Got AI? Connected Cows, Artificial Intelligence, Your Milk
'Cow Fitbits' and artificial intelligence are coming to the dairy farm, but some farmers aren't impressed.
WAYNESBORO — In the two months since Richard Watson strapped 200 remote-control-sized transmitters around his cows' necks, an artificial-intelligence system named Ida has pinged his phone with helpful alerts: when his cows are chewing the cud, when they're feeling sick, when they're ready for insemination.
"There may be 10 animals out there that have a real problem, but could you pick them?" he said one morning, standing among a grazing herd of dairy cattle wearing what he calls "cow Fitbits."
But on the neighboring pastures here in rural Georgia, other farmers say they aren't that impressed. When a cow's in heat, they know she'll start getting mounted by her bovine sisters, so they smear paint on the cows' backsides and then just look for the incriminating smudge. No fancy AI required.
"I can spot a cow across a room that don't feel great just by looking in her eyes," said Mark Rodgers, a fourth-generation dairy farmer in Dearing, Georgia, whose dad still drives a tractor at 82. "The good Lord said, "This is what you can do." I can't draw, paint or anything else, but I can watch cows."
Sophisticated AI technologies are helping reinvent how Americans work, offering powerful software that can read and react to mountains of data and save them time and stress along the way.
But its rollout is also sparking tensions in workplaces as humble and old-fashioned as the dairy farm. That down-home resistance raises a question farmers might be tackling before much of the rest of the workforce: Can new technology ever beat old intuition — even when it comes to a bunch of cows?
No fancy AI required.
The AI that Watson's farm uses - called Ida, for "The Intelligent Dairy Farmer's Assistant" - tracks his cows' tiniest movements through their collars and then graphs and dissects them en masse. Those "real-time cattle analytics' are then used by the AI to assess diet and movement and predict concerning health issues, such as lameness or udder infections.
As silly as this intricate level of maximum optimization sounds, particularly for a herd of cows that spend much of the day staring blankly or relieving themselves, Watson said it could mean the difference between a cow's healthy milking or premature death - and the difference between making or losing hundreds of thousands of dollars every year.
The Ida AI has sparked some early interest among farmers eager to compete in an industry in which low milk prices and farm layoffs have everyone on edge. And while truck drivers and cashiers see AI as a job-killing omen, the farmers say they're in a labor crunch from years of too few young people getting into farming and need all the help they can get.
Connecterra, a development team based in the Netherlands, built Ida with help from TensorFlow, the giant AI toolbox that Google created for its own apps and opened to the public in 2015. That release sparked a major wave of AI development, giving start-ups a shortcut to calculating advanced mathematics and creating learning machines.
Google has joined other tech giants in pushing forcefully into AI, with chief executive Sundar Pichai telling a town-hall crowd in January that AI "is one of the most important things humanity is working on" and "more profound than electricity or fire." (He did not explicitly mention cows.)
The "secret sauce".
Standing one March morning among his cows at Seven Oaks Dairy, one of three farms he runs as part of his Hart Agriculture brand, Watson pulls out his iPhone to show off his Ida app. The AI says he has three "potential health problems to be checked" among his herd: Cow #14433 is eating less, while cows #10172 and #3522 are "ruminating" or chewing less, a sign they might feel ill. His herd's "to be inseminated" count is at zero, as signified by a reassuring green check mark.
At 6-foot-4, with combine-wide shoulders and a Kiwi accent, the New Zealand-born Watson, 46, looks like a rugby player - which he was, playing a linebacker-like position in the late "90s for a semi-professional team called the Hurricanes. Shortly afterward, he moved to lead a cattle-grazing research program at the University of Georgia, where he taught and advocated the increasingly rare craft of letting cows amble about aimlessly on a pasture, eating as they go.
His farm's cattle - crossbreeds of America's classic black-and-white moo cow, the Holstein, and New Zealand's relatively slimmer brown Jersey bulls - spend almost all day grazing on the thousands of acres of ryegrass and bermudagrass that surround his farms. That makes tracking their free-range eating and movement harder than at the average American "confinement" dairy, where cows are kept in stalls and fattened on corn and grains.
Spotting problems the old way required closely watching the herd day and night, "unless it's really obvious - you know, she's walking or limping or there are buzzards flying overheads," Watson said. "Buzzards aren't a particularly good health program."
The cows' orange transmitters beam data over the hills of Watson's pastures to a set of antennae near the milking parlor. A "base station" computer then gulps up and processes all that sensor data, doing much of the AI work locally to avoid the problem of spotty rural internet. The sensors pay the price for much of this data exchange, Connecterra co-founder Yasir Khokhar said: "You don't want to know what cows do with them."
The Ida AI was first trained to comprehend cow behavior via thousands of hours of video and sensor inputs, as well as simpler approaches, including Khokhar mimicking bovine techniques with a sensor in his pocket. ("I was the first cow," he said.) Every day brings more cow data and farmer feedback that help the AI learn and improve. The AI, Khokhar estimates, has processed about "600 cow years of data," and is gaining about eight years of new cow data every day.
The AI now logs seven distinct cow behaviors: Walking, standing, lying down, eating, chewing, drinking and idling. Other behaviors are on the way, Khokhar said, though he could not disclose them, calling them part of the "secret sauce."
Dairy farmers have used sensors for years. But Ida's developers say its AI can do things old programs can't, by learning from the cow behavior patterns that can pinpoint injuries, predict the onset of certain diseases and "predict peak ovulation time with over 90 percent accuracy." The AI can also track how changes to cows' bedding, feed and environment can affect, for instance, how much milk they're making or how much they're lying around.
Khokhar, who said he conceived the AI idea while living on a Dutch dairy farm, launched his start-up in late 2016 and now counts a few thousand cow "subscriptions' across farms in seven countries, including the U.S., Spain and Pakistan. The company covers all the equipment and service work and sells monthly subscriptions. Farmers' prices start at about $3 a month per cow, plus a $79.99-per-cow start-up fee, and Watson estimates he has invested about $17,000 on the system so far.
Agriculture has long been one of Big Tech's juiciest target industries. Revamping the way farmers feed the planet, in the face of existential crises such as food shortages and climate change, would be audacious, revolutionary - and highly profitable. Start-ups and farmers are now using camera-equipped robots to pick apples and sort cucumbers, running driverless tractors to harvest grain, and flying scanner drones to spot poachers and survey livestock.
Beyond the Ida collars, other tech start-ups make cow pedometers, robot milkers, tail sensors and electroshock collars that can stop or shift a herd.
You cannot bore a cow to death.
But even some farmers who have invested heavily in new technology balk at the idea of paying for more. Everett Williams, the 64-year-old head of the WDairy farm near Madison, Georgia, said his farm has all kinds of sensors that print out who-knows-how-many reports on matters such as cow activity and if hogs have gotten into his pens. They give him less data than the Ida AI would, he says, but he feels as though he doesn't have the space in his brain for another data stream. "You can only handle so many text alerts," he said.
Systems such as Connecterra are also enduring early criticism beyond the farm. Because the AI can help detect early disorders and walking disabilities, conservationists have criticized the systems as encouraging the breeding of a super-cow by speeding underperformers to the slaughterhouse.
Rodgers, who runs his "daddy-daughter" family dairy farm in Dearing, Georgia, said he's no Luddite when it comes to farm technology. His "super-system" features cow-tracking transponders and, soon, a DeLaval VMS, which milks cows with lasers and robot arms and is advertised as the "ultimate automatic milking machine."
His system, unlike Ida, doesn't track cud-chewing or use AI to tell him which cows to watch or what to do. But he's okay with that. That's the way things have always been done here, and he hopes they'll be that way for a long time.
"There's no substitute for watching your animals. It's an art and a science, and I hope my daughter and nephew get better at it than I am," he said. The cattle, he added, don't care much about evolving with the times. "You cannot bore a cow to death."