Machine learning deployment is underway in the global plant breeding industry
This video is a part of the FCAI success stories series. In the video series, we explain why fundamental research in AI is needed, and how research results create solutions to the needs of people, society and companies.
The startup Yield Systems aims to have a significant sustainable impact on the global food system by creating cost efficient AI capabilities for plant breeding and speeding up the development of crop yields, as well as adaptation to climate change.
As the human population grows, it’s estimated that by 2050 the demand for food will exceed supply by as much as sixty per cent. The quantities of agronomic inputs; land, fresh water and fertilizers; cannot be increased proportionally, so further improvements in efficiency are needed. Climate change will cause significant challenges which need to be addressed to maintain food security.
“As a first step, we created a plant variety recommendation engine that enables more precise plant variety selection in changing environmental conditions. As drought, heavy rains and other extreme weather conditions become more frequent, it is more and more important to be able to select right seed to the right place, to maximise resilience and yields” says Jussi Gillberg, Yield Systems CEO.
At the moment, machine learning deployment is underway in the global breeding industry.
“Early movers have adapted genomic prediction systems and image processing with drone-based systems, and also other parts of the process are being automated. This is no wonder considering the highly data-intensive nature of the breeding process” says Gillberg.
The company originates from a research group led by Samuel Kaski that originally developed machine learning methods for personalized medicine at Aalto University. As they went on, they found that personalized medicine and breeding domains were similar in their prediction problems, and some solutions could be transferable.
The group started to focus on environment-related crop performance in arable farming, and determining which varieties would work best in different types of conditions. The results were used to develop more accurate selection methods for plant breeders.
From that research, Yield Systems developed an AI-powered observation instrument that, paired with machine learning, can get high accuracy estimates of canopy-level traits. This laboratory-like information from field conditions is then turned into rich data and understanding about characteristics, unobservable with competing technologies. This high-relevance data is the key to improving varieties for changing conditions and for reaching extremely ambitious efficiency improvement goals.
Text and video production by Mia Paju.
Yield Systems has its roots in fundamental AI research at the Department of Computer Science of Aalto University. The machine learning methods were developed in the Probabilistic Machine Learning Group.
Read more about research behind Yield Systems: