers, which is tailored to better handle Indian languages. Additionally, the Gemini 1.5 Pro model, featuring a two million token context window, was made available to Indian developers, following an earlier waitlist.
This model aims to empower developers to build advanced AI tools and applications.
Gemma's tokenizer, which breaks down text into smaller units for AI processing, is expected to be a significant asset for creating multilingual solutions that cater to India's diverse linguistic landscape. Google also released the Gemma 2 small language model (SLM) to developers.
Google open-sourced its Composition of Language Models (CALM) framework. This allows developers to integrate specialized language models with Gemma to create more efficient solutions. For instance, a developer could use CALM to build an English-language coding assistant that also offers support in Kannada by incorporating a Kannada-specific model.
Google unveiled IndicGenBench, a new benchmarking tool designed for Indian languages, covering 29 different regional languages, some of which had not been previously benchmarked. This tool aims to enhance the development and evaluation of AI models for the diverse linguistic landscape of India.
Google announced a partnership with the Ministry of Electronics and Information Technology (MeitY) Startup Hub to bolster the Indian startup ecosystem. This collaboration includes offering up to $350,000 (approximately Rs 2.9 crore) in Google Cloud credits to eligible AI startups. These credits will support the infrastructure and computational power needed to develop and deploy AI technologies.
Additionally, Google will provide AI-first programming and curriculum through existing programs like Startup School and Appscale Academy. The company also plans to host a nationwide Gen AI Hackathon, a three-month event aimed at supporting early-stage startups in their AI endeavors.
Google highlighted that over 1.5 million developers globally have used Gemini models to create AI-powered tools and applications. With one of the largest developer bases on Google AI Studio, India is poised to build new AI tools utilizing Gemini's capabilities.