Farmers look to artificial intelligence as workforce declines
Under the hot Florida sun, hundreds of strawberry pickers dot the rows of Wish Farms. It is back-breaking labor as the pickers fold their bodies to lean over the plants. Their hands skim through the leaves to find the bright ripe strawberries. With a flick of their wrists they pop the berry from the stem, carefully making sure not to tear the plant. But that’s not the hardest part about harvesting this crop.
A strawberry plant has a ripening cycle that’s much different from other fruits; every three days it will give off new ripe fruit. That means each plant will have to be revisited by the pickers every three days during that season. If they aren’t picked on that cycle the plant could quickly die.
Wish Farms has been in Gary Wishnatzki’s family for three generations, since 1922. Over the last 15 years he said he’s had a tougher time finding a consistent labor force to pick his crops.
“We’ve seen a shrinking labor force and an aging labor force,” he said, adding, “people just aren’t showing up any more to do this work.”
Wishnatzki started seeing the trend in 2011, but he realized it had actually been going on since the early 2000s.
“I started to look at what was really causing it,” Wishnatzki said. “It was more of a demographic issue than a political issue or anything else.”
He said the same pickers in the field have been there for a generation, and their children aren’t arriving to do the same work.
In order for him to keep his crop healthy, he needs 600 workers to harvest his 600 acres every two to three days. That means in season, there aren’t any days off. Currently his work force is mainly from Mexico on a short-term H2A visa program. He said the program is expensive and he knows it won’t last.
“As the Mexican economy continues to improve as more fruit is being grown in Mexico, people aren’t going to be coming anymore,” Wishnatzki said. He believes that eventually Mexico could be a net importer of labor.
Wishnatzki was convinced that automation would be a game-changer and so he joined a team of engineers to try and create a robot that could mimic what a picker in the field does. Five years later, Harvest CROO Robotics has a robotic picker which is able to look at a strawberry plant and determine which among the berries on the plant is ripe, all through artificial intelligence.
Paul Bissett, Harvest CROO’s chief operating officer, showed CBS News the robotic harvester in action. He said each component of the picker mimics what a human does. At the top of the harvester, two cameras take images of the plant.
“We’re collecting 50 to 100 images a plant, and all of those images are fed into our AI system in order to tell us, OK, this is a good berry, this is one we want to go after,” Bissett explained.
Once commercialized, they hope each harvester will be able to cover the same ground as 30 workers. Currently, Bissett recognizes the harvester, now on its fifth iteration, isn’t ready for market, but he hopes by the seventh it will be. While it does recognize and pick the fruit correctly, the next challenges become: how do you store the strawberries, pack them correctly, and make sure they are presentable to a customer?
Bissett is confident they will solve those problems, and they have a lot of people counting on them. Two-thirds of the nation’s strawberry industry has invested in the machine, all hoping that an answer to their labor issues will materialize. Bissett said their efforts are critical, not just for the strawberry industry.
“Without automation into our nation’s food basket our food availability is going to change substantially,” he explained.
If they don’t overhaul the industry, he estimates the price of strawberries could increase to $10 a pound for consumers. “That is going to eliminate this food and other specialty crops from the plates of millions of Americans,” Bissett added.
But it’s not just family businesses that are turning to automation. Farming equipment giant John Deere said they have been investing in machine learning and artificial intelligence for a decade. In the last year, the company has made a major push into Silicon Valley, opening an office in San Francisco to focus on emerging technologies. Alex Purdy is the head of that new venture, John Deere Labs.
“We know that we absolutely need to go and seek out capabilities in machine learning, in deep learning, in robotics and advanced analytics,” Purdy said. He recognizes they will “go wherever the talent needs us.” And he says Silicon Valley has been receptive.
“I have people knock on my door all the time and say, ‘I can’t believe John Deere is here,'” he joked.
One of the companies that John Deere approached is Blue River Technology, which uses a “see and spray” technique to map fields and visual recognition software to target weeds and apply herbicides only where needed.
“It will allow it to save resources to use less herbicides,” Blue River Technology CEO Jorge Heraud said. He believes the upside won’t just be for the farmer who doesn’t have to pay for as much herbicide anymore, but “also for the receiver who gets crops that are grown in a sustainable way with less herbicides.”
John Deere bought Blue River last year for over $300 million, the first large move in an effort to expand on machine learning.
“It’s really going to help a grower optimize their land and their operation by being able to make decisions in real time based on in-field conditions,” Purdy said. He said that with projections of 2.5 billion more people in the world by 2050, agricultural output will have to double. “Advancing that in a sustainable way through the use of technology really is impactful and important,” he added.
“I say always have a healthy amount of skepticism about whatever new technology you’re using. No technology is a silver bullet,” warned Danielle Nierenberg, president of Food Tank, a non-profit that studies food sustainability. While she admitted she is interested in these emerging technologies, she pushes for answers to many questions, including: What happens if farmers become tied to the machines and they break down? Will an average farmer be able to adapt to the technology? And what will happen to the workforce?
“Automation in and of itself is not good or bad,” Nierenberg said. “But we need to consider the people who are involved and make decisions accordingly and make sure workers are trained and have different skill sets so they can have other jobs in the agriculture industry.”
Purdy said John Deere recognizes all of those issues but said he hopes this turn to automation and machine learning will help farmers rather than displace jobs.
“I think our solutions are just ways to give growers more tools to make their land better and to improve their productivity and profitability,” Purdy said. “But by no means do we expect the grower to not be in the dirt and in and around the muck.”
That is what Richard Watson of Seven Oak Dairy Farm in Georgia has found. He is testing out a new technology called Ida on 200 of his 2,000 dairy cows. Ida is created by a Dutch company Connecterra, which is using TensorFlow, Google’s AI software. The data for Ida is generated from collars the cows wear. It can detect when a cow is eating, drinking, ruminating, sleeping and walking. All of those data points are digested by Ida and TensorFlow to give the farmer a readout of where his cows are, health-wise.
When we visited Watson’s farm, he found three of his cows were targeted by the app for showing signs that they were sick, and one was possibly in heat. He said the app saves him a “massive amount of time.”
“We would like for heat detection, for example, literally, we would have to go through all 2,000 cows on a daily basis and check each one individually,” Watson said.
But Watson recognizes the app is not a replacement for him and his workers. “Technology doesn’t compensate for poor farming,” he emphasized. “At the end of the day, you still gotta be a good farmer. What the technology enables you to do is become a more efficient — perhaps diagnose things quicker. And — and treat things quicker. And that’s more important when you have several thousand animals.”
Watson noted that there has been a major decrease in the number of dairy farmers in the U.S.; in 20 years they’ve dropped by almost 50 percent, but the number of dairy cows remains steady. That means that there are more cows per farmer, which makes it harder to keep track of cows and if they become sick.
“Back when it was 80 to 100 cows, the dairy farmer just about knew their animals by name — could tell you just about everything about them,” Watson said. “We just can’t do that now.”
But he believes innovations like Ida may help existing farmers and possibly entice a new generation into the field, which is critical. The latest USDA farming census showed the average age of a farmer at 58.3, a number that has been increasing since the late 1970s, and more than 60 percent were over the age of 50.
“Not everyone wants to get up at three in the morning, particularly if it’s wet and cold…and milk cows,” Watson said. But he adds there is a “misconception about agriculture that is somehow not as high-tech.” Watson points to innovations like IVF starting from agriculture and hopes that maybe artificial intelligence could be the next pioneering ground.
“Agriculture is actually a very innovative and interesting area and field to get into. I think these technologies really promote agriculture, and make it more user-friendly, and hopefully encourages these people to come and join us,” Watson said.