Subscribe
Share
Share
Embed
The NDSU Extension weekly podcast In the Pod: Soybean Updates delivers timely insights and expert advice on soybean production.
You're listening to In The Pod, Soybean Updates, a weekly trek into the latest soybean information from NDSU Extension. How useful can artificial intelligence be on the farm? One uses AI as a tool designed to help determine the optimal production practice for a field. Shawn Conley, soybean and small grain specialist with the University of Wisconsin Madison, was a keynote speaker at the NDSU Soybean Symposium. Shawn, you've been working with AI. Tell us what you've learned so far.
Shawn Conley:I think what really started this off was probably about ten years ago. My program develops a lot of data through our very testing program and all the management trials. So we kind of wanted to build an internal database to be able to handle all this, knowing at some point we would have the technology to be able to handle these larger data sets and be able to utilize those in a broader sense. You add on to that some of the research we did, like with the North Central Subway Research Program, we're able to collect farm data from over 8,000 farm fields in over 600,000 acres from across the North Central Region. And the fact that all of my colleagues across the country post for HRL information and stuff on their web pages. So over time, we've basically amassing this huge amount of information. Now what do we do with it? One of the things we've worked on, first of all, being able to figure out what tool is best. We spent a lot of time working on large language models in order to be able to figure out methodology to be able to sort through this and actually get good information out. During this time, then we've kind of learned a lot of interesting things. For one, we did a meta analysis looking at cover crop, taking out the cover crop data and trying to synthesize it all into one useful bit of information. You know, that took a lot of time. I bet it took us six months, nine months, lots of people hours to do that. I'm like, can we use AI to expedite that process? So what we did is we took those same papers, and then we brought in AI to kind of quickly and how efficiently and how close to what we decided the answers were was this. We found there was a lot of problems with how we as scientists report data in our journals. And we found a lot of gaps in terms of standard error that's not really reported. And there's a lot of specific things you need in order to do a meta analysis correctly. Like 60% of papers didn't have that information in there. And so you, by you I mean the person, would have to go in and eyeball it or take a guess. That's not really what AI is as good at doing, you know. One thing we learned is that, A, we still need human in the loop. It's not like AI is gonna take our jobs away from us. And B, we as scientists are really not publishing correct amounts of information to fully utilize these AI technologies down the road. A group of us are right now working on a paper basically trying to come up with a standard, like everything that we publish in order for meta analysis to be incorporated. That's part of it. On the flip side, we worked with an ag consulting group and built an AI tool that a farmer can drop a pin in any field in The United States and run a scenario and look at what optimal production practices are for soybean or corn. We did corn as well. And then we asked farmers to test our tool. And we're doing some work here at NDSU with Lindsay Malone and her team on this one. Basically what we want to do is have farmers try it and see does the tool work. We're basically trying to break it and then see where the data gaps are. And I think that's been really interesting because this last year, we did a first cut of data. And on average, we saw a $53 per acre increase in profit using this tool over what a farmer was normally doing. Some farmers were at zero, which means they're already optimized. 10% of the farmers usually is what aligns in these leaders. And you have the bell shaped curve where it's whatever the other 70% that I think could use this type of technology. I think what we're doing here is we're seeing some places where we could fit it, but also understand as the more we use AI, we see where the holes are and the gaps are, and that's where the scientists come back in to plug those gaps, if you will.
Bruce Sundeen:Is this tool available for farmers or is it in beta testing?
Shawn Conley:It's in beta testing, but it's freely available to any farmer. Any farmer in the country can use it free of charge. I encourage farmers to try it. Just go to my web site, Badger Crop Network. It's called Boots in the Ground version two point o, and you can get all the information. Use the tool for free. Try it on their farm and give us feedback. Let us know if it works or not. Know, that's the fun thing about this. And I think what other things we're really trying to do is just understand where AI fits into. It doesn't think for you. I just wanna remind people. You still need the humans need to do the thinking. And I think sometimes we we don't think.
Bruce Sundeen:Thanks, Shawn! Our guest has been Shawn Conley, soybean and small grain specialist with the University of Wisconsin Madison. You're listening to In the Pod, Soybean Updates, a weekly trek into the latest soybean information from NDSU Extension supported by the North Dakota Soybean Council.