By Brandon Alexander October 3, 2018

CEO of Iron Ox and former engineer at Google X's drone-delivery program, Project Wing

Growing up on my granddad’s farm—he grows cotton, peanuts, and potatoes in Texas—I often heard that technologies like genetically modified crops were required to scale food production. My granddad believed that organic practices do not scale and will not feed the world at an affordable price point. Given the state of technology then, I believe he was right.

But the industry has changed with much more than the seasons, and we need more than an iterative improvement on past technologies. We are reaching a plateau in food production. According to research published in Nature, about one third of the world’s agricultural lands have maxed out the amount of rice, wheat, and corn farmers can grow.

At the same time, the World Resources Institute suggests we will need to double our food production by 2050 to feed nearly 10 billion people. In order to boost yield, we need to systematically improve the entire grow process and maximize the potential of every plant. To enable this next age of agriculture, we will need to rely on two new advances: machine learning and robotics.

The amount of data available to farmers has skyrocketed. In addition to collecting data at a macro level from satellite or drone imagery, we can also capture data at the micro level, thanks to a combination of cheaper, lower-powered sensors. These sensors provide farmers with insights like hyperlocal measurements on soil conditions, for example.

But a firehose of data does not equate to insights, and that’s where the newer methods of applied machine learning come in. Companies like Descartes Labs and Farm Logs are applying machine learning and computer vision to glean insights from these new data streams, providing farmers not just pretty graphs, but actionable information to increase yield.

Machine learning and computer vision enables us to scan each plant in acres of land, detecting plant diseases before they spread and significantly minimizing yield loss and the need for pesticides. For example, traditionally a farmer would inspect parts of a plot of land for plant diseases like powdery mildew or signs of pest pressure like aphids. Because it was physically impossible for them to inspect each plant on acres of land, they would have to extrapolate their findings across the entire plot. Now, modern computer-vision techniques can take multiple images of every plant and stitch them together for a full 3D reconstructed model of the produce.

Data may give us the information we need to improve yield, but something still needs to perform the action. And it’s increasingly not humans.

We are experiencing a growing labor-shortage epidemic. According to the US Census, the average farmer is 58.3 years old, and new generations are not inspired to take on the laborious task that their elders did—even those who have generations of farmers in their family, like myself. This issue isn’t a shortage of food: It’s of people. Crops are rotting on bushes and vines because there aren’t enough staff to maintain and pick them. Considering that one in nine people onEarth aren’t getting adequate nutrition every day, it’s devastating.

This means that though there are more mouths to feed than ever, there will be less land to provide them food, and less calloused hands to tend to the crops that will feed them. So what do you do when you have little land to work with and fewer hands to help? You turn to technology.

Automation allows for a more accurate work environment with little human oversight. It will involve hardware that is more agile than the human eye or hand, and it will be able to give each and every plant the unique attention it needs.

Recent advancements in computing power, dexterity, motion planning, and computer vision are enabling a new generation of robotic applications. Robotics excel at rapidly performing repetitive tasks, but combined with computer vision, robots can start making real-time decisions on a per plant basis, from adjusting the nutrients to pruning. Companies like Blue River have successfully automated tasks like weeding (a manual process for non-GMO crops) to great effect, which is why John Deere bought the company for over $300 million last year.

At Iron Ox, we’ve designed the entire grow process with a robotics-first approach. That means not just adding a robot to an existing process, but designing everything, including our own hydroponic grow system, around the robotics. In an indoor farm, tasks like seeding or harvesting are happening thousands of a times a day. These labor intensive, repetitive tasks are perfect for robotics. And by integrating machine learning and computer vision, we’re able to have the robots respond to an individual plant’s needs. For example, our robot can quarantine a plant if it shows early signs of pest pressure before it contaminates others nearby or change the nutrition recipe for a plant based on phenotyping.

And we don’t even need arable land: By creating indoor farmhouses with these technologies, we can open farmhouses in any location. This means we will be able to control the weather and take chance out of the growing process; currently, we’re losing more and more crops every year to drought, extreme heat and cold, and spontaneous weather incidents. We will also be able to grow crops closer to the communities that need them, reducing the amount of miles travelled to consumers’ kitchens and the industry’s carbon footprint writ large—and for much cheaper.

If farms are to survive, we need to think about them as tech companies. And that means they should be taking advantage of what many other industries are already harnessing: automation.

This story is part of What Happens Next, our complete guide to understanding the future. Read more predictions about the Future of Food.