The Moos Room™

00;00;11;03 - 00;00;40;04
Speaker 1
And welcome to The Moos Room. Brad, here today with you back from a little bit of a time off in, in sunny Florida, and I came back to 12in of snow in western Minnesota. So it was a wonderful time away. Saw some good cows and Florida. Most people don't think that Florida has a lot of cows, but they actually do a lot of beef cows to cows everywhere in in in Florida.

00;00;40;09 - 00;01;13;06
Speaker 1
However, today I wanted to talk a little bit about AI and not artificial insemination, but artificial intelligence in agriculture in the dairy world and where it's going. But first, I must admit something. If you remember, if you listened last week we had a special guest host talking about cow calling in one of my research papers that we did here at the University of Minnesota on our dairy heard, and that was actually AI generated.

00;01;13;08 - 00;01;43;01
Speaker 1
It's just wild where things are taking us in today's world. So how did this come about? Well, I was at some meetings. I'm a editor for the Journal of Dairy Science, so a research journal and we had some meetings in Saint Louis in February, and we were talking about AI and how I could be used or not used in research or research publications.

00;01;43;01 - 00;02;13;09
Speaker 1
And one of my fellow editors had told me about a new thing from Google. It's called notebook LM what does it do? Well, it does a lot of things. So you could take, for instance, I took some research papers. I basically uploaded them to a notebook. LM it can give me a summary of this research paper. It gives me a timeline of the study.

00;02;13;11 - 00;02;54;17
Speaker 1
It will give me some frequently asked questions about the study. It will produce a study guide if I want on this paper, and cow culling, the one that we had talked about. And it's quite accurate. Well, what else does it do? It can give you an audio overview of this study. So that's what I did. I uploaded this cow culling paper that we had done from our research center and produced the documents that it has read it over, see what it was, see if it was correct.

00;02;54;19 - 00;03;27;17
Speaker 1
And then I produced an audio overview. So it will give you like it was a conversation of two people talking about this research project, which is fascinating. So it produced a 20 minute episode on cow culling and the research paper that we went through, and I thought it was so fascinating that we are at that point that we can develop audio or podcast, whatever you might be from totally AI and people might not know about that.

00;03;27;17 - 00;03;51;26
Speaker 1
So is that scary? It sure certainly can be. I look through it. I listen to this podcast, these two people talking about this research to make sure it was correct. So you still have to have human intelligence to listen to this podcast to be able to go through and see if it was correct, I would say 98% of it was correct.

00;03;52;02 - 00;04;16;09
Speaker 1
Maybe towards the end it had misspelled a few words. I know one of the big ones was Normandy, the Normandy breed. It was maybe saying it a little bit differently than what we might think, but overall, this one that I did on cow calling was pretty darn accurate. Now I've done some other ones just I've just played around with this just to see what it would do, what it could possibly do.

00;04;16;11 - 00;04;50;13
Speaker 1
Some of the other ones I put in and it really wasn't pronouncing words correctly at all. I took a research study that we did on health care of Holstein cows from one of my previous graduate students, and put that in and see if it would generate a podcast. Well, it did it generated everything generated about ten minutes worth of audio, people bantering back and forth, whether, health of Holsteins was good or bad, but it wasn't even pronouncing the word dairy correctly.

00;04;50;15 - 00;05;33;23
Speaker 1
They would say like diary. And so it's not perfect. AI is not perfect. We still, in my mind, have to have human intelligence go through all of this stuff that is happening in the dairy world, the beef world, the crop world. Human intelligence is still needed. Can I assist us in our everyday farming? Yes it can. And so we're going to talk a little bit about AI today, but I at least wanted to, get that out there that, the cow culling one was two random people, just AI generated voices talking about my research.

00;05;33;26 - 00;06;03;16
Speaker 1
Fascinating as it is. Let's dive a little bit more into AI and and really what what that is. So obviously when we talk about, AI, we're talking about artificial intelligence. And it's in simple terms, AI is when a computer can do tasks that we usually associate with human intelligence, like spotting a sick cow or predicting milk yield.

00;06;03;19 - 00;06;26;04
Speaker 1
I think most of us think about robots in the dairy world from an AI perspective, and how it can map the teats of the utter to be able to put the unit on that cow, or a lot of these activity monitoring systems for trying to spot a sick cow. That's what a lot of us probably think about. And we'll dive a little bit into those.

00;06;26;07 - 00;06;50;27
Speaker 1
But most of the time when we actually say AI in agriculture, we're actually talking about machine learning. Machine learning is basically a branch of AI that learns from the data. It's basically kind of like how you might learn to spot mastitis faster after seeing a few cases. So really, that's what it's about. And we go back to my podcast episode.

00;06;50;29 - 00;07;17;03
Speaker 1
It's learning from the data. So we've provided it data I've provided it a research paper on cow culling. And it spit out an audio format for that research. So where is AI showing up on farm? So where where do we actually see it happening on farms. And we probably don't even realize it. Have you really thought about AI on farms and where it might be used?

00;07;17;10 - 00;07;40;19
Speaker 1
We see a lot of futuristic things going on in agriculture. Some of us obviously. Well, there can be a lot of money, you know, how do we pay for all this, computers and machine learning and algorithms and everything like that? Obviously there's futuristic things, but some of us, a lot of us actually probably have AI on our farms and we don't even realize it.

00;07;40;22 - 00;08;16;04
Speaker 1
So one is animal ID and monitoring. Obviously, I can analyze videos to identify cows. We can measure body condition score. There's a lot of robotics that are measuring body condition scores or even taking a temperature of a cow could be done by AI. And we had talked a little bit about AI in a past episode, episode 273, about looking at lameness and how camera technology and dairy farms can identify lameness on cows.

00;08;16;09 - 00;08;45;07
Speaker 1
That is machine learning. All AI being able to find lameness detection from a camera. So it's happening a lot on farms. We can also on today's farms look at health detection. So a lot of people might have activity rumination, feeding behaviors a lot of sensors on their cows actually our our cows that our research center have I think four different sensors right now I'm thinking about another one.

00;08;45;10 - 00;09;25;09
Speaker 1
We have ear tags, we have collars, we have internal rumen boluses. We have pedometers on their legs, but we also have in the milking parlor, we have AP lab where we can measure inline fat, protein conductivity, lactose, all of that. That is all I generated. Basically, if you think about it from a health detection perspective, these AI models can pick up on those cows that are not feeling well, and they can send alerts for diseases like lameness or mastitis, sometimes before we even catch it ourselves.

00;09;25;16 - 00;09;46;12
Speaker 1
Think about it from a calf perspective. Automated calf feeders we can. Maybe it's sending us an alert that this cow isn't drinking or this calf isn't drinking. The drinking speed is down so we can check out this animal. Obviously there's repro. You can see a lot of these systems that use behavioral data from a reproductive standpoint. They have steps.

00;09;46;17 - 00;10;08;03
Speaker 1
Step banding how how the cows are mounting. Be able to detect these heats and maybe predict conception. You know, most of these systems now are trying to help predict when ovulation might occur. So you can breed the cow at a specific time to increase your chances of getting conception. I see it in some of the systems that we have.

00;10;08;05 - 00;10;27;15
Speaker 1
It's like, well, the optimum breeding window is maybe not at 8:00 in the morning, it's maybe at noon or 1:00, but how do you deal with that on a farm? It comes back to management. Can you be breeding cows throughout the day? And the answer probably is no. You just don't have time to pick and choose when you can breed cows.

00;10;27;20 - 00;10;51;10
Speaker 1
And then there's a lot of different things about milk yield forecasting. Some of these machine learning models can predict the cows lactation based on past performance. So we can also think about feed intake. And this one's, quite wild. Obviously some of these systems use cameras to literally watch feed disappear from in front of the cow and match it to her.

00;10;51;10 - 00;11;23;06
Speaker 1
ID so they can really estimate individual feed intake from a group pan of cows. So trying to estimate feed intake individually for individual cows. So there's lots of benefits that we can see to some of these AI models or machine learning models with sensors or computer vision or any other aspect. And that is it can improve productivity, profitability and animal care.

00;11;23;08 - 00;11;57;06
Speaker 1
We can maybe predict what the milk output would be or disease. I think most of these are important for disease detection in in livestock, and I think those machine learning ones can have that for ease of detection. And you still have to have the human there to be able to do it. However, you know, I never really thought about this is I when I got our first activity system, we've had an activity system on our farm since 2013, so almost 12 years now.

00;11;57;09 - 00;12;20;04
Speaker 1
We've had an activity monitoring system on our cows, and I never really thought about it as AI. But now definitely, if you look back, it is predicting disease at predicted heat. It's using machine learning and AI to be able to do that. But there's some challenges that go along with all of this. I in in the livestock world.

00;12;20;06 - 00;12;44;02
Speaker 1
And one is integration of data. There's lots of issues when it comes to data system compatibility. If you came into our dairy farm, you'll see we have five computers on our desk. None of them talk to each other. All of these systems, whether it's our herd management software, our activity monitoring systems, our milking parlor stuff, they're all on their own separate computer, and they don't talk well to each other.

00;12;44;04 - 00;13;03;15
Speaker 1
So that's a problem. So you got to look at like five different computers to figure out what's going on on your dairy. And there are some people trying to develop things to be able to everybody talk together. It's maybe not quite there yet, but it's getting there. One day it will be there. It's advancing very quickly, but we're not quite there yet.

00;13;03;15 - 00;13;28;23
Speaker 1
Obviously there's risks related to data privacy, security and ownership. You know. Is this data private? Is it you know, somebody's going to hack into it and spread it all over. And then there could be some issues. Obviously data security, you know, there's lots of security lapses happening all over all the time. So how safe is this data from from your farm and who really owns the data?

00;13;28;29 - 00;13;56;04
Speaker 1
I think I've debated this with a lot of people who owns it. I will say that the farmer owns the data and it's not the technology company. You know, it's generated from your cows, your animals, your livestock. You should own that data into the future. They're just using it to develop their models. Another big challenge is ongoing maintenance of these tools like sensors, cameras, just equipment.

00;13;56;06 - 00;14;26;27
Speaker 1
The maintenance sometimes can be overwhelming. And we'll maybe talk about that next week about our experience with our automatic cat feeder and the ongoing maintenance that it takes to keep these things going. And obviously, there's ethical concerns around animal welfare and responsible tech usage that's happening in in the livestock world. So what's sort of holding us back? Why why isn't everyone using this yet?

00;14;27;00 - 00;14;48;18
Speaker 1
And that's a good question. Some of these are what I talked about before. We call them data silos. You know, like I said, five computers in our dairy office. They don't connect with each other. And getting all of that into one system is difficult. False alarms. So I hate them. Other farmers hate them. We all hate false positives.

00;14;48;18 - 00;15;12;06
Speaker 1
It's like this cow is sick at flags. A cow says she's sick, and you go out there and she's chewing her cud and looks just fine. No problems. So sometimes these AI models throw too many false positives. I think at one point when we first started with our activity monitoring system, we had lots of flies at our dairy and the company was like, well, you're you're breaking the system.

00;15;12;06 - 00;15;31;23
Speaker 1
These cows, we've never seen this activity in cows before or because it was fly behavior that was messing up the algorithm. So one thing to think about all of the false positives and they're getting better. They are getting better. Sometimes these models are hard to understand. Some of even myself, I don't understand a lot of this machine learning stuff.

00;15;31;23 - 00;16;05;23
Speaker 1
It's way out of my realm. I really admit sometimes trying to. I've done a little bit of machine learning, but I still don't understand it and it can be difficult and there's a lot of farm to farm differences. The model for farm A might not work for farm B, just like my fly issue. Well, they weren't seeing this on other farms, but our farm had a lot of a lot of flies and we were a grazing farm and they were like well, but grazing cows in Ireland, we don't see this well, Minnesota here is not Ireland either.

00;16;05;25 - 00;16;37;03
Speaker 1
We have different management, we have different environmental conditions, different weather. So the model on our farm or the model that they had trained for, every farm was not working on our farm. So we had to develop a specific model for our farm. So what about looking ahead. Where where might we going? Obviously, if you think about it from an AI perspective, it's actually moving into everyday farm life quite a bit, albeit very slowly.

00;16;37;06 - 00;17;04;24
Speaker 1
But can you train employees with virtual reality? I think that's happening. We're seeing that already now when there's a lot of maybe some controversy. I think this is very interesting stuff. Is using AI video monitoring for protocol compliance? I think it's it's basically a second set of eyes on performance of employees. And I don't think I think it's useful for training.

00;17;04;27 - 00;17;34;08
Speaker 1
And we don't necessarily want to punish people for, doing some of these things. I think we can see what is happening and actually improve our employees. One of the things that I think about, it's a company from California called Cattle Care, and they use video monitoring in milking parlors to help evaluate the performance of each employee and maybe prevent some problems that are happening.

00;17;34;10 - 00;17;59;02
Speaker 1
It uses a camera. So you have a camera in the milking parlor and it can monitor employees that are in the milking parlor. And it probably does more than just monitoring. So it can monitor predict stripping, wiping and post dip. It can look at harmful activities or unusual activities that are happening in the milking parlor. It produces a lot of reports.

00;17;59;04 - 00;18;30;06
Speaker 1
It can tell you about efficiency of cows. So it can tell you when they dip an animal, when they wipe it, when they maybe brush it, when they strip it. It can tell you all of the times, and it can really tell you how well the milking parlor is working. It can tell you the order of attachment. You know, if if an employee starts wiping, from unit one to unit ten and then puts unit ten on first and goes backwards, it can tell you all of that stuff.

00;18;30;13 - 00;19;04;11
Speaker 1
So really it can give you some ideas on efficiency in the milking parlor. And I want to say it's not about trying to punish your employees, it's about trying to improve them. And I've really thought about this myself and trying to improve the employee efficiency on our dairy farm. And I'm thinking about this model as well. There's other models that can help you, use AI powered information that will help keep your farm efficient and doing other things.

00;19;04;14 - 00;19;29;19
Speaker 1
One of them that I saw at World Dairy Expo was called catalytic. It's, a Canadian company. Basically looking at scheduling of it gives you some reminders, kind of a on your tablet or on your phone talks about what's happening on the dairy today, what needs to happen, what are those, you know, assigned jobs to different people. It may pop up.

00;19;29;19 - 00;19;54;13
Speaker 1
All we need to look at this cow. So it's trying to integrate all of our data and our standard operating procedures together, and being able to help monitor our farm and make it more efficient. I think a lot of this is trying to improve efficiency and ultimately improve profitability on farms. So I think we're going to see more of this as we go ahead.

00;19;54;15 - 00;20;19;27
Speaker 1
Another one that we're actually using ourselves is a Harvest Lab from John Deere. So we have been doing some testing on our but John Deere chopper where it actually measures MDF, ash, ADF dry matter protein on the fly in the field maps the whole field. So we could actually tell where the crude protein is the best in an alfalfa field, where the knife is the best.

00;20;20;01 - 00;20;44;28
Speaker 1
So we can use AI to help determine how we might plant that field or what crops might put in that field. Or maybe we have to put a few more seeds per acre in this spot than the other spot. So that's one way that we're using it as well. And probably the last thing is looking at real time translation for maybe multilingual employees.

00;20;45;01 - 00;21;17;24
Speaker 1
We talk about AI powered speech tools. Kind of goes back to what we heard last week AI powered speech. Could that help break language barriers on farm? Can we use it to translate? Can I upload my voice and it translates it into Spanish or some other language, whereas that person could listen to it. I think the sky is the limit, definitely, when it comes to AI, and I think it's we certainly need to embrace it in my mind, obviously, we need to use human intelligence when we're talking about AI.

00;21;17;27 - 00;21;42;02
Speaker 1
Obviously there's, people can be skeptical. I think about it from my kids at school. They pretty much ban I use from kids in school. I'm not sure that I actually agree with that. I think we need to embrace it. It's going to be here. I read some reports last week, that the the Chinese are using AI and their new AI model deep seek for many different things.

00;21;42;05 - 00;22;10;14
Speaker 1
So I think we need to embrace AI. It's here and we need to think about it. But we also need to have the human intelligence standpoint for it. So where does that really leave us on? Some closing thoughts. Obviously, AI is not going to replace humans on the farm, but it certainly will change how we work, how we maybe make decisions and how we train the next generation of livestock farmers.

00;22;10;16 - 00;22;35;15
Speaker 1
In my mind, the future is really not in a laboratory. It's already showing up in some of these barns, in milking parlors, in tractors and on cows. It's everywhere. So I think we need to embrace it and but still use that human intelligence that we have. So with that, I think we'll end today. And talking about AI, I think we need to watch for it.

00;22;35;22 - 00;23;04;00
Speaker 1
It's going to happen and it's, it's going to move really fast. So with that, if you have any comments, questions or scathing rebuttals, feel free to contact me at the Moos room. That's CMO s r o m@um.edu. Or you can find us on the web at University of Minnesota Extension Livestock, or find me, on Instagram or the web at West Central Research and Outreach Center.

00;23;04;04 - 00;23;10;00
Speaker 1
With that, we'll see you next week. Bye.