AI and the Future of Agriculture

Technology continues to transform one of mankind’s oldest professions, on a much needed basis. As the global population continues to rise and the land becomes increasingly unavailable for agricultural use, farmers have needed to become more efficient and productive. So what can AI do to help? The industry is turning to artificial intelligence and machine learning to monitor soil conditions, control pests, and organize data that will help farmers yield healthier and more substantial crops.

AI and the Future of Agriculture transcript

Jason Colby (00:01):
Technology continues to transform one of mankind’s oldest professions on a much needed basis. As humanity’s population continues to grow and the land becomes more and more unavailable for agricultural use, people have needed to become increasingly more and more efficient and more productive. So what can AI do to help? The industry is turning to artificial intelligence to monitor soil conditions, control pests, organized data, and all this that will help farmers yield healthier and more substantial crops. This can be especially important for farmers that specialize in growing organics. Join us as we begin our two part series about AI and the future of agriculture right now.

Kaustubh Kapoor (00:56):
Welcome to smarter than AI, we last time we talked about AI and law enforcement and our AI can help save and abduct a child or help with roadside sobriety checks. This time in our two part episode like Jason mentioned, where tackling AI in agriculture. And agriculture is going to help farmers grow more food efficiently and effectively. The first part focuses on crop cultivation and grain handling. And how technology is transforming the traditional understanding of agriculture in 2021. I also just wanted to mention that in this first part of our podcast series, we’ve focused on very surface level introduction of AI and mentioned a lot of cool concepts and cool technologies. But don’t worry if you don’t get it right here right now, we are also going to introduce these topics in much more detail in the next coming episodes. However, if you look at AI and agriculture, the field is relatively new and somewhat untested. There are many considerations that impact the accuracy of some of the more complex algorithms that AI needs to form decisions. And there are questions around the scale ability and practical of the solutions introduced to a farm.

Jason Colby (02:12):
Right on. Cause department really excited for this episode because I actually come from a farm. We did great handling. We did a lot of oats and barley and wheat. But we also raised a lot of cattle. Specifically, charley and herford mix. So this really hits home for me this episode and I’m really excited to talk about all this stuff. But first and foremost, if you’re a farmer and you want to ask the most important question, I think it comes right down to soil conditions. We got to monitor that. How can we use machine learning to optimize the nutrients in the dirt that we plant common crops in, like canola wheat and barley, or and how do we hinder soil borne diseases?

Kaustubh Kapoor (02:54):
That’s an interesting question to get us started because I’ll be honest, I am not coming from an agriculture background, so I have to research a lot of this before we started talking today. And I came up with a few interesting things. And I’m going to I’m going to pull up my research before I start talking. So I’m not mistaken or anything. So it’s obviously a threat to food security with soil degradation. And in the U.S., we have actually a stat that puts the soil degradation at costs into numbers. So the USDA estimated the annual cost of soil erosion to be approximately $44 billion. And if AI can help with maybe 10% of that, that’s a lot of money that can be saved. So a Berlin based agricultural tech startup called Pete has developed a deep learning application called plantings that reportedly identifies potential defects and nutrient deficiencies in soil. Analysis is conducted by software algorithms, which correlate particular foliage patterns with certain soil defects. Plant pests and diseases. And this is an image recognition app that identifies the defects through images captured from a person’s cell phone. Imagine how simple and easy this makes it for farmers and big base corporation as well just to sense of people into the fields and just click some photos and up to 95% accurate results about how bad your soil is. Again, I’m not from agriculture background, but I just find that super cool that people get to go up with their smartphone take a couple photos and get some accurate results on their soil condition.

Jason Colby (04:37):
That’s crazy. That’s actually groundbreaking because if you’re a farmer, every every minute of the day counts, right? So if you can monitor your soil conditions just by taking a photo, that’s insane.

Kaustubh Kapoor (04:50):
Yeah, again, you’re definitely more of a specialist in this field. So I’m going to rely on you to come and jump in with the tidbits of how these technologies might be happening. I just know about these technologies because of my research. The second one that I found super cool was called trace genomics. I’m a taste genomics. I think I’m saying it right. It’s a machine learning for diagnosing soil effects. That’s the technology what it’s based upon. And the emphasis here is on preventing defective crops and optimizing the potential for healthy crop production. It’s somewhat similar to our Berlin based startup Pete that has the application called plantings in that you submit a sample of your soil and then through these machine learning algorithms. It gives you an in depth summary of your soil contents. And these are provided services in packages. And it can be used for pathogen screening based on focusing on certain bacteria and fungi, as well as comprehensive microbial evaluation. And there’s probably on this is probably many companies that are doing something similar. These two are definitely industry approved and industry specific, so I thought I would just mention these two on how AI is hoping to solve the soil erosion and soil degradation problem.

Jason Colby (06:09):
Yeah, I mean, it’s all about how much carbon really is in that soil, right? So if you can manage that, it’s fantastic. All right, let’s move on to the next question, though, because I’m excited about this one. It’s not a term that I’m used to personally, but I’ve heard about it. It’s precision agriculture. So what that is is basically it’s machine learning. They’re using new technologies to increase crop yields and profitability while lowering the levels of traditional inputs like water and fertilizer and herbicides and pesticides. Everything you need to grow a crop essentially, if you’re not growing at organically, of course, if you’re growing it organically, then you probably forget about most of that stuff with the exception of water, obviously, but you still have to manage it all. But so what are some of the ways that AI is used in precision agriculture?

Kaustubh Kapoor (07:06):
Yeah, so again, I came caught this term when we were talking about what to talk about in this episode and I thought it was a fun term that people are using just to make agriculture more technology focused. And it’s a crazy, it’s a crazy thing because we live in a world where we can get our pizzas delivered by robotic autonomous car, but the ingredients of the pea just don’t need to come from the ground. Jokes aside, agriculture worldwide is like a $5 trillion industry. And it’s quite progressive in that from what I’ve seen at least in different fields. It’s a field where technology is being accepted quite quicker than say maybe even just traditional banking, where if you want to implement, I think certain solution you might run into more flags and hurdles. So I thought that was super cool. But what precision farming is trying to do, it’s basically trying to control pests, monitor the soil, and going conditions, organize data for farmers, and help with workload. So basically you try and optimize the entire supply chain for food, right? And what we’re trying to do is use AI sensors that can detect and target weeds and then decide which herbicides to apply with the right buffer zone. And this is just one of the examples that I’m going to use today to just talk about what precision farming and precision agriculture can actually provide to the agricultural industry as a whole. They also farmers are also now starting to use seasonal forecasting models to improve agricultural accuracy and increase productivity. These models are used to predict upcoming weather patterns a month ahead to assist with decisions that would help them decide which crops to grow. Should we start should we start planning them earlier? Should we start removing them earlier that sort of thing? And that all comes with seasonal forecasting of weather conditions and soil conditions as well. And what’s interesting is that some of the some of the solutions that I’ll mention are actually large scale solutions for, for example, I’ll probably talk about drones and autonomous tractors and these solutions are therefore the bigger sort of growing companies that actually operate farms as businesses and large scale businesses, whereas 70% of all the farming produce actually comes from small scale farmers. And when it comes to seasonality predictions, it’s an affordable AI use for a small scale farmer, which you can see if 70% of all of our farming products is coming from small scale. This could be really helpful and useful compared to some of those more fancy quote unquote fancy technologies like drones for pesticides and autonomous tractors.

Kaustubh Kapoor (10:15):
So let’s dive a little bit deeper into the specifics of how AI is used here. AI sensors to detect diseases uses optical techniques such as RGB imaging, hyperspectral sensors and tomography. Now the exact science behind this from an AI perspective even comes from the broad broad term of computer vision. And like I’ve mentioned before, we’re not delving deep into how computer vision would do certain things. But let’s look at the AI sensors using RGB images. And the RGB color model is a simple additive color model, in which red green and blue light is combined together to make different colors. Now in an agricultural setting, these sensors take pictures of your crops and through a web of IoT sensors, which is Internet of Things, sends this these pictures back to, let’s say, a processing center where these pictures will be processed to the RGB model to basically check whether the color of the plant denotes any diseases or not. I thought that was really cool because then combining this with the drone technology that I was talking about mentioned more. You can pinpoint which plants are in need for more pesticides and which plants aren’t and which plants actually went in the time of harvest needs to be chucked out and which plants don’t. I thought that was really interesting.

Jason Colby (11:47):
Yeah, that’s very interesting.

Kaustubh Kapoor (11:49):
And the second thing is our seasonality predictions. Now this is super interesting because the weather report that we get on our phones is almost never totally accurate. It’s kind of in the ballpark, and that’s sort of what farmers are trying to do with this as well. They’re just trying to understand which how intense our season is going to be. So is this summer really going to be super intense? Or are we going to expect more than normal rain or is it going to be really humid? And that’s necessarily what farmers are trying to understand to the seasonality prediction so that when they pick crops and what to grow and how long to grow stuff, they have that information beforehand so they are not excessively watering plants or not using enough pesticides and herbicides to actually make sure the pants go grow at the rate they want them to grow. So these are the two examples that I want to talk about in how precision farming is coming into play. And how going with the flow and just farming on a daily basis like a day by day check health check is sort of getting pushed aside.

Jason Colby (13:08):
Right, yeah, it’s like more of an up to date more scientific farmers almanac from back in the day. I like it. It’s kind of like they’re bringing it back but they’re bringing it back with a lot more science involved.

Kaustubh Kapoor (13:21):
Exactly. And like I mentioned before, the uptake on AI and specifically technology has been from what I was reading at least has been super easy compared to some other industries that I’ve worked across.

Jason Colby (13:37):
Yeah. And I think the cool part is from my understanding. Anyways, and we’re just kind of leading it to the next question here. But last episode we talked a lot about image recognition and surveillance. We just spoke a little bit about the RGB images that are being taken now. How are those being taken? Could we use drones or any other devices to do that work and transmit those images to a database that helps determine the yield or spot problems?

Kaustubh Kapoor (14:10):
Yeah, finding you say that. Because in my research, I came across a couple of different companies that are doing this. Just as an example, we at the firm as well have done a very cool project with windmills where we launched drones and near a windmill farm and how this let me backtrack. So how windmills are taking care of usually is that a person actually climbs right on top of windmill and then we’ll try and assess the defect of specific fangs by looking at each bank separately. Or each blade separately. And you can imagine that an extraneous process.

Jason Colby (14:58):
That’s not something you could pay me for. That’s for sure. Not a chance.

Kaustubh Kapoor (15:02):
Exactly. So what we did was we developed a computer vision algorithm that would, with the help of drones, take photos of different windmills and different blades. And using, again, it’s an RGB model that we developed. It would try and depict what part of this blade is chipped or what part of this blade is suffering from decoloration, which is a sign of an early sign of blade defect. And so that people will go ahead and fix it. And that saved the frame that we were working with a lot of money. And a lot of manpower at the same time. And it helps people because there’s less accidents driven around this industry in itself is quite accident prone. So yeah, that’s one of the things that we were focused on and just like that. In my research, I came across this company called skyscraper technologies, which is working with drones and computer visions for crop analysis. One of the most interesting thing was that the market for drones in agriculture is actually projected to be four 80 million by 2027.

Jason Colby (16:11):
It’s huge.

Kaustubh Kapoor (16:12):
And that’s huge. So in addition to ground data, farmers are taking to the sky to monitor the farm, right? Computer vision and deep learning technologies are helping farmers depict which part of the crops or which part of their entire harvest is gone bad or which soil doesn’t look like it should at that point this season or what part of the crops actually need more water because they look dehydrated. And before drones you would have to physically walk through the entire harvest to access a 100%. You done it? No way, okay. That’s awesome. Yeah, so that what this drone does, it uses a pre identified route. If you watch the Roomba episode, you might kind of relate because the Roomba also has a pre identified route after it takes photos off your house and identifies a route that it’s going to clean its way through. And drones kind of do the same thing. They have appeared in 5 route where they’ll go and take photos and after the photos have been taken. You can transfer this data onto a USB drive or upload it to the cloud. And then skyscraper will do its thing and come up with come up with analysis on similar things, you know, which part of your which part of your crop needs more pesticides, which part of your crop needs excessive watering or which part is definitely gone bad because of. An infestation of mold or something. And it needs to be discarded away with. So these are the things that I thought were super cool because these are unmanned drone that drones that will just go ahead and click all the photos and you have all the data and without needing to spend hours and hours walking across the field. You have this amazing analysis at your feet.

Jason Colby (18:06):
You know what? That leads right into my next question. And you mentioned it before the autonomous tractors. And I’m excited for these. But I got to know, is this more hype versus reality? I mean, we talked about this in our very first episode. We talked about self-driving cars. And right now, we think those are more hype than reality. More or less. But given today, there’s a shortage of agricultural workers, farms don’t make as much money as they used to. Does that mean that making AI and machine learning based smart tractors spraying drones? And all these other types of robotics, more of a viable option for many remote agriculture operations that struggle to find workers.

Kaustubh Kapoor (18:55):
This is sort of like a quantitative and a quality question at the same time, because in my head, when I talk to a consumer of AI, there’s always like, it could always go both ways. One, they could be super into the idea and super amazed by the cool stuff that we can do. Our second, they could just be closed off and try and test us. And that happens quite a bit in this industry where clients will try and test you on their reality of this, right? And with AI in farming, I think the biggest hurdle that there is is that even though, like I mentioned in the drone conversation, it’s an autonomous drone that doesn’t need you to fly it. There’s still some technological aspects that are farmer needs to learn to get it actually functioning at the farm, right? And fortunately what I’m seeing in the field is that agricultural implementations of AI are maturing at a rate quicker and they’re being adopted and can be relied upon. So let’s look at a few examples the same similar with the drone example. Let’s look at this drone company called DJI that has created multiple different types of educational videos have onsite support so that farmers that might not be the most fluent with technology can actually use this so that it makes sense to them. And they do some cool stuff. They take pictures and then create a 3D model of the farm, which farmers can then play with and then set based on their subject matter expert expertise. They can take a suggest which parts of the farm need more water, herbicides and stuff and pesticides, and I thought that was super cool. But again, there’s a lot of support there so that first time user doesn’t feel overwhelmed.

Kaustubh Kapoor (20:40):
And the second thing I want to talk about is the John Deere series and their innovations. They have this thing called C and spray, which is similar to the technology I was mentioned. Mentioning, it’s a high resolution camera that captures 20 images per second and based on the images in artificial intelligence. The system recognizes which parts need to be, again, controlled for weed or put more water into and then a farmer is the ultimate decision maker that. And secondly, they are doing it, which is really cool. It’s called command cap. If this is to make sure that the farmers always in control and this is where our autonomous tractors come into play, it’s called it’s basically like if you were in a VR reality, like, for example, if you went to the arcade, but like in real life, you’d be sitting at in your office and you would have full control of tractors using a joystick. And you could then drive these things sitting inside without having to burn from the sun. So things eventually what I’m trying to say is that things are becoming a reality, they’re not maybe they’re not quite there yet, but I think the pickup is quicker than other industries because I think the agriculture industry as a whole is realizing that technology is needed to sustain what they did in the past because like you said, the reduced labor force, the need for more reduction of land availability, the need for more efficient crops is just growing.

Jason Colby (22:15):
A 100%, so let’s talk about, I think that pretty much wraps the time up. Again, this is a two parter, right? So let’s talk about what we’ve talked about. We know that we can monitor soil conditions more effectively over time and increase the quality of the soil that we plant in using AI based technologies, such as drones such as autonomous tractors come out. I’m so excited about which, let’s see. I don’t know if that’s going to be a thing or not, but I’m excited to find out. What else is going to happen? Farmers can also minimize the inputs that are necessary to grow a healthy crop using machine learning. And we can use drones to monitor the health and overall wellness of anything planted in that soil over the course of a growing season. And last but not least, we can also reduce the amount of work that it takes to manage it all, right?

Kaustubh Kapoor (23:10):
Exactly. You’ve summarized everything that we talked about quite quite well.

Jason Colby (23:19):
Okay, that’s a wrap for part one, then. Join us next week as we talk about the opportunities and challenges that AI can help with farmers and ranchers that raise livestock.

(23:30):
And we would love to hear from you. If you have any feedback or questions for our team or topics that you would like us to cover, please let us know. Drop us a line on Twitter, Facebook, or LinkedIn, or visit us at mnpdigital.ca.