Ag Geek Speak

10. AI and Human Interactions Pt. 2 OR How I Learned to Stop Worrying and Love AI feat. Travis Yeik

• A Podcast for Precision Agriculture Geeks • Season 2 • Episode 10

Share episode

Travis Yeik joins us for part two of our deep dive into the interaction of artificial intelligence and humans in agriculture, exploring current applications and future possibilities of this transformative technology in farming.

We chat about... 


• Weather forecasting is already utilizing AI to provide crucial planning information to farmers
• Precision agriculture applications include weed identification, disease monitoring, and targeted spraying
• Livestock operations benefit from AI monitoring of animal health, feed consumption, and production metrics
• Irrigation management systems use AI to optimize water usage and placement
• Biogenetics represents a revolutionary frontier where AI is accelerating crop development
• Protein structure mapping has advanced from mapping dozens to millions of proteins in just years
• Future AI applications could include profit maximization planning and automated stand count analysis
• Adoption challenges mirror previous farm technology introductions like auto-steer systems
• Data collection remains a critical challenge for AI implementation in agriculture
• Human oversight will remain essential even as AI capabilities continue to advance

Join us as we continue exploring how technology is transforming agriculture and what it means for the future of farming.

https://gktechinc.com/

Sarah:

And now it's time for Ag Geek Speak with GK Technology's, Sarah and Jodi, friends and I can't wait to get in the fields again. No, I can't wait to get in the fields again.

Jodi:

Welcome back to another episode of Ag Geek Speak, where I am joined with my co-host, Sarah Lovas, and GK Technology's own Travis Yeik, and in this part two of our episode about artificial intelligence, we are going to continue our excellent conversation that we started last week, where Travis helped us define what artificial intelligence is, what its history has been and, kind of like, what its limitations are, and we thought for this episode we're going to go into more depth about you know what AI tools are here now and how could they help us be more efficient as agronomists and farmers, Because there's a lot of capacity for AI to do that.

Sarah:

My question is what is AI doing in agriculture right now? How is AI helping us right now? Travis, can you help us understand that, maybe a little bit?

Travis Yeik:

Yeah. So there's several applications, I think, where we can definitely say AI is already in the workforce and some of those are, as we discussed here last week is weather right. Being able to have farmers know, you know ahead of time what is going to happen, whether that is you know two hours ahead of time or even two days ahead of time, is super important for making those decisions. Yeah, so weather is one we go on and there's a lot of precision agriculture for room for AI to grow and where AI already is, and some of those include weed control and disease monitoring. I think there's John Deere has a scene spray thing where they use AI in it and they use images to come out and say, hey, this is this type of weed and spray specifically just that weed. Yeah, we don't get into it a lot.

Travis Yeik:

But livestock health monitoring, I think is a big one. You'll see it in a bunch of dairies, whether it monitors milk outputs or the type of feed, uh, the animal when it comes into the, to the bunk, or getting water. That is as a big one for for doing animal production that way, and we talked other ways I guess would be irrigation, I think is another one. There's several companies doing irrigation, on determining when and how much and where to water, so some of that precision ag stuff there, that way again. Uh, then we, we can move out and go broader into such as, uh, being able to determine where we need products or or just, you know, shipping and distribution in that way with agriculture, that's, that's a, that's a huge part of agriculture. Um, and doing predictability of what, what crops are being grown and what needs to be grown, and doing commodities and stuff. Those are some of the major ones just off the top of my head. I think that is big in agriculture here already.

Jodi:

So, now that I think about this and where AI is and where it isn't in agriculture right now, I feel like we're seeing AI in places where there's already data in ag right. So like yield data, satellite imagery these are pieces of agriculture where we already have data. Is that why we've got AI there already, because we've got data, or are we seeing a push for AI? And you know, maybe some of the more challenging pieces of what I'm trying to say here is like is this really, is this being driven? And like where AI is and where its potential. Is it driven from low-hanging fruits or is it really trying to address, like, problems that exist in the market? Now, maybe that's too too like, uh, theoretical for this conversation, but like I mean, there has to be data to do ai right, but you also have to start with the theoretical idea.

Sarah:

And if in a, in a in a concept of where you want to go, you got to dream a little bit, right? You know, I have this problem.

Travis Yeik:

And it also calls money, of course.

Jodi:

Yeah, people have to pay for it, right yeah?

Travis Yeik:

Yeah, I think there's some you know statistics that you know. Like right now I can't remember what it was, maybe I had it in my presentation. Let me see if I had that. It was interesting, Like, let me see if I had that it was interesting Like where the money in AI development is. Yeah.

Sarah:

So Travis actually attended GK Technology's first precision agriculture symposium that we ever put on and he was a speaker for us. He spoke about AI and precision agriculture. It was a great presentation. It was fantastic. So if anybody's interested in that sort of stuff, come to the GK Symposium into the future, because there's going to be another one, not that I would ever, Travis, do you want to give your same presentation again?

Jodi:

Just kidding Not that I would ever throw out a shameless plug for our programming.

Sarah:

But there it is.

Travis Yeik:

There's your commercial so, yeah, we are spending like billions of dollars already and it's just crazy to think of how much energy is going to be put into ai in agriculture how do we keep that money focused on products that actually benefit decision making in the field from a practical standpoint?

Sarah:

and I and grant that. You know there has to be a certain amount of dreaming that goes into you know, what if I could do this, what if I could do that? You kind of have to dream about things that you want to make better, but what are some things that could go in our industry to help make sure that that money is going into places where it's actually going to affect the outcome for agriculture in a very positive way?

Travis Yeik:

I think there's like four or five like major agriculture sectors that we can think of on where where AI is going to be focused, and I think so for us, obviously it's precision ag, right. That's that's a big deal. So for us, obviously it's precision ag right that's a big deal and that's being able to use satellite images and be able to to use some of that precision ag stuff to make those decisions and enhance like efficiencies of how we're managing our crops and then I think another part is is robotics, right?

Travis Yeik:

uh, some of these, these big companies, will get into John Deere and Case and whatnot and they'll get into the robotics and being able to use the AI to automate stuff or to automate machinery. And then I think another one will be supplies and change and distribution. It's another major sector there. That's another major sector there. I guess we could include livestock, even though we are in the part of the different part of agriculture is what we're dealing with, and then I think the very last one would be some of these other companies such as Monsanto Say Bayer, Bayer, Bayer.

Travis Yeik:

Yeah, Monsanto doesn't exist anymore. Yep.

Sarah:

They were purchased by Bayer. Say Bayer yeah, sorry, Monsanto doesn't exist anymore. Yep. They were purchased by Bayer oh okay, Thanks.

Travis Yeik:

And then the other big one then would be companies such as Bayer and doing biogenetics right To breed and to genetically alter some plants and make them crop resistant, disease resistant, drought resistant, and that'll be a huge, huge money making adventure there for AI coming into agriculture.

Jodi:

To speed up the process of choosing which genes, which genetics, make up the best combination to maximize yield and survivability in a given area to maximize yield and survivability in a given area.

Travis Yeik:

Yeah, yeah, yeah. And really it's being able to have AI do that and be able to look at whether the genetics or the proteins of the plant. I was listening to a podcast here the other day of they were doing that and whether you call it GMO or genetic altering right, but it's just taking a plant and carrying it down about 10 evolutions, uh, or combining and saying, hey, we can see this crop, that that grows in this part of the country, and take out that specific piece of dna sequence and put it into a different plant that that needs, that needs that trout resistance or whatever it is.

Sarah:

So AI is going to be able to help identify those genes.

Travis Yeik:

Yeah, help identify them or help figure out. Hey, how can it be implemented into this different cell structure?

Sarah:

Wow, Wow. To actually help with the actual implementation of where it needs to go, I suppose in the DNA sequence itself when you splice it out and put it back in.

Jodi:

Can you send that link to that podcast?

Travis Yeik:

Yeah, it was a good one. It was this girl in Europe. Another really good one, though, is he's a really famous YouTuber guy and he's called Veritesium, and he does. He had one that come out on. What are the future benefits of AI and one that we're currently using right now? Is it's actually the same algorithm that I was talking about here last week, about the AlphaGo and the Moo0ero and the EfficientZero, where they're using this one that was able to play games now and they changed. Instead of playing a game, it is able to change protein structures.

Travis Yeik:

And this came out here. I remember reading about it. It was about eight years ago and we didn't have any protein structures, hardly mapped out. And so this uh or this professor, he came in and he made a program and he gave it to gamers and these gamers from all over the world were able to figure out this protein structure in like three days, something that we hadn't been able to do in like 50 years, right holy buckets and now they changed it so that not only can gamers do it, well, now this ai can do it, and within, like the last, I want to say two or three years.

Travis Yeik:

Right, we had maybe 20 proteins mapped out. Now we have like a million of them mapped out. That's amazing, so so we're able to, and not only can it map out the proteins, but it can create new proteins.

Sarah:

Could it change the protein that's there?

Travis Yeik:

So like let's say that you're lactose intolerant, although that's a sugar, but like but it's an enzyme, right, which is a protein then that controls and how we can process that sugar is a protein than the controls and how we can, we can process that sugar.

Jodi:

So I don't know a whole lot about most things, but I know some things about weed science.

Jodi:

Um, and that was a lot of what I worked on was like herbicide resistance and like the cool thing that I'm thinking about right now is like a lot of these like we talk about target sequence mutations that cause herbicide resistance and you know, for like ALS inhibitor resistance, most of those are target resistance.

Jodi:

But the point is is like what happens there is you have just maybe a single amino acid or like a single nucleotide change and that causes an amino acid inside of a protein to change shape and so maybe that changes the protein to change shape, so where an ALS inhibitor can no longer fit in that binding site and can't kill that plant, and what we always wanted like when I was in grad school just a couple of years ago, like the thought was like how can we predict where new mutations might occur that might cause herbicide resistance?

Jodi:

And if you could model or predict, you know what changes on the gene could change the shape of it and make it so the herbicide couldn't fit there anymore, you'd know like what, what could be likely to happen for herbicide resistance? But on the other hand too, like if you can predict how these proteins take shape, then you can hypothetically design better herbicides that fit these molecules to kill plants right. I mean, most of our herbicides that we work with are all about inhibiting enzymes in plants. So if we can model these proteins in weeds and figure out how to back engineer a molecule of herbicide to fit in that protein and make it stop, that could help us develop new herbicides too.

Sarah:

Or we actually have a weed geneticist at North Dakota State University who works on the genetics of of weed resistance, like all the time. Wouldn't it be cool if we could change the weed back so it was not resistant anymore and make it susceptible to the herbicides that are already in the marketplace? Yeah, right, like hypothetically, could you just change it back and it's ai that could find that spot and do that. That's, that's fun.

Travis Yeik:

I think about the scary side of it now like we're changing proteins right, which is a prion, and so we can change and create our own prions, and you know, so you got so mad cow disease right.

Jodi:

Yeah, yeah. So prions is like, if we can make a prion like mad cow disease, right, yeah, yeah. So prions is like, if we can make a prion like mad cow disease, like our ability to bioengineer, like weapons, would be drastically better so so you mentioned the other episode. If ai can learn how to kill humans, is this the approach.

Travis Yeik:

We should probably not transcribe this, so this doesn't end up in a large language model for ai to discover and get back at us cut this part out no but I was listening, you know, so that podcast I mentioned earlier about this uh girl, she was using some of that biogenetic engineering and so one of her things that she mentioned in the podcast was being able to take a plant right. It takes, you know, 15, 16 weeks or whatever for the plant to grow.

Travis Yeik:

Well, she was just cutting out the plant tissues and, uh, so you can get it within a day or two and change it and see how how these genetic markers react within the plant tissues, rather than yeah and I I think just to provide some context.

Jodi:

So like I read a novel about like barbara mcclintock's life, so barbara mcclintock, she was a geneticist that discovered jumping genes, um, and basically was helped helped to map out the corn chromosomes back in like the 1930s. But at that that point in time, right, that was before Watson and Crick. So like we didn't even know what DNA looked like. And so in the 1930s, the 1940s, we were still trying to figure out like it was during that time period that we even found out that one piece of DNA codes for a protein. Like that's not even a hundred years ago, that's almost a hundred years ago we didn't even know that DNA is coded for proteins. And now we're talking about using an AI technology to predict, like if we made a change in this tissue, you know how could it. It's crazy to think how fast this technology is progressing, based on how recently we've learned about a lot of these things. Just nuts.

Sarah:

Okay. So, jodi, I got an idea. And this might be scary, but here we go. I have this idea. What if we play a little game right now called if AI could do anything for me in the field? It would be. And then fill in the blank, and then Travis can tell us whether we're pie in the sky, ridiculous with the idea, or whether you know there's not a. You know there's no way it would ever work. So, okay, jodi, you go first. If. Think about farming, think about agronomy, think about whatever you want, if there was one thing that AI could do for you in the field, what would it be?

Jodi:

It would be profit maximization decisions at the beginning of the year. Help me decide what herbicides, what mix of crops, given like market conditions, what sets of crops should I grow, given my equipment makeup, given the prices of what I can have and then what I can hire, and then given the possible range of like prices I could get for the, the commodities at like. There's a lot of variabilities of it but we'll say I'll give it a data set of like what I think I can market it for. But if they could help me profit maximize and help me with that decision making, that's what I'd want it for.

Travis Yeik:

Okay, I think that's definitely coming'd want it for. Okay, that's definitely coming right. One of those algorithms I talked about was a mu zero and efficient zero, and what they do specifically is they look not only at the reward now with this observation, but it plans it out. So, 30 steps ahead. What will my reward be?

Jodi:

And so is, and if it can.

Travis Yeik:

And then also the other part of the model is saying hey, if we have this observation now and I take this action, what is my next observation going to be in my next one and my next one. And so it gets really good at planning the future and it makes those those future rewards based on what is happening and where you're, where your current situation is at now, and that's why it does really well and it does a lot of times better than humans at understanding these very complex models, because it has ran through these simulations millions of times and it knows, hey, based on this, this is where the most likely action will be, this way or, probability wise, this will be the best outcome based on my decision now.

Sarah:

That's fun. That's really fun Because that's a lot of different factors that go in and I feel like you know, in the spring of the year, you know we're really well. It actually starts like the fall before, like after harvest, when you're trying to figure out what am I going to plant next year and and what. Then you're going through the whole year of fertilizer, of everything else. So that's that's pretty exciting.

Travis Yeik:

And then obviously the tough part is, as we talked about, is this combining all these different decisions into having that data for it to make those decisions. So we, uh, we are. The algorithms are being developed, but having that data put together and processed, um, it could be ways in the future, but, um, I, I could foresee it happening.

Jodi:

And I have a question for later. I'm I'm going to put this on my later notes. Okay, sarah, okay, same question for you Okay.

Sarah:

So this time of the year as an agronomist, you know we're recording this over like right after Memorial day weekend and pretty much if you are an agronomist in North Dakota, Manitoba, Saskatchewan, Minnesota, South Dakota, if you were an agronomist and you've got a four-wheeler, you are putting on probably 100 miles a day right now, at least on your four-wheeler, if not more.

Sarah:

It's go time and one of the the parts of scouting crops at this time of the year that I find time consuming and annoying is stand counts. Now I am aware that we can go out and we can log stand counts with a drone. You know you can take a picture. It's high, detailed. You can get those pictures to a certain extent. Like sugar beets can be really hard if they're in the cotyledon stage because they're so small that even sometimes the resolution isn't good enough on drones UAS. But what I think would be really great is if we could get accurate stand counts before and after an event and then have some sort of a decision model in there that can take into effect the previous weather, the future predicted weather and the market conditions to determine whether we should replant or not.

Travis Yeik:

So that is a lot right, that's. That's like several different systems coming into one. I think about what was. We were in grade school and you had the man versus the machine. Right it was. It was this guy digging a tunnel or a for the, for the railroad, and then he had a machine and he was trying to beat the machine. You guys remember what that was called.

Jodi:

No.

Travis Yeik:

Oh, okay, maybe this was a Wyoming thing. No.

Sarah:

You're younger than me too. I mean, when I was in elementary school we had Apple II computers with basic language.

Travis Yeik:

I'm going to look at that real quick, just because I'm curious.

Sarah:

I know you guys have heard this the chalk all looked pink when I got back to my classroom after computer class because of the green screen on the computers. Oh yeah, and, by the way, we actually had chalkboards with chalk.

Jodi:

Oh. I went to an elementary school that had about 50, 60 kids total. So yeah, all we could afford were Apple computers and chalkboards too.

Sarah:

Then I moved and look at how he turned out. Jody. The legend of John.

Travis Yeik:

Henry.

Sarah:

What was it called Travis?

Travis Yeik:

So it's the legend of John Henry, and it was this drill that could drill faster than any man, and he was drilling through a tunnel. And so he went and said you know what? We don't need these machines. These machines are going to replace us. Right, I can beat this machine. And so he went up against this machine and I believe I'm trying to remember the outcome, but I think he lost.

Travis Yeik:

But it's the same thing, right? I'm thinking about automation and how we have assembly lines. Right, when they came and I think a lot of people were scared at that time too that assembly lines are going to replace jobs, and in a lot of cases, they did and I think AI with robots, I think they could be there someday where they take over these processes. It's just an assembly machine that can move around, and so, whether you're pitching hay or doing manual labor tasks, fixing fence and stuff like that, I think you could have an ai someday that has the ability to to move um and make these very simple um task oriented decisions on putting a post in the ground or doing this or doing that, and I think think in this case, it could be in your example, driving a food out to this place, or a drone, flying a drone out to this location and being able to make these tasks and bring back the data for some of these tasks.

Travis Yeik:

And I think once that happens, then AI will be able to be carried along quite a ways, because now we have that data that is being automatically generated by robots, by AI itself, and it just compounds right even more and more. And already, who knows when that's going to happen? That could be 50 years down the road, I don't know. But you know, robots and ai are getting a lot more intelligent even now, and so, yeah, if we have a moving assembly line that can do manual tasks like that and then being able to come in and process that data, which which we are already doing in some sense, um, you know, with satellite imagery or aerial imagery, and processing that data to make decisions, I feel like it could be. That is a lot farther away, though. It may not be in our lifetime, but yeah, someday I think it could be.

Jodi:

That's you know you brought up. You know robots would collect the data instead of having to rely on humans to collect the data. I never thought about that. Right, how fast or how much better will AI models get when they don't have to rely on imperfect humans to collect data, but they can just rely on robots to collect data? That's interesting because that's what I keep thinking back to Someone that did a research project in grad school. Like it's it's not easy to collect good data. Like it's a hard process.

Travis Yeik:

It's a lot of yeah, it's a lot of resources, it's a lot of time and, yeah, a lot, yeah, and, as you said, it has to be accurate, right? So you have one person that does it here versus another person that does it over there.

Jodi:

They're going to do things entirely different ways and maybe that I think that leads into my next question too is like, okay, so in order for all of us, because at the end of this, like I'll just cut to the chase and say I don't think that AI will replace agronomists or farmers anytime soon, but I do think, you know, with the legend of John Henry and thinking about assembly lines replacing humans and like the Luddites deciding that they didn't want to embrace technology, but it seems to be the best.

Jodi:

At least, looking at history, it seems to be that those that can utilize it, learn about it and grow with it are the ones that quote unquote win in society. So, thinking about that, you know, like us, as farmers and agronomists, what can we do now to start growing with AI? And I God, that sounds like such a weird question, but like, what can we do now to make all of us better right now, with our current you know what we do in decision-making and farming and agriculture. What can we do now so that we can take advantage of this as technology progresses, collect better data? I don't know?

Travis Yeik:

well, I mean, that may be part of it. I, I think for me, um, it boils down to one of the very basic things is accepting it right, like anything new that comes in it? It's just like this is never going to be right, or I'm against this, like I don't know. It took me it was probably 10 years after your smartphone came out that I had a smartphone. It's been within maybe the last eight years. I had that. I switched from a flip phone. Even though I'm a developer, I'm kind of anti-technology.

Sarah:

Even though I'm a developer, I'm kind of anti-technology. Do you think that's just because you're skeptical of new technology that's coming out? Do you think it's because you question the validity of what the claims are? And the reason I ask that is because you know we find this in farming all the time. Where so, for example, we had auto steer. Well, it wasn't that long ago and I remember you know my dad's generation saying who is so lazy that they can't drive their own tractor? And why would you ever not want to drive your tractor? And now, like, honestly, most farms right around here in North Dakota, they usually have.

Sarah:

God bless them but a retired guy that wants to come out and do tillage. Well, part of that tillage application is going to involve auto steer, and if the auto steer quits, that guy is calling and he's going to be like, hey, I don't know what to do, the auto steer isn't working. It's like that's all right, grab the wheel and drive what? No, yes, you know, and so I think that's that's yeah. How do we? How do we handle that skepticism?

Travis Yeik:

well, I think what you bring in is great point is the cost of what it is right, not just the financial cost of it. I mean that is a huge deal like does using this technology offset what I, what I would be doing otherwise?

Travis Yeik:

does getting that uh you know that rtk gps and using it to it for spraying or for auto steer. Does that offset my cost of what this technology is and using it or the support if it breaks down? Right, you have to have somebody there that can give you the support and know how to use that technology, and that's another turnaround as well. I think you know to use that technology, and that's another turnaround as well, I think. And with AI, for now or even in the near future, we're going to need that human oversight. So even with a robot, I would think if you had one, you're not just going to let that thing go and, you know, terrorize the farm and all your animals. You're going to have to have someone watching it and making sure that what what it is doing is right, cause if it breaks something that could be way more costly than what it's actually worth.

Sarah:

That's a really good point. And you know, going back to the auto steer example, yeah, we're getting to have some autonomous tractors in the field, but you know we don't have the operator that far away and I realize we're getting further away from it to the point where we can trust it. But think about how long auto steer has been in a tractor to get to this point and it's not like it's driving down the highway. You know, going between fields, you know going between fields, uh, and in some instances you know with the with the grain cart, that that tractor with the grain cart behind it is not loading the truck, it's just going over to the docking area and then the truck pulls in and then the truck driver gets out and jumps in the grain cart and loads that. So it's it's. It has taken us so long to get to this point in in the whole concept of autonomous tractors. So the idea that we're not going to have oversight in AI, at least for a while, is blows my mind. I mean, we're going to have to have that.

Sarah:

Yeah, that's such a great point, sarah. Well, and Travis kind of brought that point up. But you know, another question that kind of goes with this is it. It feels like Travis, when, when we visit about what you're doing with AI. You are actively training or teaching that AI model how to do something, so it has to learn. Teaching that AI model how to do something, so it has to learn. So in a way, it kind of almost feels like it's a child that you're actually like training how to do something.

Travis Yeik:

Yeah, would you trust your four-year-old to go out in the tractor by himself? I don't know.

Sarah:

Yeah, is that way off, though, on the way that I'm thinking about that, that you are actually like teaching that AI model how to do something.

Travis Yeik:

Yeah no I think that's one of like the one of the cruxes, I guess, in teaching AI is that we are teaching every AI from infant stage, right, and there's not an AI that's already like 10 years old that we can just, hey, okay, now teach you to do this, and teach you to do that, and make these connections, and I think that's part of the point, or, you know, one of the struggles with ai, and I don't see this perhaps happening, and then I don't know, let's say, the next 20, 50 years, even of being able to make these important decisions that humans do, that require this oversight.

Sarah:

Imagine what this is going to be like when AI reaches its teenage years and young 20s, when it's going to make a few bad decisions along the way. Uff da maida, huh prions, just kidding pre-ons oh, that's.

Sarah:

this is a really interesting conversation. It's. There's so much here to think about. Hopefully, ai is going to be a great thing for agriculture into the future, but it really does sound like it is up to us as humans to decide what that's going to look like, how we train it and how we choose to incorporate it into our daily lives. Travis, I want to thank you so much for this great conversation. The last two episodes that we've done here this one and the one before just it was very insightful and there was a lot of thoughts, so thank you for that. Jodi, as always, it's super fun to co-host with you and I guess with that at GK Technology. We have a map and an app For that. I can't wait to get in the fields again.

Jodi:

No, I can't wait to get in the fields again and an app for that.