Click Here Cloudy with a chance of Algorithms
12 snips
Jan 2, 2026 Paris Perdikaris, an AI researcher and associate professor, shares insights on advancing weather forecasting with deep learning. He discusses how AI can augment meteorologists' efforts, improving predictions but not replacing human expertise. Paris highlights the importance of public data from NOAA and ECMWF in training models like Aurora, which predicts hurricanes more accurately. However, he warns that staffing cuts could jeopardize data collection, impacting AI's effectiveness. Optimism remains as new data sources could enhance future forecasts.
AI Snips
Chapters
Transcript
Episode notes
Pattern Learning Beats Equation Solving
- AI models learn weather by recognizing patterns in massive historical observations rather than solving physics equations.
- Paris Perdikaris says this makes AI faster and sometimes more accurate than traditional forecasts.
An Education For An AI Meteorologist
- Paris compares training Aurora to giving an AI a liberal arts education in Earth systems before specialized graduate work.
- He then fine-tunes the model for tasks like hurricane tracking, air quality, and ocean waves.
AI Strong On Hurricane Tracks
- Aurora outperforms operational forecasts on hurricane track predictions several days ahead.
- The system still struggles with finer-scale events like flash floods and microbursts, critics note.