On-Farm Brains, Digital Twins and Data Resilience
Let’s think this through, with a hyper trained data cruncher to help.
Intelligence is all the rage these days. While we’ve previously covered augmented reality farming and the impact of artificial intelligence on farm work, we haven’t yet talked about farm data ownership, on-farm data storage or hyper-localized intelligence.
Today at AgPunk, we’re featuring a new voice - that of Loren King, a farm-raised fan of drones and all things AgTech. Loren helps us look at the kind of brain our farms might have in the future. Not just human brains, but the potential for a server bank that stores, recommends and outlines potential solutions throughout a farm.
To underscore what this might look like in a positive future, let’s spend the rest of this edition in a story.
The future, now
It’s a late night in the farm shop for Jared, a diversified, regenerative producer who raises dry beans and carrots for local food purchasers. He also uses a few fields to grow mushrooms for biodegradable packaging. Right now, though, he’s tinkering with a critical piece of farm hardware. His entire farm and all the bits it produces are summed up and stored in an on-farm brain. It’s a personal, private server network that he and his employees use to make decisions nearly every day.
Sifting that much data and making it usable is a huge process, one that takes place entirely in a quiet room, right smack dab in the middle of his humming farm shop. For security reasons, Jared works directly in the room when making major adjustments to the data that powers the farm’s digital twin. Adding new efficiency parameters for equipment isn’t exciting, but it’ll save him an estimated 10% in fuel costs once the brain has calculated the best times to start equipment, and the best paths to get it to the field. After a quick upload, Jared finishes his updates and heads out for the night. The brain? Its work is just beginning.
Like biological brains, the farm server doesn’t really turn off. When Jared leaves the room for the night, the system reanalyzes data and compresses old files, prepping for farm work the next day. His data models and structures for field reports are updated with years’ worth of information, and an entire digital twin of his farm changes ever so slightly in response to a new weather report.
By the time he gets back in the morning, the brain will have generated a completely new view of his farm. Every field updated, every bushel of storage mapped and every piece of equipment loaded with a new game plan, just in case he needs to access it from there.
This comes in handy when weather estimates predict an early fall frost and heavy snowfall winter. Jared’s on-farm brain allows him to take that prediction and use the digital twin of his fields to prepare a plan for what to harvest and when. It’s not perfect by any means, but the combination of hyper-local data, advanced computing and Jared’s tried and true farm experience creates a plan that gets the most plants out of the ground, even in extreme weather.
It takes a village
Jared’s got a lot going on, but so does his small community. The area still deals with frequent interruptions to web access due to climate shocks and a lack of available infrastructure. To get around the access, Jared uses a unique solution. His farm brain turns the Wifi repeaters on his equipment, silos and barns into a defacto cellular network. Then comes the big part. The on-farm brain partitions itself for public access. Any extra processing needs for the community? Taken care of. Local storage of Netflix shows and university courses? Made available for easy access. Getting a forecast on a big winter storm? Definitely a possibility.
The system is part of a larger network, one where other farms, libraries and utility stations communicate wirelessly. In most cases people in their homes won’t even notice when the service switches from a local provider to the backup.
In return for this service, Jared gets an extra source of income, and diversification on his farm. There’s an automatic payment system built around how many megabytes of data the Wi-Fi repeaters ferry throughout the community, and the amount of processing of local data that happens in the on-farm brain.
The last time a winter storm swept through the area, Jared had a brief scare when it came to farm storage. A panel on the farm shop swept away, exposing the inside of the building to the elements. There wasn’t any damage, but it did convince Jared to take a look at a backup option. Using quantum cryptography, his entire farm-brain is backed up securely and frequently in a secondary data center kept by a regional telecommunications company. In the event the worst happens, he’s able to access it and restore his system completely.
Jared’s farm supports the community, and it gets a lot of support in return from agronomists, neighbors and the team running daily operations. But on the data side there’s one stream in particular that helps him out. That’s an online access point updated regularly with the latest publicly available data from statewide and local field trials. Extension services, researchers and a litany of experts keep a regular flow of information. All of it is made available for free and Jared regularly takes a look.
One particular night, Jared found a doozy. A state-wide university research hub has some updated findings on micronutrient impacts on yields in a wet spring. Jared took a look at the results and copied the files over into the on-farm brain before heading home.
There is precedent for government data being made available to farmers in a regular, and usable form. Even as far back as 1916, government-distributed data was baked into farmers’ everyday lives. A 1916 edition of Michigan Farm Laws outlines the manner and timing of some of these releases:
“The state board of agriculture is authorized to provide from time to time… the results of experiments made in any of the different departments of the agricultural college, and such other information that they may deem of sufficient importance to require it to come to the immediate knowledge of the farmers and horticulturists of the state.”
Today there’s even more data available. Despite advances, drawing insights from a report takes time and the ability to compare results at a macro level. Jared relies on his farm brain to stay on top of this data and run estimates, calculate probabilities, and spit out potential solutions based on new research data.
Jared’s view is that the ability to fine tune these responses, determine what stays and what doesn’t, would - and even should - stay in the hands of the farms themselves. He’s seen how this accounts for regional variance and personal choice, providing significant benefits to personalized algorithms trained and stored on a farm’s servers.
The models stored in Jared’s on-farm brain are a key asset, one that he needs ready to go at a moment’s notice. Keeping them unbiased is the key to his success. In fact, he has lots of security solutions, simply to make sure that only the right data is feeding back into the brain. Lots of farms do this for training, transaction protection and data privacy so making it standard practice didn’t take long in Jared’s neck of the woods.
No limits
When he first started farming years ago, Jared remembers hearing his farming mentor discussing the then-emergent AI technology like ChatGPT and large language models. His mentor would share impassioned dialogue about the potential for AI to shape farm decisions and workflows. He would cite the research at the time, which outlined the challenges of emerging AI technologies for farmers. Jared remembers these challenges including things like response time and accuracy, the need for big data, high data costs and flexibility, among other concerns.
As Jared reflects on those memories, he feels awe for how far the technology has come. Those limits that concerned his mentor years ago have largely been solved on his farm today. Years of data have helped train and refine accuracy in the system, while access and storage of data are tied to the very server room built into his shop. Implementation required relying on experts, but by working with the right people Jared was able to design a brain to meet his farm needs, and be flexible enough to work in just about every situation he’s come across. Anything that he wasn’t prepared for? Well that’s just farming.
Back to the future
Jared’s work today brings him out in the field. He directs a stream of small, swarm-based equipment around his field to measure nutrient levels, check foliage and green cover for anything that might need replanting and evaluate the merits of utilizing a new cover crop this fall and switching to sugar beets next year. He looks at his phone, checks in with the farm brain through a secure VPN, and gets a green checkmark next to the farm brain’s dry bean estimate. In a few months, the field he’s standing in is going to look a lot different.
Until next time, futurists.
Future Questions
Here’s a few thoughts that came to mind while drafting this issue.
Can we train individual on-farm AI models?
Will farms serve as local data centers when rural areas need excess capacity?
Are digital twins a technology that should be incorporated into a farm brain?
What’s the best way to collect data to train these models?
Contributors
Author: Loren King. Farm-raised fan of drones and all things AgTech. He wants to know what people think. Find him on LinkedIn.
Other Contributors:
Editing: Sami Tellatin
Sources
Web-based sources hyperlinked where relevant.
Additional source without hyperlink: Michigan Farm Laws, 1916 edition
AgPunk Playlist
For some positive punk vibes to accompany your AgPunk reading, check out our playlist on Spotify. In it are themes of liberation, positivity, revolution and pure punk spirit.
 | A guest post by
|