Artificial intelligence (AI) has been around for decades, but has become hard to miss in the past few years. Today, what many people refer to as AI—OpenAI’s ChatGPT, Microsoft’s Copilot or Google’s Gemini—for example, are a newer advancement in AI called large language models (LLMs). LLMs were developed from machine learning research and use statistical prediction to capture the deeper context of human language and generate text.

Just last fall, the University of Illinois launched Illinois Chat, an LLM-based chatbot that is available to students and employees of the university. This AI software tool developed by the National Center for Supercomputing Applications (NCSA) allows users to create personalized LLM-based chatbots with local data and privacy safeguards.

CROPWIZARD

It just so happens that one of the first significant research applications built on the Illinois Chat platform was designed for the agricultural community. The application, called CropWizard, serves as an interactive question-answering and decision-support service powered by AI. Users can ask CropWizard about pest management, crop rotation or nutrient requirements, or even upload images about which they have questions.

Once a question is inputted, such as “What are the best pest control options for aphids in my soybean crop?” CropWizard then consults over 200,000 trusted agricultural publications. The publications are mainly Extension documents from land-grant universities, a growing set of open-access research papers and a portfolio of computational tools specific to agriculture to answer data-driven questions.

The result is a virtual agronomist capable of offering tailored farming advice, research insights and computational analyses of user data. CropWizard can be found at uiuc.chat/cropwizard.

This tool, developed by the Center for Digital Agriculture, AI for Future Agricultural Resilience, Management and Sustainability (AIFARMS), and NCSA, is just
one example of AI being used in agriculture.

AI IN RESEARCH

“AI is a big field with a long history,” said Matthew Hudson, Ph.D., Crop Sciences Professor at the University of Illinois. “There is a lot of machine learning and machine vision technology that is used in plant-breeding research. AI in the broader sense is very widely used in research around phenotyping, genomic selection and DNA sequencing, as well as increasingly in automated farm equipment.”

Hudson’s research project that is partly funded by the Illinois Soybean Checkoff, “Using Genetic Engineering to Help Control Soybean Cyst Nematode (SCN),” is researching new ways to reduce SCN reproduction and survival. The project utilizes some AI tools in the background of the research, especially around DNA. Using an AI-based algorithm, Hudson’s team is able to more quickly analyze the genome sequence to define DNA markers.

“The underlying thing about all of this is machine learning,” said Hudson. “Rather than programming a computer to do exactly what you want it to do, you teach it to learn. The way the machine learns is customized, for example to identify different kinds of pests or to make a breeding selection.”

AI technology has the potential to reduce the time it takes to develop new crop varieties— something that usually takes several years.

One area of advancement has been in phenotyping, which is the process of measuring and analyzing observable plant characteristics.

For example, images of plants can be taken, and AI can examine those images for stress, or traits related with yield. This can be done, for example, by a rover or drone going through the field to take photos or videos. These then get analyzed using AI, which can count details such as the number of pods in a soybean plot or spikes in a wheat plot.

“There is a lot of interesting work going on to try to get those measurements,” said Jessica Rutkoski, Ph.D., Crop Sciences Associate Professor at the University of Illinois. “These are measurements we don’t normally collect because it wouldn’t be possible to collect the number of pods or spikes manually on a large scale. With AI, we are now able to collect a lot of data that we wouldn’t be able to collect before.”

One interdisciplinary research team at the University of Illinois is going beyond phenotyping to collect data via rovers in soybean fields before any signs of distress might be visible. The team’s rover, created by 3-D printing, is about the size of a shoebox. It collects information via sensors and can move under and around the soybean canopy.

“These multimodal data systems give us what we call a ‘data cube’ that basically stacks different data points per time point coming from different sensors,” said Elhan Ersoz Ph.D., Crop Sciences Clinical Assistant Professor at the University of Illinois. “This gives us high fidelity and repeatability capabilities and reduces our error rates.”

The rovers, through AI, algorithms and data transmissions, have allowed the researchers to cut their workload significantly because seed composition components, such as soybean oil and protein content, and stressors such as disease or insect pressure, can be determined in real time through different sensors.

“We can collect this data while the plant is growing instead of waiting for the plant to grow, dry, be harvested and then measuring the composition,” said Ersoz.

One of the biggest challenges of a plant breeding program is the scientific difficulties of collecting information and the time it takes to develop new breeds.

“Breeding is a numbers game,” explained Ersoz. “We are lacking the infrastructure to get the numbers we need in a public breeding program. With AI and all of this technology, we are finding ways to do that without having people physically in the fields.”

DATA CHALLENGES

With AI comes an increase in the quantity of data collected and available.

“It’s amazing because we have all of this data, but how are we going to use that?” said Rutkoski. “We know how we use yield data, but we don’t entirely know how we’re going to use these new data sources that we didn’t have before.”

In general, researchers are trying to figure out where AI might be used in their work. But one large challenge is that a lot of the AI models are not built with crop research methods in mind. This means researchers are trying to utilize something made for a completely different industry or purpose and apply it to breeding.

U.S. SOYBEAN GENETICS COLLABORATIVE

The U.S. Soybean Genetics Collaborative (USSGC) is a new checkoff-sponsored project that aims to share expertise, build strategic alignment and drive technology advancement across public soybean breeding and genetics programs. Partially funded by the Illinois Soybean Checkoff for fiscal year 2026, USSGC is led by Bryan Stobaugh, Director of Licensing and Commercialization at the Missouri Soybean Association and Missouri Soybean Merchandising Council.

“Public soybean breeding and genetics programs have played a vital role in developing traits, germplasm and, most importantly, training and educating future scientists,” said Stobaugh. “However, accessibility to public-bred soybean varieties, traits and other technologies has been
hindered over the last 30 years. Varieties developed from public breeding programs currently occupy less than 10% of the U.S. soybean acreage.”

To address these challenges, USSGC is engaging with individual breeding programs to gain a comprehensive understanding of their strengths, weaknesses, opportunities and threats. The Collaborative then will compile current public breeding assets and formulate a way to promote those assets and facilitate industry engagement, in addition to fostering a network for knowledge and resource sharing among stakeholders.

Ultimately, USSGC aims to create a Seed Guide for all publicly bred soybean lines to allow for U.S. soybean farmers to find their checkoff return on investment in one spot.