Deterministic AI Agents for Downstream Tasks and Application to Hurricane Forecasting

Explore how deterministic AI agents can revolutionize Earth science research and operational forecasting in this 53-minute conference talk from IPAM's Mathematics and Machine Learning for Earth System Simulation Workshop. Discover the current state of agentic AI technology and learn how Large Language Models (LLMs) are transforming complex knowledge work tasks that traditionally require hours of human effort into processes completed in minutes. Examine the specific application of AI agents to operational tropical cyclone forecasting, focusing on methods to ensure deterministic output from LLMs to meet operational requirements. Investigate a comprehensive 20-year archive of unstructured operational tropical cyclone discussions and forecasts, and understand how deterministic AI agents can enhance and accelerate critical human forecaster tasks including search, interpretation, and prediction capabilities. Gain insights into the potential role of LLM-based AI in Earth science research and forecasting applications, with practical demonstrations of how these technologies can be implemented in real-world atmospheric and oceanic science scenarios.