By Ismahane Elouafi
Over many decades, the agricultural research community has supported vulnerable smallholder farmers by improving crops, animals, and the resilience of farming systems, amassing a substantial and valuable body of data along the way. As global agricultural challenges deepen, artificial intelligence (AI) now offers powerful ways to unlock these data and enhance agricultural science. The potential of AI is compelling given the plight of vulnerable, smallholder farmers, including marginalized groups such as women, youth, Indigenous peoples, and remote, underserved communities. But amid the rush to develop and deploy AI tools, the sector must address key risks to keep expectations grounded and outcomes relevant and equitable.
While AI encompasses a broad suite of technologies, machine learning (ML) has already delivered considerable advances in agricultural science by driving the rapid analysis of large, complex datasets. ML speeds up pattern detection and improves prediction and scenario testing in climate modeling. It also enhances remote sensing, supply-chain analysis, resource management, and livestock disease detection, and accelerates the development of climate-smart crops. Large language models (LLMs) have democratized and enhanced knowledge delivery into wider policy and farmer-level decision-making through AI-supported advisory systems and multi-language interfaces like Digital Green.
Along with common questions about data sources and privacy, concerns about AI center on inflated expectations and users’ overreliance. This can overlook AI’s tendency to amplify errors, gaps in data, and biases via feedback loops that can be introduced via decisions, assumptions, and biases made far from the field during design - often from Western contexts - as well as by current researchers and users.
Click here to see more...