Bill Foster Questions AI CEO on Supercomputing for Weather Modeling

The Role of AI in Weather Prediction

• 💡 Bill Foster, an AI programmer, highlights the critical role of weather prediction systems, referencing a tornado in his hometown that was mitigated by prompt warnings.
• 🎯 He contrasts two approaches to weather modeling: one relying solely on vast amounts of sensor data, and a hybrid approach combining data with known physics.
• 🧠 Foster notes that in fields like protein folding, hybrid approaches using AI to supervise physics-based models have been most effective.

AI, Data Width, and Supercomputing Investment

• ⚡ The discussion shifts to the impact of AI on supercomputing, specifically the trend towards smaller data widths (e.g., single-bit) in AI machines, contrasting with traditional high-precision (32/64-bit floating point) methods.
• ⚠️ Foster questions whether government investment in supercomputing should shift towards these narrower data path architectures for weather applications, or if high-precision computing will always be necessary.
• 📊 The AI CEO suggests that while high-precision is beneficial for physics-based models, AI/ML approaches can achieve results more cheaply with lower precision, making lower-width data paths attractive for commercial AI applications.

Future of Operational Weather Systems

• 🚀 The CEO indicates that AI methods are likely to become the future of operational weather systems, as supported by previous chairmen of ECMWF (European Centre for Medium-Range Weather Forecasts).
• 🔬 However, he emphasizes that physics still plays a crucial role in understanding model behavior, guiding necessary observations, and improving overall scientific understanding.
• 🤝 It is anticipated that physics-based and AI-driven approaches will coexist and remain parallel for a significant period, with AI models already providing operational necessities.
• 💬 Foster expresses a sense of awe at the rapid advancements, suggesting that traditional physics-based understanding might become less relevant in favor of AI-driven answers, calling it a "brave new world."