This is an archive article published on July 18, 2024

Google to launch tool to make Indian agriculture data driven; renews focus on Indian languages

This tool, utilising high-resolution satellite imagery and machine learning, aims to improve yields, access to capital, and market access for farm products.

This was the 2nd I/O Connect held in Bengaluru (Image credit: Google)

3 min readBengaluruJul 18, 2024 08:35 AM IST First published on: Jul 18, 2024 at 08:23 AM IST

Written by Venkatesh Kannaiah

Google will soon launch the Agricultural Landscape Understanding (ALU) Research API, a limited availability tool designed to make agricultural practices more data-driven and efficient.

This was announced at the Google I/O Connect Bengaluru 2024 by Jeanine Banks, Vice President and General Manager, Developer X, Google.

According to Google, the Agricultural Landscape Understanding API looks to address improving yields at farms, improving access to capital and providing market access to farm products. The use of ALU information is already being explored by select partners like Ninjacart, Skymet, Team-Up, IIT Bombay, and the Government of India.

This tool will provide granular landscape insights at the farm field level. Google says that generating agricultural insights at an individual field level is critical to a change in the agricultural ecosystem. It refers to the challenges with the huge diversity of landscape and the crops leading to varied requirements for even fields in close proximity to each other.

As of now, these insights are available at an aggregate level, but interventions are needed at an individual farm level. Using high-resolution satellite imagery and machine learning, it proposes to draw boundaries between fields, and with the establishment of clear demarcation of fields, it hopes to resolve multiple issues ranging from drought preparedness and irrigation issues to market access problems. The tool can get into such granular details to provide for crop type, field size, distance to water, distance to road and distance to market.

In the Indian language space, the Google DeepMind India team shared updates to empower developers building language solutions for India. This includes the expansion of Project Vaani, in collaboration with the Indian Institute of Science (IISc), which provides developers with over 14,000 hours of speech data across 58 languages, collected from 80,000 speakers in 80 districts. The team also introduced IndicGenBench, a comprehensive benchmark to evaluate generation capabilities of Large Language Models on Indic languages. IndicGenBench is designed specifically for Indian languages. Covering 29 languages, including many that have never been benchmarked before, IndicGenBench provides a valuable resource to assess and fine-tune language models.

Google is also an open-sourcing CALM (Composition of Language Models) framework that allows developers to combine their specialized language models with Gemma models. This enables the creation of better solutions taking into account India’s linguistic variations.

Ambharish Kenghe, Vice President, Google said that the company was fully committed to empowering Indian innovators to harness AI’s full potential, creating solutions that not only address India’s unique needs but also shape the future of AI globally. “The opportunities with multimodal, mobile, and multilingual AI are immense, and we’re thrilled to be a part of India’s AI journey.”

Google is also working with MeitY Startup Hub to support 10,000 Indian startups in their AI journeys, through Google Cloud credits, AI-first programming curriculum, and launch of a nationwide GenAI Hackathon and AI Startup Bootcamp. Developers in India now have expanded access to Google’s powerful AI models with the 2 million token context window in Gemini 1.5 Pro and Gemma 2, the next generation of open models.

# Google