John Deere’s AI vision for future farms

Note: Transcripts are automatically generated from episode audio, and are not fully corrected for spelling, grammar, and formatting.

JAHMY HINDMAN: All of the corn and soybeans in the U.S. are generally planted in a two- to three-week window in the spring, so every farmer is busy planting, right?

Labor’s a challenge, so we started working in the space of full autonomy.

We’ve started to show the world what we’re up to in terms of replacing, or giving the farmer the choice to replace, themselves in the cab of the machine if they choose to do so. We’ve had full autonomy on tractor products.

That sensor modality is primarily a camera array around the top of that tractor. So, 16 cameras with overlapping fields of view. We do frame-by-frame calibration so that we get depth from that sensor array as well.

The compute to do that is not a trivial task. We run embedded GPUs, Nvidia GPUs, on that. Importantly, in our application, they need to be hardened. They need to be able to survive shock, vibration, and temperature extremes.

RANA EL KALIOUBY: Right. They’re not sitting in an air-conditioned data center.

HINDMAN: They are definitely not. We take a lot of pride in our ability to harden some of these devices because we need the compute capability in order to do that perception problem on the machine.

EL KALIOUBY: That’s Jahmy Hindman, CTO of John Deere.

The almost 200-year-old agriculture super-company is fascinating as a technology enterprise. They’re developing AI-powered tools that help farmers be more efficient and productive. And the stakes are high — for humanity and the planet.

I first met Jahmy several years ago at CES, the legendary show for consumer tech. I have to admit I’d never thought deeply about the technology powering farm equipment before then, but I’ve wanted to talk to him ever since.

Our conversation covers why John Deere owns the tech stack for farming, the company’s vision for the future of agriculture, and the most important question of all — what would a fully autonomous mango farm look like?

It’s fascinating stuff, so let’s get to it.

I’m Rana el Kaliouby, and this is Pioneers of AI, a podcast taking you behind the scenes of the AI revolution.

[THEME MUSIC]

EL KALIOUBY: Hi, Jahmy. Welcome to Pioneers of AI. I’m so excited to have you on the show.

HINDMAN: Rana, it’s great to see you. Thank you for having me on the show.

EL KALIOUBY: All right, so you’re the CTO, chief technology officer, of John Deere. It’s safe to say that if most non-farming people were to name one brand in farming, it would be John Deere. But I also think most people don’t associate John Deere with being a technology company.

I do want to start with your background. You grew up around farming. Your grandfather was a farmer in Iowa, and I guess you still have family members who are.

HINDMAN: Yeah. It’s a story that’s pretty common in the upper Midwest. A lot of people can trace their roots to agriculture in some way, shape, or form. My grandfather farmed a little farm in southwest Iowa. In a story that repeats itself over and over again, it was not large enough to support the whole family.

So my dad became a college professor. He taught aerospace engineering for his career, and that farm got consolidated into other farms.

EL KALIOUBY: How has that upbringing shaped what you do today and your perspective on the whole business?

HINDMAN: I’ve been around agriculture for a long time — my whole life. Maybe not directly involved in it at some periods of my life, but always around it. And I’m a technologist at heart. I grew up with an engineering professor as a father, so I was around technology my whole life. I think those two things go together.

Agriculture is an application that begs for efficiency improvements. If you were to rewind the clock 50 years ago, roughly 30 to 40 percent of the U.S. population would’ve been involved in agriculture directly.

Today that number is like 1.5 percent. If you think about that, it’s enabled people like you and me to do the things that we do. We no longer have to worry about how to get our food. We go to the grocery store and it’s there for us. But that’s done on the backs of 1.5 percent of the whole population.

Agriculture has been a story of efficiency, and technology is the underpinning, or the foundation, for how that efficiency has happened.

Copy LinkWhy agriculture has always been a story of efficiency

EL KALIOUBY: Amazing. So you started your career at John Deere in 1996. That’s like 30 years ago, as a test engineer.

HINDMAN: Sounds terribly old, but yes.

EL KALIOUBY: As a test engineer, you’ve seen the evolution of the company.

I would love for you to walk us through the company’s very first product and then the range of products you guys have today.

HINDMAN: Yeah.

EL KALIOUBY: I love all the models in the background.

HINDMAN: Yeah, for sure. We’re a 189-year-old company, which we’re proud of. That means we’ve had to reinvent ourselves multiple times in the history of the company. We trace our roots back to John Deere, the man himself, who lived in Vermont but moved to the Midwest, to the state of Illinois, as the country was being built.

He was a blacksmith.

The plows of the day were largely wooden plows and sometimes cast-iron plows, and soil was always sticking to these things. Farmers had to stop and clean the plow off every 10 meters or so.

John Deere’s claim to fame is that he fashioned the first self-scouring steel plow, and we’re proud of that product, obviously.

One of our more pivotal moments was in the early 1900s, when this thing called the internal combustion engine started to happen. We no longer had to rely on animal power to do farming, and the tractor was born. John Deere actually didn’t develop the tractor or start the tractor. That was a company we bought, the Waterloo Gas Engine Co. They happened to have an engine that they put in the form of a tractor, and we purchased that company.

EL KALIOUBY: About the inventor’s dilemma — I think that’s really relevant still today.

HINDMAN: It is. And you can look at the models behind me — it’s the ubiquitous product form for the company today. That really started us on this path, right? This ability to take advantage of the efficiency of mechanization and have people go do other things with their creative potential that, obviously, I think the world has benefited from.

We’re in the middle of that, like many industries right now, with artificial intelligence. It obviously has huge potential, I think, in the agricultural industry.

We view it as a responsibility for our company to be able to utilize that technology for the benefit of the customer and to walk hand in hand with our customers to make sure they agree with us that it’s doing the things that are beneficial.

My experience sort of falls into the category of I’d rather be lucky than good. I’m a mechanical engineer — bachelor’s, master’s, and Ph.D. — but my graduate work was focused in an area called artificial neural networks, of all things, 15 years ago.

It was not a very compelling area of research at the time. It was challenged by lack of compute, not-great data sets, and, in our space on the edge, the inability to access equipment through communication paths. All sorts of things were hurdles and roadblocks that are largely removed today.

Copy LinkUnpacking John Deere’s tech stack

EL KALIOUBY: Very cool. So you oversee the entire tech stack for John Deere — the hardware, as you said, the software, all of it. But it would be helpful to unpack what a tech stack means in the agriculture space.

What does it look like?

HINDMAN: Yeah.

I’m a mechanical engineer — bachelor’s, master’s, and Ph.D. — but my graduate work was focused in an area called artificial neural networks, of all things, 15 years ago.

So I’ve been in this place where I understand the mechanical side of our business, which is still really important. We still have to provide these robust, durable, highly productive pieces of mechanical equipment. But we also need to weave through those pieces of equipment the modern technology — artificial intelligence and software — that can improve the productivity of that equipment.

It’s really a handful of things that are critical and that we build on. The first is the ability to locate the piece of equipment on the surface of the planet — GPS.

EL KALIOUBY: Why is that so important?

HINDMAN: Our own GNSS. Great question. In agriculture, it’s important for a couple of reasons. One, plants live their best life when they get the opportunity to equally compete with one another. That makes it interesting for us to create the same row spacing, right?

So precisely putting that seed in the ground at exactly 30-inch rows, time after time after time, is a core component of agricultural efficiency.

In addition to that, we don’t like to overlap things. If you’re putting seed in the ground, you don’t want to have seed on top of seed, right? It gives you the ability to know where you’re at. If you imagine you’re in a 1,000-acre field, this ability to know where your 16 rows of planters are in 1,000 acres is a very challenging thing to do.

Before GNSS, you were often guessing about where the tractor had already been or where the piece of equipment had already been. So it gives you the ability to keep yourself from duplicating the work in the field or skipping some areas of the field and not doing the work at all. Those are some of the ways that it’s important.

We think about it as this idea of plant-level management. If you look at the corn crop in the United States, there are 4 trillion corn seeds planted in the United States every year, give or take. Our mission is to be able to provide, in mechanized form, the master gardener experience for every one of those seeds. We want every one of them to be treated exactly where it needs to be treated, how it needs to be treated, and when it needs to be treated to live its best life.

EL KALIOUBY: This just occurred to me. I don’t know if the analogy makes sense, but as humans, I’m really into my wearables, right? I track my sleep, I track my steps, I track my activity. If there were an easy way to track my nutrition and hormones — all of it, right? It just occurred to me that in this mission to help plants live their best lives, do you also have a picture of a plant’s health and wellness?

HINDMAN: I love the analogy. We’re an organism, the plants are organisms — it fits generally. We don’t know as much about the plants as you know about yourself, because you’re more instrumented than the plants are. But we’re on a path to instrument plants in a similar way.

One of the interesting technologies that we’re exploring at the moment is to give a plant the ability to communicate what it needs, what its stresses are in its life. There’s a company called Interplant that we partnered with that is genetically modifying plants to allow them to fluoresce at a certain wavelength based on what stress they’re seeing in their life.

So if they’re seeing, in the case of soybeans, a stress due to a fungus attacking the plant, they fluoresce in one wavelength. If it were nitrogen deficiency, they would fluoresce in a different wavelength. If it was water deficiency, they would fluoresce in a different wavelength. Pretty soon, you can conjure up this mental image of the ability to sort of listen to the plant and understand exactly what it needs and then treat it.

So, yeah, absolutely. I think the analogy holds true.

EL KALIOUBY: That is so cool. I spent my entire career building emotion recognition and emotion-sensing technology for humans, but that would be the equivalent of nonverbal communication for plants. That’s so cool.

HINDMAN: For sure.

EL KALIOUBY: Coming up: a walk-through of how John Deere brings together sensors, data, and AI in one of its high-performance machines — and what the company does with all that data to help farmers get better results.

Copy LinkWhat sensors and AI actually do on a modern tractor

Now, as you know, I also spent a lot of time in the automotive industry bringing together this idea of sensors, data, and AI.

So I want to unpack what that trifecta looks like for John Deere vehicles. Maybe we start with a specific product, the John Deere 9RX 830. I don’t know if it’s one of the ones behind you. Is it?

HINDMAN: It is very similar to that tractor right there.

EL KALIOUBY: Okay, that’s great. This model can retail for up to like $2 million.

So what is this tractor built to do, and what kind of sensors sit on it?

HINDMAN: Yeah. A long time ago, we started putting high-performance compute and sensing elements on these products that enabled them to start collecting information that was useful to the farmer.

So in the case of that 9R, those units would be responsible for collecting and communicating information about how the planting process went. Like, how many seeds did you plant? Where did you plant them? What depth did you plant them at?

Those sorts of things, so that a farmer can take that agronomic information and understand, OK, what did that produce in terms of germination? How many seeds germinated and are going to create a new plant? I can then use that information either in the next season to determine how many seeds to plant and how deep to plant them, or I can use it in season to understand where I might need to replant my field.

I didn’t have good germination. I need to replant those seeds. All of that data goes from the planter behind that tractor, through the tractor, up to a cloud instance, and into a mobile application that the farmer would be able to access.

EL KALIOUBY: Yeah. Let me stay on the sensors a little bit too. Do you use computer vision? Do you use radar? Lidar? What kind of sensing?

HINDMAN: On that machine and on another machine, in two different applications, we do use computer vision. In many cases, there’s not enough labor to do all the work on the farm, or at least labor is a challenge on the farm.

Primarily because in most of the production systems we’re involved in, you need all of the labor at a very specific point of the year. We plant all of the corn and soybeans, for example, in the U.S. in a two- to three-week window in the spring, so every farmer is busy planting.

You need these maximum amounts of labor in a very short period of time. So labor is a challenge. We’ve already got hands-free guidance on these pieces of equipment, but we haven’t replicated the greatest sensor of all, which is the human in the cab of the machine. So we started working in the space of full autonomy.

We’ve been working in that space probably for 30 years, but more recently, publicly. We’ve started to show the world what we’re up to in terms of replacing — or giving the farmer the choice to replace — themselves in the cab of the machine if they choose to do so. We’ve had full autonomy on those tractor products in limited applications with customers over the last four years.

That sensor modality is primarily a camera array around the top of the operator station of that tractor. So, 16 cameras with overlapping fields of view. We do frame-by-frame calibration so that we get depth from that sensor array as well.

The compute to do that is not a trivial task. We run embedded GPUs, Nvidia GPUs, on that product. Importantly, in our application, they need to be hardened. They need to be able to survive shock, vibration, and temperatures that traditionally these compute devices would not survive.

EL KALIOUBY: Right. They’re not sitting in an air-conditioned data center.

HINDMAN: They are definitely not. We take a lot of pride in our ability to harden some of these devices that traditionally would not find their way into our applications because we need them. We need the compute capability in order to do that perception problem on the machine.

In the machine right below it — this is called a self-propelled sprayer — we use computer vision in a very different way. That machine has a boom that will spread out to 120 feet. It will travel at about 15 miles an hour. Traditionally, it would apply herbicide, pesticide, or fungicide across every square meter of the field, even if that part of the field didn’t need the application, because there was no way to sense whether the crops needed it in that particular area.

So we put 36 cameras across that boom. We put nine embedded GPUs on that product. We do what we call See & Spray. We sense the ground at 15 miles an hour and look for pixels that contain weeds and pixels that don’t contain weeds. The idea is, you don’t need to spray herbicide on ground that doesn’t have weeds, so you only spray the weeds.

It’s good for the farmer. It’s good for the planet. Nobody wants to spray more herbicide than they need to. It’s good business for us. So it’s a bit of a triple win, and there are great applications of AI in agriculture.

Copy LinkTurning field data into decisions farmers can use

EL KALIOUBY: Yeah, that’s awesome. One way to think about these tractors is that they’re a massive data-collection machine, right? To your point, you’re processing all this data and uploading it to a cloud somewhere. What do you then do with the data, and how does that show up on the operator or the farmer’s side?

HINDMAN: The data goes to a cloud instance, but in many of these locations — remember, they’re rural locations — you may not have terrestrial cell service, even here in the U.S. So we partner with Starlink to put low-Earth-orbit satellite terminals on product in areas where we don’t have great terrestrial cellular connectivity. But one way, shape, or form, data gets pushed into a cloud instance.

We do a lot of different things to it. We process the data into usable pieces for the farmer so they can look at things like a yield coverage map, for example. Because we know where every seed was planted and we know where the machine is when it’s being harvested, we can tell you, in a discretization about the size of a pizza box — maybe one square meter of resolution — how much corn or soybeans, or whatever the crop is, came from that particular area.

I refer to it as the farmer’s report card. It’s: How good did we do? Did we produce 150 bushels of corn on this particular piece of ground, or did we produce 300 bushels of corn on this piece of ground?

At the end of the day, that is what the farmer wants to know because it provides the information necessary for them to make better farming decisions, better agronomic decisions — whether that’s genetic variety of the crop, seed density, nutrient schedule, all of these things for the next year.

They get that exposed to them in two different ways. We expose it in what we call John Deere Operations Center. It’s our digital offering to customers. They can see that either in a desktop version, which is the way most of them will do their back-office data understanding and data analysis post-harvest, or we expose it in a mobile environment, Android or iOS, which is where the feature set that is more near-term resides so they can understand logistics on the farm.

EL KALIOUBY: How has generative AI changed any of your products and how farmers experience the products?

HINDMAN: The predictive side of things — we’ve done that for a long time. We do things like try to determine, based on anonymized aggregated data sets, when the best time to plant is.

That’s the conventional way that data sets can be used. Generative AI — and I would argue transformer networks in general — is interesting to us for a whole variety of reasons. I think the first is that agricultural data is often poorly structured. It’s messy, it’s complicated.

But generative models have given us the ability to sort of reject the noise in the data and focus on the signal, and to be able to do that at faster clock speeds than we’ve traditionally been able to do it. So you can unpack more insights out of the data. We’re also interested in them for edge use cases. I talked to you about the autonomous use case as an example.

This notion that you can do an end-to-end understanding of what a farmer would do — how they manipulate the controls when their eyes see these things, when their ears hear these things. In that way, you can sort of completely emulate what the farmer is doing within the operation, as opposed to just assuming that they’re going to make these decisions when the sensory inputs are X.

EL KALIOUBY: OK, so that’s really interesting. What you’re saying, I think, is that if you take one of these tractors, the default model is it’s collecting all this data, it’s going to the cloud, and then it’s going to go back to the farmer’s mobile phone. The farmer’s going to say, you know what, click — action A taken. Instead, you can actually have all this run on the edge, basically in the tractor, where the tractor can say, OK, I have all this data. It’s very likely that I need to make decision X, or decision A.

HINDMAN: Right.

EL KALIOUBY: So cool. It’s very interesting. I have more on that.

We’ll come back to it.

HINDMAN: It is only an idea that you can contemplate tractably doing today because the compute is expensive — don’t get me wrong — but it’s not the limiting factor anymore. Traditionally, it’s always been the limiting factor for us on the edge.

The embedded GPU compute trails, roughly, six years behind data center compute in terms of the operations you can run on a given unit of power. So you can kind of think forward. Think of what’s happening in data center compute today, and in five or six years, you’re going to have that capability in your hands at the field edge.

What are you going to do with it then? It’s a tantalizing intellectual thought experiment to go through.

EL KALIOUBY: Yeah. I mean, the whole semiconductor space — there are a lot of new players who are essentially just focusing on the inference problem.

HINDMAN: And you saw Google — they’re separating it, right? Inference is going to be a separate compute device. I think that’s interesting in its own way. We still need some inference on the edge, no doubt, but separating that out of the equation gives us the ability to sort of play with the compute topologies on the edge in a way that we traditionally have not been able to.

EL KALIOUBY: We have more to get into — how farmers are using AI on the ground today and what role robots might play in the field. Stay tuned.

Copy LinkHow farmers use generative AI

Do you envision farmers using natural language? I’m envisioning a chat interface, basically, to prompt for data and insights. Is that right?

HINDMAN: Yeah. I was meeting with growers in Pasco, Washington, probably a year ago, and there were 16 or 17 folks. I started out the presentation and the conversation with, “How many of you have heard of ChatGPT?” And all their hands went up. I said, “How many of you use it on the farm?” And almost all their hands went up.

Then the next question was, “How many of you use it regularly, like every day?” And about half of them did. That’s an interesting observation. They’re using it as a thought partner to assimilate the data that’s on the farm and juxtapose that against decisions they would make, and to spar with it intellectually a little bit about what decisions they should be making on the farm.

So I do think there is an appetite for it, for sure. And there is an opportunity for that to start contributing to making farming practices more efficient and more effective over time.

EL KALIOUBY: Yeah, it’s so cool because, again, you’re collecting so much data. I imagine there are set dashboards where you can see different views, but it would be so cool to be able to interrogate that data and just ask natural questions.

HINDMAN: Exactly right. That’s the reason they’re drawn to it. The interface, the user experience, is very natural.

Copy LinkExplaining the right-to-repair issue

EL KALIOUBY: This is a very strong business model, to own the entire tech stack. But from a farmer’s perspective, it feels a little bit like a monopoly, right? I want to talk about the right to repair for a moment. John Deere just settled a lawsuit around this. I would love to first have you explain what the right-to-repair issue is, and then your perspective on this and what happens next.

HINDMAN: Yeah, sure. Maybe I’ll rewind and start 25 or 30 years ago when we started to put microcontrollers on equipment. This idea that a farmer was going to want to reprogram those microcontrollers was never a thought. It just wasn’t 25 or 30 years ago.

I would say the current state is an evolution from that point. In fact, we wouldn’t have thought about reprogramming those controllers back then either, unless something had failed.

EL KALIOUBY: If there are no over-the-air updates, right?

HINDMAN: Right. There were no over-the-air updates. We didn’t get here intentionally. We got here as a consequence of technology changing over time and customer sentiment changing over time. There are customers today who want the ability to update software on their controllers. That’s at the core of the argument.

Up until we produced a product called Customer Service Advisor that gave customers the ability to do that, they had to purchase the tool that was very similar to the dealer tool that was used to update controllers. But there was still friction in the system. They had to go to a dealer to get that tool.

They had to have a laptop that could run it.

EL KALIOUBY: They had to pay for it too, right?

HINDMAN: Right. There were technical hurdles between here and there, and it wasn’t very efficient, probably, is the right way to describe it. So last July, in response to that, we came out with Operations Center Pro Service. That’s the trade name, but it’s effectively the ability for you to download any controller payload to your phone, to your smart device, and then walk out to your tractor or your sprayer and push that payload file to the controller that you want to update, as the owner of the piece of equipment.

So we’ve made that update process much more efficient, much more effective, and you can do it through your smart device. That, in a nutshell, was our response to this idea of right to repair, because the argument is mostly centered around the software space as opposed to the hardware side of things.

For our complete existence, we’ve provided service parts and service manuals and those sorts of things to enable customers to do maintenance and service on their equipment as they want, and we still do that today. There are tractors that are 70 or 80 years old that we still provide service parts for. You can still go to a John Deere dealer, buy the service part, and update the piece of equipment yourself.

So for a long time, we’ve been proficient at the mechanical side of making sure that customers have the ability to repair their equipment. I’d say the right to repair is us catching up on the digital side of being able to do that.

EL KALIOUBY: So this is slightly different from the idea that, for example, if my car breaks down, I can take it to my dealership, but I can also go to the mechanic down the street and just fix it for a lot less money.

HINDMAN: Yeah, it’s similar. This Operations Center Pro Service is also available to independent repair shops. So you can take your tractor to an independent repair shop and they can do the software update as well. They can do the mechanical updates too, if you want them to do that.

Copy LinkWhat a fully autonomous mango farm could look like

EL KALIOUBY: OK. Very interesting. All right, so let’s look forward a bit. I want to set the stage. There’s an alter ego version of me that — I’m originally from Egypt, I love mangoes. My grandma had not a mango farm, but she had mango trees in Cairo.

HINDMAN: Cool. Yeah.

EL KALIOUBY: And I’ve actually visited a couple of mango farms in Cairo because I’m like, someday I’ll build a mango farm. I don’t know. But it’s going to be the farm of the future. So I would love for you to describe what that future farm looks like. I want to pull it all together and set a scene.

You get to play.

HINDMAN: Yeah, for sure. Let’s do it.

EL KALIOUBY: Build my mango farm together.

HINDMAN: We’ll do it.

EL KALIOUBY: So we’re starting in the morning. The farmer gets up, and then what’s next? Do you pull up your laptop and look at a data dashboard? Is that step one?

HINDMAN: I actually think there’s a digital assistant that’s probably talking to you once you get up, telling you what the weather’s like on the mango fields that you have, telling you what the potential steps you should take that day are that maximize your farm productivity.

Whether that’s telling you the health of those trees, the state of the trees relative to harvest, what nutrients those trees need, what sort of illness the trees might have — all of those things probably are being communicated to you. I think you’re not reading them. I think you’re listening to them.

EL KALIOUBY: I have a voice AI agent that’s talking to me. OK, all right. And then are the tractors already out in the field? Because they’re autonomous and at 4 a.m. they started themselves and just headed out?

HINDMAN: I think the way I think about autonomy on the farm is that it is a tool in the toolbox for the grower, and it is their choice. So it could be. You could absolutely have that tractor go start its work unprompted if you wanted. That can be a thing that happens. If you want to prompt the work, that is also a thing that can happen.

I think the key thing is, you don’t have to be in it if you don’t want to be in it.

EL KALIOUBY: OK, cool. And then I am really fascinated by the field of humanoid robots. We’re looking at a number of robotics companies that are specifically building humanoid robots for heavy industry, like ship welding and whatnot. Are there going to be teams of humanoid robots in the field, and they’ll have straw hats and overalls or something?

HINDMAN: I like the idea of humanoids in agriculture for a whole host of reasons. There are some jobs that no humans want to do, like in the upper Midwest, going into a grain bin at this time of year and cleaning it out and preparing it for the next season’s harvest.

It’s a dusty, dirty, dark job. You have to wear a respiratory mask. It is not a place that anybody — no farmer, no honest farmer — is going to say, “I love to do the grain-bin job.” What a perfect application for something like a humanoid.

I also think in the mango farm, there’s a whole host of crops that are challenging to harvest mechanically. Fruits and nuts — fruits especially — are more challenging.

So I do think there is, and lots of companies have created bespoke robotic harvesters for grapes, strawberries, tomatoes, and these sorts of things. But I think the thing they all struggle with is the manipulation that humans can provide — this is a hard thing to replicate.

Humanoids have a space in the harvesting part of agriculture if we can get the hand right, because there’s a dexterity associated with that that is just difficult to replicate in other form factors, especially in things like citrus and mangoes and these higher-value crops.

EL KALIOUBY: Berries, exactly.

HINDMAN: Exactly. I think there’s an incredible opportunity in that space for them.

Copy LinkHow AI can improve the entire food system

EL KALIOUBY: Do you see the role of AI in feeding the planet even beyond what it can do for farming?

HINDMAN: Yeah. The food chain — farmers are the nucleus for the things that you and I get to consume on a daily basis. But there’s a whole host of inefficiencies that exist in the food-production system from the point of the farmer all the way through the vertical integration of those feedstocks into the food that we consume, and then the distribution of that food.

So I think there’s a role for AI to play across the whole value chain of agriculture.

We, as humans, compartmentalize this into the agriculture space, and then maybe the conversion of those crops into the raw materials for food production, and then the actual food-production part of things — the products we would see in the store.

And then the consumption and the delivery and logistics of those things. I think nobody has ever been able to look at it and put their hands completely around that whole cycle, if you want to think about it that way. I think AI gives you the opportunity to do that and interrogate where the inefficiencies in this system are and how we can be more effective moving forward.

EL KALIOUBY: And how to reimagine it. Jahmy, this was such a fascinating conversation. Thank you so much for joining us.

HINDMAN: Yeah, it was my pleasure.

EL KALIOUBY: Jahmy shared so many fascinating details on how AI is transforming one of the world’s most important industries, like how they’re using AI to track crop and soil health. I love how Jahmy put it: They’re helping plants live their best lives.

If you’re not in the industry, I think it’s easy to have an image of farming that’s stuck in the past. But who knew the amount of technology powering equipment like tractors? I didn’t.

I also appreciated our discussion on robots. There’s a fear that humanoid robots could replace jobs, but in farming they can take on the work that no one wants to do. This is a great example of AI augmenting — not replacing — humans.

If you’re building in this space, I’d love to hear from you. Find me on Instagram or LinkedIn.

Thanks so much for listening. We’ll be back with a new episode next week.