These days, there's a lot of hand-wringing over the possibility of self-driving cars on our roads. But we've largely overlooked the fact that, in the fields just off the side of the highway, vehicles have already been driving themselves for the last 15 years.
Self-driving tractors are becoming more common. A John Deere spokesperson told me the company currently has about 200,000 self-driving tractors on farms around the world, from the US to Germany. And they're just one example of a major investment that the agriculture sector is making in artificial intelligence and the Internet of Things. When it comes to farming, AI can take many forms, from providing simple navigation like with self-driving tractors, to more complicated crop analysis and decision-making.
There are economic drivers for this: automation and AI can make agriculture more efficient and cost-effective. But there's a very existential reason to integrate AI into farming, too. The global population is growing, but the amount of land on the Earth we can use for farming isn't. By 2050, there could be as many as 9 billion people on the planet, according to the Food and Agriculture Organization of the United Nations, and to meet that demand, we need to increase our global food production by 60 percent.
"The population on the planet is still growing but agricultural areas are limited, so the land that is there has to be managed more effectively and efficiently," said Christian Bauckhage, a data scientist from the Fraunhofer Institute for Intelligent Analysis and Information Systems. "This is where these kinds of technologies come into play."
Bauckhage first got interested in the agricultural applications of AI when he was tapped to contribute to a project at the University of Bonn in Germany. Botanists there had collected reams of hyperspectral image data—photos that capture light outside of the visible spectrum—for different plants and needed a data expert to help them sort through it. These hyperspectral images show early signs of plant disease that aren't visible with the naked eye, allowing outbreaks to be detected, tracked, and treated earlier. Bauckhage and his colleagues were able to create an AI system that analyzed the images to detect disease before a farmer would ever spot it, and recommend early courses of action.
This is just one discrete example of the ways AI is being integrated into food production, but it follows similar patterns to a lot of other work. The major focus is on data collection and analysis, which is then fed into smart, connected farm equipment, which in turn can communicate with other pieces of equipment, and with the farmer.
The goal is to inject more technology into precision agriculture: a modern approach that focuses on the needs of small, specific sites within a single farm, rather than treating it as single thing. So rather than spreading the same amount of fertilizer all over your 40 acres, you'd use AI to analyze data for each individual acre, and determine needs at a microlevel. Ideally, this approach allows farms to not only be more productive, but also use less water, fertilizers, and pesticides.
"We need to double the world's food output by 2050."
"Agriculture is becoming even more industrialized," Bauckhage told me. "It's moving away from the traditional image we have of a farmer who walks his or her fields, and pats his or her cows. Like many areas of modern life, it's becoming data intensive. That might a bad thing. It might be a good thing in the sense that yields may increase, and we don't need to use as much pesticides, things like that."
So far, AI in agriculture still functions with assistance from humans, and most of the systems aren't able to run independently. But they're beginning to.
"With our self-driving tractor, we have two levels of adoption," said Ron Zink, the director of on-board applications at John Deere. "A lot of people will have the tractor do the straight lines, but then manually turn the corners for the tractor. But others are actually planning out the entire route for the field."
Zink told me John Deere estimates there's at least a 10 percent increase in productivity when farmers use self-driving machines, because they're equipped with GPS and low-Earth satellite telematics navigation systems that enable these machines to drive more precisely and avoid overlap. That way, farmers avoid harvesting already-plucked rows, or throwing down seed where it's already been dropped.
Beyond automation, data analytics are creating important spaces for innovation in farming. There are dozens of examples, from facial tracking livestock to keep wildlife from bringing possible disease into farms, to a process called variable-rate seeding. With this, smart harvesting equipment is rigged up with computers that collect data about annual yields. That data can then be communicated to a program that analyzes it and suggests adjusted seeding plans so the most productive areas are being taken advantage of.
"Let's say one part of the field has very good soil and a lot of variability when it comes to yield, and another part of the field isn't so good," Zink explained. "This gives you the ability to say 'OK I'm going to plant more seeds where it's more productive, fewer seeds where the soil's not as good,' and then I might do the same thing with my nutrients."
Again, the farmer is still involved in this process—the farming machines aren't sharing data with each other and making decisions independently—but it's getting easier to imagine a future where farmers are able to take more of a hands-off approach, and just monitor, allowing them to multitask. While this could potentially eliminate jobs, the reality is that the average age of farmers in most of the developed world is around 60, so along with pressing demand for increased food production, our food producers are starting to age out, with not enough young people coming to take their place. It could be one of the few industries where robot workers start to seem more like a savior than a threat.
"Some experts are saying we need to double the world's food output by 2050," Zink said. "You're not going to substantially increase the arable farmland, so what's going to help us do this? We think technology will be more than 50 percent of the answer."
Correction: An earlier version of this story indicated John Deere has 2,000 self-driving tractors on farms around the world. It has 200,000.