Recently, I’ve heard podcast, 科技浪Tech.wav, has been discussing how DeepMind uses TPUs to achieve more accurate and energy-efficient weather forecasting compared to supercomputers using NWP (Numerical Weather Prediction). Traditional weather forecasting relies on various formulas involving atmospheric dynamics, thermodynamics, and more to predict the future.

This development is analogous to actuarial science, where actuaries use population models, economic models, and technological parameters to estimate values. Just as AI models are now outperforming traditional methods in weather prediction, it suggests that actuarial science may also follow a similar path in the future, potentially reducing the need for actuaries.