LSU's Thanos Gentimis Discusses Digital Agriculture and AI in Farming

In this episode, President William F. Tate IV speaks to LSU assistant professor in experimental statistics, Thanos Gentimis. He discusses his field of expertise, which is data analytics with a special interest in machine learning and neural networks. Thanos created the first digital agriculture class at LSU and discusses how AI will soon shape the world of agriculture and farming.

Thanos received his Ph.D. in theoretical mathematics (algebraic topology) from the University of Florida in 2011. His dissertation was titled: “Properties of Groups at Infinity.” He earned his master’s degree in theoretical mathematics from the National and Kapodistrian University of Athens, Greece, where he also received his bachelor’s degree in Mathematics and Education. He is fluent in Greek (his native tongue), English, and has a passion for playing chess (he holds the Greek ELO rating of 1880 as a competitive player).

Full Transcript

Related Links

 

 

Transcript

[00:00:00] President William F. Tate IV: Dr. Thanos Gentimis is an assistant professor of Experimental Statistics at LSU. His fields of expertise include data analytics with a special interest in machine learning in neuro networks. He created the first digital agriculture class at LSU.

[00:00:19] Thanos Gentimis: Thank you, Dr. Tate. Glad to be here.

[00:00:20] President William F. Tate IV: You know, I love to hang out with statisticians. You have all the answers.

[00:00:24] Thanos Gentimis: Really?

[00:00:24] President William F. Tate IV: That's what they tell me. So we're gonna tee off, tell us about your journey to Baton Rouge and LSU.

[00:00:30] Thanos Gentimis: First of all, I am from Greece. That's where I got my bachelor's and masters in mathematics, with a specialty in math education, so we have something in common.

[00:00:38] President William F. Tate IV: Nice.

[00:00:39] Thanos Gentimis: And then I got lucky and got a scholarship for UF. That's where I finished my PhD in theoretical mathematics. And then me and my wife started chasing each other, trying to solve the two- body problem. And then finally in 2017, I got hired here. I was a spatial hire by the way, and here I am.

[00:00:58] President William F. Tate IV: Well, that sounds like a great journey of two great academics finding their way. We're really excited to have you here at LSU. Talk to us about your research focus.

[00:01:07] Thanos Gentimis: Right. So, as I said, I started with theoretical mathematics, but pretty soon I discovered that I'm more applied. I got it interested in artificial intelligence and machine learning specifically when I was in North Carolina, and that has been my field ever since. Now, I started with applications of AI and health informatics, but when I joined LSU, I saw that there was an opening in agriculture. So, that's what I've been doing since. Digital ag is basically my field.

[00:01:35] President William F. Tate IV: So, help people understand you jumped from theoretical mathematics to AI. What got you interested in that?

[00:01:42] Thanos Gentimis: Honestly, when I was at NC State, there was a lab that I was working in, the Lab of Analytical Sciences and there were various groups of people working there. And we all knew what everybody else was doing, and everybody was trying to kind of jump into this new idea of AI, machine learning and all that. And I was in those meetings and I saw that: A, it was interesting, and B, it was actually successful. They had good prediction rates. They had, you know, revolutionized the way that we do analysis, right? So I just had to learn about it, and especially since I had the math background, it was relatively easy to understand the math underpinnings. So the transition was simple. It was mathematically interesting and very useful. So there we go.

[00:02:28] President William F. Tate IV: I like the fact that you said it was simple.

[00:02:30] Thanos Gentimis: [laughs]

[00:02:30] President William F. Tate IV: Help us to understand a bit-- how is AI solving problems in the world around us? You mentioned health analytics and how you're working in agriculture.

[00:02:39] Thanos Gentimis: AI has optimized so many things. First of all, computer vision. There was computer vision before AI and afterwards, and we're talking now from satellites to endoscopy when we're talking about health. Second, all your social networks, your Google engines, all of these are driven by AI. But AI has also optimized manufacturing. It’s going to replace driving, right? All of this has some sort of AI in the background. So yeah, I believe AI is changing our lives.

[00:03:10] President William F. Tate IV: So you used the word optimize several times. Help us understand how AI might optimize something, but it might present a danger or ethical challenge.

[00:03:21] Thanos Gentimis: Right. So, the optimized portion is basically the machine learning techniques can create predictions based on complex patterns that human beings cannot see, right? So, they look into the data, they find a pattern that you and I cannot immediately see and say, "Oh, if you have all these characteristics, then the answer is blue, red, whatever." And their prediction is much better than our prediction in a sense. The problem is with a lot of these techniques, we just don't know how it happens. It seems like a black box. Even though we are the creators of the algorithms, we don't know exactly what's going on in the background. So you may have a lot of, you know, ethical problems there. I've seen a lot of research on that area where, for example, minorities are excluded because they're underrepresented in data, okay? So you have these problems that are not really tractable by humans because the machine code is also not so attractable by us. So that can create a lot of issues. That's why I strongly support, you know, research in that area. The ethical implications of AI.

[00:04:23] President William F. Tate IV: Did you ever think we'll get to a point where the AI actually flags the ethical challenge? I mean, that would be some sophisticated coding.

[00:04:30] Thanos Gentimis: That would be interesting. That would be interesting.

[00:04:32] President William F. Tate IV: I just solved the problem.

[00:04:33] Thanos Gentimis: [laughs] That would be very interesting, actually. If we could, in a sense, train an AI to check itself. The research is not to the point that we can actually understand what we're doing, and I think we need to spend more time on it. But you're right, we could do that, too. Hopefully.

[00:04:49] President William F. Tate IV: So, help us understand the difference between machine learning and deep learning in AI.

[00:04:54] Thanos Gentimis: Okay, so AI is the big umbrella, right? It contains, you know, machine learning, obviously, but ideas from robotics, computer vision, things like that. Now, machine learning is, as I said, a subset. All the techniques that machine learning has are techniques of AI. But let's say you can think of neural networks, support vector machines, random force, so applications in data. And then deep learning is a subset of machine learning. Let's just say deep because it looks complicated and also deep because it can, in a sense, identify deep patterns. Do something like, you know, a deep thought like our brains can do. One is a subset of the other with deep learning being all the way inside, and AI being the overarching umbrella.

[00:05:39] President William F. Tate IV: One of the things that you focused on in your research is helping to make farming more efficient. And now you've created a course, digital agriculture course. Can you share with us what is the digital agriculture course all about?

[00:05:54] Thanos Gentimis: So, the digital ag course is basically an introduction. Ideas from machine learning, data analysis, a little bit of image processing, computer vision and all that. Because you see, in agriculture nowadays, we have the introduction of drones, we have the introduction of smart machines, we have satellites, and then we have the Field of Precision Act, which is basically taking all that and creating variable application rates or farming plans. So all that exists, but it has become very complicated. So now we're reaching the point that the data being accumulated cannot really be processed by humans. So that's where the digital part comes in. We have more and more people using codes to clear the data or make sense of it. And so, the agronomists, let's say, of tomorrow, will have to know besides that ergonomic stuff, they need to know Python. They might need to know databases, they might need to know image processing and things like that. So, digital ag is a gentle introduction to those. I'm not going deep in all these ideas. I just give them a taste. I do cover practically everything you need to be dangerous, as I say [laughs]. You learn this and then you can start going deeper.

[00:07:05] President William F. Tate IV: So what are some prerequisites for that? Would you need to know anything before you came in that class? I mean, maybe something about the field? Do you need any technical skills, statistical background?

[00:07:13] Thanos Gentimis: It helps if you have a stats background, maybe like one or two simple stats classes. It really helps if you have taken coding, but it's not a prerequisite. And on the other end, it helps if you know a little bit about agronomy. Perhaps you've taken a GIS class or perhaps you've done some sort of precision ag class. But, in its inception, the class had no prerequisite. It was literally, let's try to get a couple of people interested in this.

[00:07:41] President William F. Tate IV: So, when you think about the class, when it's over, what do you want your students to be able to do? I know you said they might be dangerous because they just know a little bit, but is there something that you would imagine that they would be able to do as a result of having the experience?

[00:07:55] Thanos Gentimis: Right. So, first of all, mine is a very introductory class, so what ends up happening is people like it and start taking more classes. Especially, I've seen a lot of agronomists taking some computer science classes, some statistics classes. So, the first thing is they get over their fear of, this is brand new, I don't know how to do it, and I'm intimidated. So, they dive into that. The second thing is they know that these techniques exist. So, besides their typical agronomic stuff, they have them in the background. And finally, it's easier for them to connect with the experts in the field. So even if you don't become a data analyst, let's say, it's much easier to talk to one. Show that this is not some alien field that nobody can understand and build bridges with subject matter experts on both ends.

[00:08:43] President William F. Tate IV: Help people understand how AI technology and agriculture is really important for the state of Louisiana.

[00:08:48] Thanos Gentimis: So, in Louisiana we have some special crops like rice and sugar cane. A lot of the research that happens on agriculture does not directly translate, okay? So, we need to have specific tools for Louisiana. We have so many different soil types. We have the Louisiana around the Mississippi, and away from the Mississippi, we have north and south. There's so many variations, so many varieties on the products that you plant. We have the hurricane season, we have this weather that is crazy, alright? There's so many things that are ready to be optimized. And right now, we just basically don't have enough data to do so, or if we do, we don't have that much adoption. Another thing that's happening in Louisiana is the change in the adoption of techniques like drones and farming, smart farming. If you go back 20 years, 1% of the farmers had a drone, and now we're at 13%. So that's going to change. So, as we move forward, we'll have more and more data, and that data's going to be Louisiana-based. Precision ag, with the help of digital ag, can actually optimize the whole process. And I think that's the future.

[00:09:59] President William F. Tate IV: Say a little bit more about the future of AI in ag. I mean, where are we really headed with this in your -- I don't wanna put you on a spot in terms of predicting the future, but you started down a pathway of how these different kinds of ways of learning are merged. Where's all this headed?

[00:10:18] Thanos Gentimis: What I see is Agriculture turning into this fully automated process where practically everything is handled by some sort of smart machine. You're not gonna inspect your field anymore by walking around. You're gonna fly a drone or look at the satellite imagery. You're not gonna create a variable rate plan by asking consultant. You're probably gonna feed it into a machine. You probably won't have somebody in the tractor anymore. It's gonna be like completely automated. I mean, if we're talking about automated cars, imagine like automated farm equipment, which is much easier because there are no obstacles over there. The process after things are harvested, like optimization and the mills for sugar cane. Again, I don't know if this is gonna happen in the next 10 years, 20 years, or 40 years, but that's probably what the future is gonna be like.

[00:11:04] President William F. Tate IV: Help us to understand how you think AI could help Louisiana be a better competitor in the marketplace.

[00:11:14] Thanos Gentimis: Right now, in Louisiana, we have some commodities that are very, very important, like lumber and rice, sugar. None of them is optimal. Right now, sugar cane is subsidized because it's a little bit expensive to grow it, but if these techniques actually work and you can cut down on fertilizer, you can cut down on irrigation. Then, basically the production becomes cheaper. The price will stay fixed, so there you go.

[00:11:42] President William F. Tate IV: Now, when I listen to you talk about that, it strikes me that the type of people who will be involved in agriculture in the future might be different in terms of their educational background, their expertise than you'd have had in 1960 or 1990 or 2000. Help us understand what that person's going to look like. What is that training gonna look like for that?

[00:12:10] Thanos Gentimis: Right. And this has been actually one of the reasons why this digital ag initiative started here at LSU. We recognize that the consultants, we have extension agents, consultants, that work with LSU and work with the farmers. They now had to answer questions like, "How do I make sense of what my drone says? How do I make sense of what my smart machine says?" So, besides the agronomic practices, you will need to know a little bit more about databases, and you'll need to know a little bit about variable rate planning and all that. So you'll need some sort of computer science background. We're gonna be introducing classes like digital ag and more and more people will take whatever they're doing, like math, statistics, agronomy, science, and AI based classes, I see that at least knowing a little bit about how to code, how to manage databases. This will be second nature. Just like 30 years ago, every agronomist needed to know statistics. 10 years from now, every agronomist will need to know a little bit about computer science. From databases to image processing, to coding.

[00:13:18] President William F. Tate IV: Wow. Well, I guess Mendel was king, so to speak, in statistics, and now it's gonna be the digital revolution.

[00:13:26] Thanos Gentimis: Perhaps.

[00:13:27] President William F. Tate IV: That's exciting. Now, earlier you talked about the hurricanes. I had experienced my first one last year. It was quite something. Help us understand how this digital revolution in AI is gonna prepare farmers to recover from natural disasters like hurricanes.

[00:13:45] Thanos Gentimis: So, there's already some progress in utilizing drones to assess the problems after hurricanes. So that already exists. We're thinking about then optimizing that and keeping it as some sort of record that you can then use next year to predict yield based on these issues. And that could also be used for claims, insurance claims. And it can also be used for future prediction on in terms of where I'm gonna put my money on, like when you have a big field. Should I put my money on this side or that side? So having that assessed by a drone rather than by a person will optimize basically your yield next year. The prediction of the hurricane and its results can be, in a sense, known prior to the hurricane coming. So that will help the farmer make sense of where they need to again, put their money. So, if you know, for example, that that field is gonna be directly in a path or that field gets flooded more easily, perhaps you don't spend all your money there and you diversify your portfolio and things like that. There's definitely some value in first knowing where the hurricane is gonna strike and then afterwards completely assessing what the results were.

[00:15:00] President William F. Tate IV: That's a motivation to actually even learn. How can AI technology impact agriculture and potentially influence what's happening with climate change?

[00:15:10] Thanos Gentimis: AI has been hailed as the technology that will help us optimize things like water consumption or fertilizer application. If we move to smart machines and all that, we're talking about a reduction in the consumption fuel, right? So, all of these are moving towards like a greener agriculture in terms of like how the energy is being used. Optimizing planting date windows reduces the amount of extra effort that you need to put in order to grow the same crops, right? And that extra effort translates to less fuel consumption, less water consumption, less nitrogen, less fertilizer overrunning our Mississippi River and all that. And right now, we are in the early stages of that. But everybody's pointing that direction. We need to do something to reverse the issue. And that might be something that we can do. Industry's hiring like crazy. If you look at the big players in agriculture, they are acquiring those startups that have to do with applications of AI and agriculture, and they're trying to staff them with people that know both ends, like both the agronomy and the computer science. But also, you see people from computer science jumping into agriculture, which was not something that happened the last five years because there's money involved, okay? All the consultants now, the farm consultants start putting in their portfolio some knowledge of smart machines or computer sciencey stuff, or databases. So now is the time, honestly, if you are an agronomist and you're thinking, what kind of future should I, if you're, let's say, finishing up with your bachelor's-- I would strongly advise you add a little bit coding into your resume. Maybe try your hands with Python. Figure out besides, you know how to fly a drone, how to analyze the data from a drone. That will make you very marketable nowadays. Plus, if it doesn't work out with agriculture, you'll still have a skill that is on a very high demand right now.

[00:17:12] President William F. Tate IV: That's good to hear. Since we are an agriculture mechanical institution, right?

[00:17:15] Thanos Gentimis: Mm-hmm.

[00:17:16] President William F. Tate IV: So, fun questions.

[00:17:18] Thanos Gentimis: Okay.

[00:17:18] President William F. Tate IV: I understand you like chess.

[00:17:20] Thanos Gentimis: Mm-hmm.

[00:17:21] President William F. Tate IV: How did you get into chess and why do you enjoy it so much?

[00:17:24] Thanos Gentimis: Well, okay, so I got into chess when I was eight years old. My mother took me to a chess club. I think it was probably like a cheap babysitter [laughs]. But I loved it. It was, uh, it spoke to me. And since then, I've been in various clubs. I played competitively for a little bit. When I came to the United States, I couldn't play anymore, but I kept at it. I was in the UF chess club, then I formed the Florida Poll Club. I'm here at the LSU Chess Club. It's also a great way to connect with my kids now.

[00:17:56] President William F. Tate IV: Well, I understand you had some questions for me.

[00:17:59] Thanos Gentimis: So, I know that you're very interested in applications of AI in general. I was actually part of the committee that you had set up to explore AI at LSU. So, what do you think about it? Where are we going with AI? Why is there this drive?

[00:18:15] President William F. Tate IV: Well, partly in response to your question, you really answered it. I mean, it really is underpinning almost everything we do. We can't pick an area, whether it's healthcare, and we're looking at healthcare applications. And if you think about what I've been promoting here, this Pentagon is agriculture. Clearly, you've made the case that AI undergirds agriculture is biomedical science. With access to health records, Epic is now driving a lot of what happens in the biomedical area. So, there's a massive amount of data. Clearly in the biomedical side, it's a big deal. Coast is our other area. This is a massive amount of data, both visual and historical. AI clearly is going to be better at giving us insights into the coast. And then we're very much interested in defense. ROTC's one part of it where we want to have people really understand how to use those techniques if they're gonna be in military science, but also cyber security. And on the cyber side, there's a lot of data that could be gathered, and clearly AI could be helpful. You need the edge in that area. And AI is very important for our last part of the Pentagon: energy. Energy is foundational to the state. It's the largest industry. We want, as you use in the language, optimize our opportunities there. It is both an engineering challenge, it's an economic challenge. It is a challenge around people and communities. And so, when we think about the various data sources that are gonna be related to the work in energy, AI is gonna be very important for us. And so, pulling it all together in that Pentagon, really undergirding that, are the people and really the methodology. And AI is an important methodology to help us understand the world around us. And so, I see AI, computational science, all of our abilities and statistics. Your background is perfect for what we're trying to get done. Just really pulling together so we can understand better the world around us in our A, B, C, D, and E areas. So, it's quite exciting for me. And what I would love to see are more people like you at LSU. We need more people with strong computational skills who really want to help students learn and then be able to apply and then help secure and protect the state of Louisiana across those five industries.

[00:20:37] President William F. Tate IV: Well, this has been a wonderful conversation and I'm excited about the research you're doing and the work you're doing with our students. I want to thank you for coming on the podcast, and I'm glad that you and your family are here adding value at LSU and Louisiana.

[00:20:53] Thanos Gentimis: Well, again, thank you for having me here. This has been an excellent experience and indeed let's keep the communications open, and I promise to put the best effort to help with promoting, let's say, AI here at LSU.

[00:21:16] President William F. Tate IV: Thank you.