It was not long ago when the term “artificial intelligence” was something largely reserved for sci-fi movies. But, increasingly, daily life is being influenced with the use of artificial intelligence (AI).
“AI is changing everything from the way we shop with products like Amazon’s Echo using voice commands to initiate the purchase of products while other AI devices like Nest keep our homes safe and comfortable,” said Christopher Wiegman, a graduate student in the Ohio State University Department of Food, Agricultural and Biological Engineering. “These devices represent a new type of ‘smart’ technology that utilizes AI or machine learning. Machine learning distills large amounts of input data into algorithms based on patterns. The amount of investment in the field of AI has grown substantially spanning all economic sectors ranging from industrial to consumer goods, health care and even banking. Technology titans such as IBM, Microsoft, Google, Amazon, and Facebook are committing heavily to continuing development of AI.”
Agriculture is certainly a potential beneficiary of the AI progress.
“AI could analyze all the inputs of agricultural production such as hybrid performance based on historical yield and disease pressure. AI may also help with selecting plant populations based on weather models, soil types and topography, or nutrient and water availability,” Wiegman said. “If there is one area that the digital agriculture revolution has given producers, it is a mountain of data. AI will help producers interpret this data and extract actionable information with which to enhance profitability while ensuring the sustainability of their enterprises.”
These types of applications are at the forefront of the pipeline development for The Climate Corporation.
“One of the biggest things we see on the horizon that we are excited about is getting more data pulled together in a different model that relies on advanced tools like artificial intelligence. This allows for more insights from the data because we can extract more meaningful information that we can bring to bear on management decisions,” said Brian Lutz, yield analytics lead for The Climate Corporation. “Machine learning is a tool that thrives on extremely large amounts of data. Farming is complex and there is lots of information. It is hard to know what the most important patterns are. The machine learning models can pull the important patterns out for us that we just can’t do for ourselves without the use of these kinds of models. They dive into the data and surface the most important patterns so we can translate them into insights for better management.”
This type of work is ongoing, but will be translated into some new user options being offered by The Climate Corporation.
“When we first announced our pipeline a year ago we had over 35 different projects. Since then we have made significant advances in half the pipeline projects and we are moving forward. The majority of those 17 advancements over the last year are in the three buckets of seed scripting tools, fertility management tools and what we are doing with disease management,” Lutz said. “Disease can be difficult to manage because it can be hard to positively identify what disease you have. Lesions of different diseases can look quite similar and even some abiotic stresses look similar. We are developing technology similar to what is in Facebook where they do facial recognition. So, with that technology growers can use their cell phones to take a picture of the corn leaf and the models that are embedded will automatically identify what disease it is. It really helps with the diagnosis part of it. We are also able to marry that with our high-resolution weather data. Not only do we want to know when a disease is present, we also want to know when to alert growers that the weather conditions are prone to fostering that disease. With disease management, timing is everything. That helps them to be timely and make decisions faster. We offer geospatial and scouting tools too so everything is geo-referenced.”
The development of seed prescriptions also harnesses a large amount of data.
“With seed scripting we are bringing all of these types of data together. One area of focus is seed placement — what type of seed to plant and where. As we add more information about soil types, management practices and weather conditions, we marry that up with our plant genetics data at Monsanto so we can really hone in on which hybrid or variety is able to perform best in that particular acre of the field,” Lutz said. “The other side of the coin is the seeding rate. Variable rate seed scripts have been really popular. We know that every acre performs differently and we have the tools to make that variable rate scripting easy.”
Ohio agriculture understands very well the complexities of efficient nutrient management and AI can help offer more tools to help.
“One of the big focal points in our fertility platform now is our P and K scripting tools. We are also starting to focus on soil pH and lime. One of our flagship products has been our nitrogen management tool. Growers have seen the value in getting a more in-depth look at nitrogen in their fields. But one of the biggest asks that we’ve had in the last couple of years was to expand beyond nitrogen to P and K,” Lutz said. “We are trying to maximize the use of all inputs, including fertilizer, on every acre so that we are not over applying but also maximizing output. That is what these tools are all about. We want to make sure we are doing the best possible job of taking all of the information into account.”
The viability of these types of applications moving forward depends on not only gathering on-farm data, but then also analysis with the help of AI and machine learning.
“In the last couple of years we have had tremendous success in getting all the data in one place. That has led the value for us. Now as we bring more data streams we can go beyond just having your data all in one place and derive the most insight from it,” Lutz said. “Growers are increasingly collecting more data. We have grid soils samples, as-applied data, high-res information about the soils — when it is all spread out, that information doesn’t do much by itself. It is only when it is all integrated into one place that it can be used to make better decisions.”