Artificial intelligence (AI) is permeating nearly every facet of society, and soon it will be a regular fixture on dairy farms, too, according to Miel Hostens, Robert and Anne Everett endowed associate professor of digital dairy management and data analytics at Cornell University.
On a recent episode of the “Cornell Cow Convos” podcast, Hostens says he currently considers AI to be in the exploratory stage.
“At this moment, I think we’re in a bit of a ‘hype cycle,’” he says. “We’re just starting to see some of the advantages that could potentially be applicable to the dairy industry.”
Hostens says there currently are very few AI-based technologies that are ready for commercial farms, but that will be changing quickly as companies strive to apply the technology to assist dairies in meaningful ways. Some of the potential applications he mentioned are parlor management, lameness/locomotion scoring, automated body condition scoring (BCS) and calving detection.
He says the origins of AI have been in place for decades but have been bolstered recently by cheaper and more robust computational power, coupled with natural language processing techniques and photography and video technology.
Some “baby steps” toward full-on AI have already be implemented on dairies, such as using basic statistics to program systems to signal alerts. But in that case, humans made the decisions and set the data thresholds, versus the machine learning that is the cornerstone of AI.
“Nowadays, with machine learning, you can throw a whole bunch of data at these algorithms, and the algorithms are able to find patterns themselves without humans being involved anymore,” he explains.
There will be questions and even ethical considerations along the way. For example, it’s one thing to capture and apply data on cows for BCS or locomotion scoring, but is it an invasion of privacy to analyze and apply human behavior in parlor management? Hostens says that is yet to be determined.
And who, exactly, owns the data?
“The ownership stays forever with the one who creates the data and that’s the farmer,” declares Hostens, who also is director of Cornell’s Bovi-Analytics Lab.
But, he adds, unlike a tractor or piece of land, it is possible for more than one entity to own data. In addition to the farm, it could be legally possible for the AI company to own a common set of data.
Just like a robotic milker or feed pusher, Hostens says it’s also important to have a plan in place for breakdowns. If the system is knocked offline or quits functioning for some reason, who will step in to take its place?
In preparation for the onset of AI, Hostens advises dairy managers to take the following steps to maximize it in their business:
- Get your data ready – Think about the issues on your dairy that could be solvable through AI, and begin collecting data sets on them now that could potentially be plugged into an algorithm later.
- Consider monetizing it -- Hostens says AI companies cannot train their algorithms and deliver services to dairies without the ground truth that happens on a dairy.
“If you are generating data for an AI company, you can ask for something in return,” he advises.
Hostens says AI will obviously not eliminate the need for people to care for animals on dairies, but it will shift employee needs in some ways. For example, fewer people might be needed to manually perform BCS or walk pens to monitor calving, but someone will need to regularly maintain the equipment that performs those tasks instead.
“Those systems will have to be running all the time,” he says. “What we tend to do is put those cameras high on the ceiling and hope that they will monitor forever. But you will need to have some kind of maintenance system around it.”
AI systems also still need to be developed to a point that they are generalizable across farms, Hostens says.
“But the promises are big, that’s for sure,” he adds.
Your Next Read:
Robotic Milking Success: It’s More About the Management Than the Technology
