Zoom Recording
Notes DocEarth science applications of artificial intelligence and machine learning (AI/ML) have seen a flurry of interest in recent years, as models become more effective at predicting patterns and processes across multiple scales. However, despite this recent focus there still exist a number of common challenges in the development, deployment, and assessment of AI/ML projects which can hinder their usefulness in various domains. In order to fully realize the potential of AI/ML as a practical tool for approaching Earth science problems, practitioners will need to better understand and address these common challenges in a standard, cross-domain way.
This session, organized as part of the ESIP Machine Learning cluster’s ongoing Practical AI initiative, will bring together AI/ML practitioners and users to talk about the generation, use, and understanding of AI/ML systems in the Earth sciences. Talks will focus on practical, successful, and useful applied AI/ML systems, and the approaches taken to overcome the common challenges inherent in producing AI/ML solutions. The session will additionally inform the ongoing Machine Learning cluster white paper, Practical AI for Geospatial Data-driven Applied Sciences, by highlighting the commonalities between successful practical AI initiatives and the “gaps” still to be solved in years to come. Here is the agenda:
- Amruta Kale, Marshall Ma - Explainable AI and Provenance in Earth AI Applications
- Michael Mahoney - AI Use Case on Tree Quantification
- Doug Newman - AI and NASA Systems
- Chung Nga - Cloud-based Data Match-Up Service and AI in Oceanography
- Ziheng Sun - Geoweaver for Productivity and Reusability of AI for Earth scientific workflows
