How AI Can Help Fight Poverty

Machine learning (ML) and artificial intelligence (AI) are among the new technologies leaders are relying on to help alleviate a global food crisis.

By Lisa Rabasca Roepe, Contributor

The world population is expected to hit 9.6 billion by 2050, according to the United Nations, and experts warn that if scientists don’t find more efficient ways to use and protect limited agricultural resources such as land, water, and energy, there could be a global food crisis. It’s for this reason, eliminating world hunger is a top priority among the UN’s Sustainable Development Goals, which the supranational organization hopes to attain by 2030.

Currently, one in nine people, or 795 million people, do not have enough food to lead a healthy, active life. For the third year in a row, world hunger has increased — in 2017, around 821 million people faced undernourishment from chronic food deprivation, the Food and Agriculture Organization of the United Nations reported.

As famine continues to increase to levels we haven’t seen since a decade ago, it’s becoming more critical to examine the way emerging technologies can curb this reoccurring problem. “We’re in the midst of biggest [societal] shift since the Iron Age,” said Elisabeth Mason, founding director at Stanford University’s Poverty & Technology Lab. “How are we leveraging new technologies and skills to address these issues?”

Machine learning (ML) and artificial intelligence (AI) are among the new technologies leaders are relying on to help alleviate a global food crisis. Already, ML and AI technologies are predicting impoverished regions across the globe, using the data to find solutions to mitigate global hunger.

At Stanford, researchers are using these technologies to help humanitarian organizations measure the impact of their efforts, while Carnegie Mellon is tapping their potential to improve crops on a global scale. Here’s a look at a few of these solutions in action.

An Abundant Food Source

Researchers at Carnegie Mellon are working with U.S. farmers to grow more high-value crops, such as grapes and apples, using machine learning, robots, and drones. The scientists plan to apply what they learn about fruit to breed staple crops. For example, “Any information we learn about growing high-value crops can be applied to growing sorghum,” explained George A. Kantor, a senior systems scientist at Carnegie Mellon. Sorghum is common in the Sub-Saharan African and Asian regions and can withstand extreme heat and drought.

So how, exactly, does the technology work? The program uses a robot, sensors, and a high-quality camera to take photos of sorghum’s grain head. On the back-end, AI technology looks at the photos and extracts information, such as the size of the grain head and the number and size of the seeds, then estimates the quality and ripeness of the crop.

The process allows crop breeders to compare over 1,000 varieties of sorghum and make better decisions about planting, cultivating, and harvesting. The ultimate goal, as the university’s site states, is to help farmers develop “plants that produce more food on fewer acres with less water.”

“We use robots and sensors and AI to improve the breeding process, and as a result the breeders end up with a new variety of sorghum that is higher yielding,” Kantor explained. Eventually, instead of giving farmers in Africa and India advanced technology like robots and sensors, Carnegie Mellon researchers hope to hand these new seeds directly to local farmers to produce higher yielding crops. Rather than having to learn all new equipment, farmers would soon be able to plant seeds that just perform better.

Carnegie Mellon has also partnered with Clemson University to analyze plant growth. Plant breeders see potential with sorghum because, like corn, sorghum can be used as a source of grain for humans and livestock. However, the process to yield stronger sorghum is not instantaneous. “Plant breeding is slow and it takes multiple years to get a new product,” Kantor said.

Predicting Areas of Poverty

At Stanford University, researchers from the Sustainability and Artificial Intelligence Lab are using machine learning and remote-sensing data to predict crop yields, specifically, as it relates to soybeans.

“If we have a model that works for U.S. soybeans, maybe we can train that model for areas with less data,” said Marshall Burke, an assistant professor of earth system science at Stanford and a fellow at the Center on Food Security and the Environment. Researchers believe understanding crop yields will help farmers around the world make better planting decisions by increasing their ability to identify low-yield regions.

Stanford scientists are also working on locating areas of poverty, which can be difficult, as accurate and reliable data from impoverished regions is often scarce. To compensate for limited data, they are using machine learning technology to extract food scarcity information from high-resolution satellite imagery.

Researchers start out with data from household surveys that report agriculture and food insecurity on the ground, then use satellite imagery to model areas of poverty. When survey data isn’t available, scientists can use the satellite imagery to predict areas of poverty, Burke explained. The ML algorithm can then use the imagery to determine whether areas have roads, farmland, and healthy vegetation.

Stanford is also working with a number of small NGOs and UN agencies, including the UN’s World Food Programme, to determine if their relief efforts are having an impact. This research could help global organizations improve their humanitarian response to food shortages and distribute resources more effectively, Burke emphasized.

Disseminating Sustainable Solutions

At research institutions like Stanford and Carnegie Mellon, the focus isn’t just trying to incubate solutions. According to Mason, it is also to build a network of practitioners across the United States, and to get technology companies to make investments in those solutions.

“AI and technology in general offer huge opportunities for us to universalize access to valuable information,” she explained, “but also to target it in a way that can be more effective in opening up new opportunities.”