How Artificial Intelligence Can Help Close the Loop in Greenhouse Production
In August 2017, iUNU announced it had secured $6 million in venture capital funding from big-name investors – including Reddit’s Alexis Ohanian and football great Joe Montana to name a few – to officially launch its new greenhouse artificial intelligence platform, LUNA.
iUNU CEO Adam Greenberg says for his company, developing an AI solution is not about replacing growers; it’s about augmenting their operations by allowing them to work proactively instead of reactively to solve plant issues in real time, and redirect how they use their resources.
A good analogy for the way iUNU approaches artificial intelligence with its LUNA platform is iUNU Chief Technology Officer Matt King’s frustration with his inability to recreate a favorite recipe that his grandmother makes – white beans and ham.
“She knows exactly when to check it, when to taste it, when to tweak it, and how to make it right – and it comes out amazing every time she cooks it,” King says. “When I read the recipe that she’s written down for me, and follow the steps, it comes out okay sometimes. The difference isn’t just that she knows more and has years of experience. It’s that she knows when to check, what to check, and what can’t be ignored. She knows how to instrument it and collect the data, and she can think back and adjust what she does, and of course, she gets the same desired result every time.”
That, King explains, is the difference between a closed loop system and an open loop system. His grandmother is working in a manual, but closed loop system. When he tries to blindly copy the recipe, he is working in an open loop and hoping for the best. Grower operations experience the same challenges when they lose experienced growers, he says.
“Without technology, growers and their experience are the only data-collection apparatus, and the only link to the feedback,” he says.
There are many companies trying to implement technology in agriculture that aims to replace a grower with artificial intelligence by automating diagnostics and decision making. But King says iUNU’s approach is different.
“We think growers are really good at being the decision maker,” he says. “What they need is the technology to automate the data collection, and get the feedback to them faster, to integrate the greenhouse automation, close the information loop, and let the grower focus on what they do best. We’re building an AI assistant, not an AI overlord. This is not telling growers what to do – it’s giving them information.”
The theme used for building LUNA focuses on closing the loop, and putting the plant at the center of the loop that’s measuring the plant, King says. The data that is collected can then drive growers’ decision making.
“Our AI is a bit more under the hood, and what it’s doing is recognizing the plants as inventory units, as a manufacturing process; it measures the performance of the plants in a standardized process with standardized production techniques. Our AI is instrumentation that automates looking at an industry that produces the contextual understanding of which plant is which, where it is in the growth cycle, how it’s doing as a baseline of that particular grower.”
Other systems look at massive amounts of data generated, and then makes predictions based on that data. King says LUNA is much more specific than that.
“Our system counts individual flowers and plants, and millions of plants, which allows the system to be specific. It’s not generalized or a guess as to what may happen based on a trend. It’s what is actually going on right now, with every single one of your plants,” he says.
The Value of Retaining Tribal Knowledge
King’s difficulty recreating his grandmother’s recipe is truly analogous to the challenges greenhouse operations have when an experienced grower moves on and takes that tribal knowledge along – knowledge about growing at that operation that often is not recorded anywhere.
Greenberg says the turnover in the industry is one of the problems the company homed in on when creating the LUNA platform. Oftentimes a recipe binder or worksheet would have been created, but it wasn’t referenced or the nuances of the recipes that were not written down were lost when the grower left, he says.
“From a tribal knowledge perspective, when companies have to hire somebody new, then these growers would learn on company time how to get back to the recipe and the standardization and repeatability that took the grower before them three years to perfect,” Greenberg says. “So the idea that this knowledge stays with them in the system, and this knowledge stays in the company, not only allows for them to have better repeatability, but it also allows them to learn faster, too.”
That baseline knowledge is built on the idea of continuous improvement, as opposed to getting back to the original baseline, so greenhouse operations don’t get less efficient when they lose a grower – they stay as efficient or better, Greenberg says.
“This tribal knowledge is not only stored within the company, but it also adds value to the company,” he says.
That added value translates to a faster return on investment, despite a probably higher price point than most growers may be accustomed to for implementation of a system like LUNA. But while AI may seem intimidating and expensive at first, Greenberg asks growers to consider the cost of not having this technology.
“It’s only expensive if you’re not seeing more value out of it than what you’re getting,” Greenberg says. “There is a value proposition that in almost every aspect of the organization, whether you’re talking about sales and sell through or the organizational structure, or the financial visibility and cost accounting, or the growers’ time, or even the facilities, of where do we fix problems. All of that has everything to do with business throughout the organization. And when you talk to our customers, they don’t actually look at the expense, because we’re only taking a fraction of what they’re actually saving. In the grander scheme of things, it’s almost like a commission for helping their businesses as a partner.”
How Growers Can Use Their Collected Data With AI
The data generated through the system is always owned by the grower operation where the system is in place. LUNA makes the data accessible through the computer and handheld devices, and it’s stored in growers’ ERP systems and the cloud. The infrastructure of the system allows growers to put their data to use, Greenberg says.
Once the system is in place, it provides a recorded history from that time on, and growers can use the data collected to compare plant growth patterns, crop health, and a number of different variables.
“It’s so simple yet so powerful, just having an actual, recorded memory,” King says. “Because trying to remember what plants looked like a year ago and look through your notes and see what was going on, instead of being able to actually bring it up, play back the video, look at all the data – the temperature, pH, watering schedule, when did somebody trim, when did we harvest, what was the temperature outside, what was the weather like when that HVAC unit broke and we had to move crops – what happened? When you look at the data, you realize that probably every year, we’re solving the exact same problem, because we don’t have a record of solving it last year.”
Greenberg adds that without this type of record, “growers will solve the same problem three years in a row, and it will take them three years to realize that pattern.”
Another perk of the system is the image recall, Greenberg says.
“When you have some discoloration or some kind of issue on your plants, oftentimes they’re not 100% sure of what it is, so they’ll watch over it the next day or two to solve the issue. But with LUNA, you can go back in time, and then you can actually solve the issue right then and there, as opposed to having to wait and see how it spreads. You can solve the problem as soon as you catch it, instead of waiting to understand the root cause of the problem.”
Growers’ Biggest Surprise About AI: Automated Inventory
The biggest thing growers have reacted to regarding LUNA and its capabilities, Greenberg says, is that it offers automated inventory, so there’s no need to use bar codes and RFID.
“The simplest way to put it is, the steps that were created for facial recognition, we have repurposed for plant recognition,” Greenberg says. “Every plant has its own unique number. So even if it has the same genetics, the branches and the petioles all have very different leaf articulation. No plants grow the exact same way, with the exact same structure, with the exact same leaf and flowers at the exact same time.”
King offers another analogy to humanize that idea.
“LUNA knows plants like you know your children,” he says. “That’s the AI.”
Brian Sparks is senior editor of Greenhouse Grower and editor of Greenhouse Grower Technology. See all author stories here.