After 54 years of existence and delivering 1,200 seed varieties in pulses and millets, the International Crops Research Institute for Semi-Arid Tropics (ICRISAT) has set its sights on deploying artificial intelligence, machine learning, and large language models (LLMs) to improve the quality of the output.
To start with, it launched an AI-based Plant Health Detector, a tool that helps farmers and other ecosystem players quickly assess the health of a plant and take remedial measures, if necessary. It also developed a digital soil library that helps in quickly assessing the health of the soil.
“This is a very emerging area, not only for CGIAR centres like ICRISAT. Even Indian Council of Agricultural Research has started working on this. These technologies will be really helpful,” ICRISAT Director-General Himanshu Pathak has said.
The institute has plans to harness its knowledge bank to benefit from the power of LLMs.
Dr Stanford Blade, Deputy Director General – Research and Innovation, ICRISAT
“With over 1,30,000 accessions of the mandated crops, we can start to look for new traits in ways never before possible using machine learning and reinforcement learning. We have terabytes of data, and the goal is to find unique traits and develop them into the varieties we are looking for,” Stanford Blade, Deputy Director General (Research and Innovation) of ICRISAT, told businessline.
“It’s for us to find those very unique traits, not only to find those traits but then to develop them and put them into the varieties that we’re looking for. More generally, if I may, around this whole aspect of digital analysis—so whether it’s geospatial sciences, we’re working with the government of Telangana concerning crop insurance and how to make that more efficient,” he said.
“We’re trying to find these very specific areas where the tools will be beneficial to the farmers here in Telangana and across all of the geographies that we work,” he said.
Digital soil library
Pathak said the institute has developed a unique digital Soil Library with data from over 15,000 samples. Stating that the current method of testing soils involves using chemicals, he said the unique technology, using some hyperspectral tools, helps in testing the samples without any chemicals.
“We’ll be able to test the soil and know its health in no time. We are going to calibrate and validate this along with our National Agricultural Research Systems (NARS) partners. Once it is validated, it will revolutionise the whole soil health assessment,” he said.
The institute feels that this is going to be transformative because it is not possible to reach all farmers with the current practices of soil health assessment. “But we need to reach them. We can link this to remote sensing and ensure taking decisions quickly using AI/ML tools,” a scientist involved in the library said.
