An Arizona State University researcher is using artificial intelligence to transform data collected from satellites and other earth observation instruments into information, which could help decision makers become better stewards of the planet.
Hannah Kerner, assistant professor of computer science at Arizona State University, noted AI has become a hot-button issue, especially in the areas of consumer products and social media. But Kerner said there are few discussions about how AI can be applied to help combat big adversaries from climate change to poverty and food insecurity.
“AI really does have a huge potential to help us address these challenges that society is facing,” Kerner contended. “We really need to focus our research and efforts in AI towards tackling those challenges.”
Kerner and her colleagues have developed a more effective way of collecting what is known as “labeled data,” which then can be used to train AI models to better complete tasks. She acknowledged there are associated risks but encouraged people to learn about other uses of AI which could have, as she puts it, “transformative benefits.”
While AI technology is advanced and has led to some distinguished accomplishments, Kerner pointed out most people would be surprised by how poorly some AI models perform in “real-world” applications. She added algorithms can struggle to complete some tasks most folks would consider trivial, like interpreting visual data on satellite images.
“Even if you haven’t ever seen agriculture — in, say, China — in a satellite image, you would still be able to recognize that it is agriculture,” Kerner explained. “Current models fail miserably at what we call ‘geographic generalization,’ or transferring knowledge across different parts of the world.”
Kerner stressed the models need to become more effective, especially as the global climate is changing, often rapidly.
