TL;DR:
- Dr. Hannah Kerner of ASU leverages AI to address climate change and global challenges.
- ASU collects extensive data, including satellite and ocean sensor data, which can be processed with AI for practical solutions.
- Kerner leads AI initiatives for NASA Acres and NASA Harvest, focusing on agricultural and food security challenges.
- A remarkable success story involves aiding Togo during the COVID-19 crisis by creating a high-resolution cropland map.
- The challenge of limited labeled data is overcome through innovative solutions like “Street2Sat” and “few-shot learning.”
- Kerner envisions AI tools, exemplified by “Presto,” tackling various global issues at a planetary scale.
- Real-world deployment and interdisciplinary collaboration are key factors in the success of AI research.
Main AI News:
In a world grappling with climate change and other pressing global issues, the role of artificial intelligence (AI) is gaining prominence. Dr. Hannah Kerner, an assistant professor of computer science at Arizona State University (ASU), believes that AI can play a crucial role in addressing these challenges, given the wealth of data already available.
ASU has been collecting vast amounts of data, including data from orbiting satellites and ocean sensors, for decades. According to Dr. Kerner, this data can now be harnessed with the power of AI to provide actionable insights and real-world solutions. She emphasizes the potential of AI in making informed decisions that can lead us towards a more equitable and sustainable future.
Dr. Kerner’s work extends beyond theory; she is at the forefront of developing and deploying AI and machine learning tools for NASA Acres and NASA Harvest consortia. These initiatives utilize satellite observations to address critical agricultural and food security challenges, both in the United States and globally.
One notable example of the impact of Dr. Kerner’s work occurred in 2020 when the West African country of Togo faced a crisis due to the COVID-19 pandemic disrupting supply chains and agricultural activities. The Togolese government urgently needed to allocate aid funds to smallholder farmers across the country, but they lacked the necessary demographic data.
Dr. Kerner and her research group rose to the challenge by using their machine learning methods to create a high-resolution cropland classification map of Togo. This map enabled the government to efficiently distribute aid to more than 50,000 farmers, demonstrating the potential of AI in real-world problem-solving.
One common obstacle in interpreting remote-sensing data globally is the scarcity of labeled data. Creating labels for agricultural fields, for instance, is a slow and expensive process that often requires physical visits to each location. Dr. Kerner and her team devised an innovative solution called “Street2Sat,” using GoPro cameras mounted on vehicles to capture geotagged and timestamped images alongside farm fields. These images were then analyzed by computer vision algorithms to identify crops, providing valuable ground-truth data.
In addition to overcoming the challenge of limited labeled data, Dr. Kerner is pioneering “few-shot learning” techniques for satellite remote-sensing data. Unlike traditional computer vision models, which require large labeled datasets, few-shot learning aims to make efficient use of existing labeled data. Dr. Kerner is confident that the lessons learned from models like ChatGPT, which utilizes large amounts of unlabeled text data, can be applied to develop “Large Earth Models” (LEMs) for satellite data.
The ultimate goal is to create AI tools capable of addressing a wide range of planetary-scale challenges, including forest classification, deforestation monitoring, methane emission detection, weather tracking, livestock counting, and more. Dr. Kerner and her team have designed “Presto,” a pre-trained remote sensing transformer, specifically tailored to the nuances of working with remote-sensing data. This model’s versatility and efficiency have already earned it recognition and adoption by organizations such as NASA and the European Space Agency.
Dr. Kerner emphasizes that real-world deployment and interdisciplinary collaboration are essential for the success of AI research. Solutions need to be developed in tandem with end-users to ensure they meet real-world needs and deliver tangible impacts. In her view, these partnerships and user-driven metrics are the keys to advancing the field of AI and solving some of the world’s most pressing challenges.
Conclusion:
Dr. Hannah Kerner’s pioneering work with AI, combined with ASU’s data resources, demonstrates the immense potential for AI to address global challenges effectively. This signifies a growing demand for AI solutions in various industries, particularly those related to environmental and agricultural sustainability, disaster response, and resource allocation. Businesses should consider investing in AI research and development to stay competitive in an evolving market focused on solving pressing global issues.
