The Growth of AI in the U.S. Food System

Ever since the First Agricultural Revolution, farmers have been using technology to advance their craft and improve their yields. The scythe, the silo, steam tractors, and GPS sensors have all helped them grow food and, ultimately, remake society. These innovations have accelerated the development of new economies while also making it possible to feed billions of people at scale. We must explore the role of technology not only in improving production efficiencies, but also in ensuring that everyone has access to healthy and fresh food.

illustration illustration illustration

Today, artificial intelligence and machine learning offer the opportunity to bring a greater level of certainty to the notoriously uncertain business of farming. By using the processing power of AI to collect and analyze multiple sources of data out in the field, new tools are allowing farmers to benefit from the insights that data analytics can generate. Precision technologies are replacing the Farmer’s Almanac, predicting everything from temperature and rainfall to pests and commodity prices with greater accuracy. Agricultural engineer Amanda Ramcharan is helping to transform the “field” itself through her work introducing predictive technologies to measure the resilience of soil in the face of droughts or floods. Ramcharan sees these innovations as vital to the future of farming. “You have to start using data, you have to start using machine learning technology,” she says. “That’s the only way that we can keep growing more food with less labor.”

Meanwhile, apps like FARMWAVE, PlantVillage, and Plantix, use deep learning algorithms to diagnose plant diseases and pests. Cameras attached to drones can detect weeds and monitor crop health from hundreds of feet above. With collaborative online platforms for chatting about shared pests, farmers can increase their yields, reduce overhead costs, and produce more fresh food for their communities. And much of this farming technology can learn on its own, improving its performance without engineers having to constantly make adjustments.

The symbiosis between technology and farming isn’t new—there’s even a link between the cultivation of cereal crops and the birth of written language—but recent advances have the potential for big breakthroughs in the ways we grow, store, transport, distribute, and consume food. From production to consumption, this digital transformation, in tandem with new ecological services, will prove critical to reducing greenhouse gasses, addressing the multiple causes of food insecurity, and feeding the planet in the 21st century. It couldn’t come at a more opportune time: It’s estimated that global food production will need to increase 25-70% by 20504 in order to nourish a population predicted to reach 9-10 billion in the next thirty years.

We also have the potential to utilize the valuable insights and predictions produced by AI to create more equitable food systems. By using advanced technologies to look for patterns across vast reams of historical data, we can predict future outcomes, helping us address critical food security issues—from global hunger to obesity, food waste to famine. Grocers can now use “predictive ordering”5 to accurately estimate how much of their food will be sold each day, thereby reducing costs and food waste, while retailers using Shelf Engine’s6 predictive ordering algorithm are cutting the food they discard by half. Consumers are regularly using AI baked into their apps to track their diet and exercise and improve their health. From FitGenie’s nutrition coaching to Calorie Mama’s visual food journal, these new tools are making it easier for people from all backgrounds to plan meals and maintain healthy eating habits.

Many of these food tech ventures are also fostering collaboration and improving the well-being of communities. The Flint Eats app uses crowdsourcing and GPS to help Flint, Michigan area residents pinpoint what’s convenient, fresh, and nutritious in their neck of the woods—a critical service for those in a food insecure community. In Oakland, California, The Town Kitchen is building on the meal delivery service model by training underserved youth to prepare and deliver locally-sourced lunches ordered online.


Amanda Ramcharan

“If you’re a farmer, soil is like your bank account... We rely on soil so much, and we don’t realize it.”

As a research scientist who studies the intersection of AI and agriculture, Ramcharan has been integral to the evolution of soil mapping, which is needed to visualize where and how to grow plants. Using machine learning tied to historic data, she builds models of soil characteristics that affect what can be grown and where. For example, she built a model to predict bulk density—or the mass of soil in a given volume—over the entire continental U.S.

Ramcharan wasn’t initially interested in the day-to-day challenges of farming. Born in Trinidad and Tobago, she first studied mechanical engineering at Princeton with a focus on sustainable energy. “I really got into reading about the environmental impact of fossil fuels and human activity on the globe,” she says. After graduating, she worked in Kenya, building rainwater harvesting systems for a rural community. “One of the things we talked about was, what does sustainable agriculture actually mean?” she explains. “How can you measure that?” She began using data modeling to look at different scenarios in soil and weather patterns in order to establish baselines of agricultural sustainability.

Seeing the science and the math behind these complex environmental problems really opened her eyes to how soils affect food and energy systems. She began researching the history of soil mapping in the United States, which has its roots in the era after the dust bowl. After catastrophic crop loss, the government embarked on an ambitious plan to reduce erosion by looking at how factors in the environment, like the slope of the land, depth of the soil, amount of rocks, and even native vegetation, affect crop health.

Soil mapping has grown a lot since then. Today, machine learning models increase the value of the information that humans venturing out into the fields collect. Hyperspectral imaging in the visual and near-infrared spectrum, combined with these models, can predict the organic matter in soil at deeper depths. Plants are also benefiting from machine learning. Multispectral cameras attached to drones collect data that can be used to look at a field’s photosynthetic capacity and predict future health. Machine learning can also be fed geolocal data about past weather conditions to divine upcoming precipitation events.

For researchers and farmers concerned about the viability of their yields, these innovations could be game changers. “You can have apps that aggregate all this data and answer big questions that we haven’t been able to answer before,” says Ramcharan. Through her work, Ramcharan has become an unexpected expert in the rules surrounding the use of drones, in particular. “I was kind of blown away when I realized how demarcated our airspace is,” she says.

Technologies like plant-spotting drones could be marshaled to protect soil—and our society—from the effects of climate change while simultaneously helping farmers grow more resilient crops, creating a win-win scenario. “Farmers are most concerned with their return on investment,” adds Ramcharan. “They’re always doing the math in their heads.”

From soil to supper, this report explores how technology, including AI and machine learning, could accelerate the creation of a more sustainable, scalable, and equitable food system. Leveraging technology to improve our food system will require working across sectors, finding pathways to both harness and protect data, and devising new policies and workforce training. But tackling these issues could ensure that healthier food makes its way down our entire food chain and reaches more of the folks who need it most.

  1. Campolo, Sanfilippo, Whittaker, Crawford, “AI Now 2017 Report,” AI Now Institute, 2017.
  2. Philip Napoli, “The Algorithm as Institution,” Fordham University Schools of Business Research Paper, May 5, 2013.
  3. Pedro Domingos, “A few useful things to know about machine learning,” Communications of the ACM, volume 55.10, October 2012.
  4. Mitch Hunter, “We don’t need to double world food production by 2050 – here’s why,”, March 8, 2017.
  5. Jeff Wells, “How can predictive ordering impact fresh departments?”, February 13, 2017.
  6. Catherine  Shu, “Shelf Engine  uses machine learning  to stop food waste from  eating into store margins,”, August 15, 2018.
  7. SAS, “Big data analytics: What it is and why it matters,”
  8. National Academies of Sciences, Engineering, and Medicine, “Overview of data science methods,” Strengthening Data Science Methods for Department of Defense Personnel and Readiness, National Academies Press, 2017.

Artificial Intelligence

A catch-all term used to refer to automated systems that can make decisions or simulate a form of reasoning that resembles that of humans. In non-technical settings, AI is often used to describe or categorize a broad range of technologies that span speech recognition, language translation, image recognition, predictions and determinations, tasks and functions that, until recently, have not been mechanized to the extent that they are now.1

Machine Learning

An AI system that makes predictions based on instruction from a human teacher. The system is built such that it can automatically modify its knowledge and procedures to improve its predictions based on new data, without being explicitly programmed to do so. Machine learning is used in web searches, ad placements, credit scoring, fraud detection, stock trading, and many other applications.3

Data Analytics

Business analytics software developer SAS explains that “big data analytics examines large amounts of data to uncover hidden patterns, correlations, and other insights.” Today’s technology can automatically analyze large volumes of data and draw conclusions from it almost instantaneously.7


A set of rules computer programs follow in problem-solving operations. Algorithms are used in AI systems in order to rank websites, make recommendations, and predict outcomes. Public policy scholar Philip Napoli calls them “institutions” because of their power to govern behaviors (social, political, economic, etc.), influence user preferences, and manipulate consumer decisions.2

Predictive Analytics

The National Academies of Science describes predictive analytics as “the extensive use of data and mathematical techniques to uncover explanatory and predictive models for an organization’s [or farm’s] performance as represented by the inherent relationship between data inputs and outputs/outcomes.” By employing statistical analysis techniques to mine current and historical data for insights, this process yields predictions about what will happen in the future.8