Podcast Extra: Rooted in Organic Podcast - What AI in Ag Can Look Like with John Kempf

You're listening to Rooted Inorganic Podcast. I'm Mike Klein and along with Brian Wood. Today we have a returning guest. He's been on a couple times before. One of the early episodes, John Kempf, who is the founder of AEA Advancing Eco Agriculture and is very involved in the regenerative ag movement, and he has some exciting new things we want to talk about today. So, John, thanks so much for joining us again.

Yeah, thanks Mike, and thanks, Brian, for having me back on. I've always enjoyed our conversations.

So, yeah, John, usually we start these with a good introduction and that usually takes us down wherever we wind up. But since you're a returning guest, I think, you know, folks should be aware of who you are and if not, I mean, you can certainly give a quick introduction, but I think it would be most interesting to just talk about what, what it is that your, some of your latest projects you've been working on and how that's advancing agriculture, what your, what your goals are behind them.

Yeah, thanks, Brian. There's, you know, the world of agriculture, and particularly this world that's now being called regenerative agriculture or biological agriculture, whatever you want to call it, has changed so much in the 20 years since I've been involved. And even just in the last half a dozen years, there has been this upswelling of accessible information where now there are dozens of YouTube channels, there's all the information out there. And you know, this method of farming differently, of a different type of agronomy that seeks to enhance soil health and enhance ecosystem health generally tends to be much more, somewhat more management intensive sometimes and certainly more knowledge intensive, more information intensive. It's like there's a need to know more and to constantly be asking different questions and be asking better questions. And while it's true that we live in an information age, we have access to information in ways that we've never had before. It's also, it's sometimes hard to know where to find that information because the best information relative to this type of agriculture, a lot of it is being put on the Internet very fast, but the best of it is still held in books and held in people's experience. It's not on YouTube. There's a lot of great information on YouTube, but there's still a lot of information that is more difficult to access. And so one of the things that I've been thinking about for quite some time is in our work at advancing eco agriculture, our, our job as a consultant is to have read the books, to have benefited from the experiences of hundreds of growers that we have worked with and to bring that to the farms that we are coaching and that we're working with and to, to help to give them access to that information. Well, the problem with that is, is that if you really want agriculture, regenerative agriculture, to scale rapidly, then there are limitations to how rapidly we can bring people on board, how rapidly we can train them, how rapidly they can get exposure to that type of experience. And there are limitations to how many people we can support, how many people we can help. And so for the last decade I've been thinking about how can we democratize knowledge, how can we digitize it, how can we make it readily accessible? And I first sat down and outlined a sketch, a scope of what at the time I called an AI agronomist. And well, AI today turns out to one. It's developing extremely rapidly, but it's developing very differently from what I had envisioned. And I don't particularly care for the term AI anymore to describe what it is that we're doing, but for the last several years our team at AEA has been working on, on digitizing all the knowledge that we have collected and making it more broadly readily accessible. So in just at the beginning of this week, well, actually end of last week, I guess, we launched our, that regenerative AI agronomist that I first envisioned and sketched 10 years ago, 11 years ago, and we're calling it Field Lark. So it is the features that are currently available. It is a chatbot that you can use and you can ask it any type of questions about. It's focused obviously on agronomic questions, but any type of soil and plant ecosystem management questions related to agronomy, related to what types of biologicals, what nutrition interactions, what causes different diseases and insects. You can ask it any of those types of questions and the answers that you get back are, I mean this, this is technology that is not dissimilar to ChatGPT or Gemini or Grok or these other AI tools that are out there. But this is, we are not. When you have, when you're using those general chatbots to scour the Internet and to come up with publicly available information and the results you get from those versus the results you get from fieldlark are, they're not even in the same league. We're not making slight incremental improvements of 10 or 20%. We are making step change improvements and getting much higher results back than we used to, than we're able to get with those general AIs. So that's pretty exciting. We're quite excited about that. And then there's some new product stuff that we've been working on the last year or two that we, that we're now just releasing that I'm also pretty excited about. So there's, there's, there's a lot of momentum, a lot of energy.

Yeah, I'd say so. And my first question on the AI would be or the chatbot or field lark. I think I'll just go with that for consistency sake. So, you know, when I think of what is out there in terms of, you know, human agronomists, so to speak, you know, their basis of their knowledge usually comes from varied sources, whether that be, you know, university educated or experiential, you know, their local experience in the area or I'm trying to think of the, the author that wrote some of the books and I'm drawing a blank right now that Dr. Paul always referenced. But either way, that, that's kind of where my question starts with the AI is, you know, is there a basis, a base point that feeds into the intelligence that it has or where does that come from?

Yeah, it's a great question. So the base points are a number of written resources, materials and books out of my library and the types of books, I'm not sure who you're referring to in reference to Dr. Paul, but books by Acres USA and just generally the scientific reference books that in the open published literature that are aligned with regenerative agriculture, that talk about the interactions of microbes and minerals and soil and organic matter and so forth. So it is published information. And then the other significant data source is all of the database that we have developed in our work at advancing eco agriculture, where we have this large database of soil analysis and SAP analysis data and the crop reports that came with those. What were the yields like? What was the performance like? What were disease and insect susceptibilities, what were, what is the microbiome like? And what does the DNA profile of the soil and the plant look like? And you combine all that data together and the, this is one of the places where computers and machines in some ways have, they have complementary analytical abilities to humans. Because we have, we can only hold so many data points in our mind at once and compare so many variables at once and they can compare many more. So all of a sudden you start looking at this data and you realize, oh, wait a minute, for this particular crop, every time we have this disease show up, we have this particular nutritional profile show up, or we have this particular microbiome profile show up. And so there is an abundance of very rich, high quality data there that feeds into the background of this model. That is. And that is information that's been collected over the last 20 years. But I'm not aware of any other similarly integrated database that exists.

So it works similarly to any other chatbot, but you would have given it a different baseline to begin with.

It's got a completely different training set. It's not trained on the information that is available on the Internet. It's trained on the information that we have vetted and proofed and said, all right, this information is worthy of including. So you know the old. You know the old saying of garbage in, garbage out. And the only way. The only way you get something better than garbage out is to not put garbage in.

So it does not scour the Internet like the others for the information or does it, using certain parameters that you would have put in.

It uses the parameters that we have put in.

Yeah, I'm gonna definitely play with it.

What have you observed so far, Mike? What have you. What have you. Have you played with it at all?

Yeah, I did. When you had sent me that link a couple of weeks ago, I started, like, playing around with it, and my wife actually played around with it more, but it was. We have some brown spots on strawberry leaves. So she went into that and came up with a whole plethora of problems or, you know, things we should be doing. And then also asked on, like your strawberries runnering, like, do you let them run or do you not let them run, or. And it was. The response was quite different than a Google search.

Yeah, yeah, that's. That's exactly what we're looking for, Mike, is to be able to. You know, one of the things that has always intrigued me, and you know this from experience, if you. If you think about the knowledge base of a dairyman who is very good with cows and who has 20 years of experience treating cow health organically, their knowledge of the physiology of the cow and the husbandry and what is required to care for cows, versus a homesteader that has had one cow for five years, the delta, the gap.