filmov
tv
How @Google uses #artificialintelligence for weather and #climate modeling with NeuralGCM
Показать описание
Modeling weather and #climatechange is one of the toughest challenges in the field of #machinelearning . Meet NeuralGCM — a hybrid machine learning and #physics model of the Earth’s atmosphere, featured in @NatureVideoChannel , that delivers state-of-the-art global weather forecasts and climate simulations.
Our guest, Stephan Hoyer, a Senior Staff Software Engineer at Google Research, introduces NeuralGCM to the @BuzzRobot community and shares the technical intricacies behind the project.
In this lecture, Stephan also offers his broader perspective on the rapidly growing field of AI weather and climate modeling
#artificialintelligence #ai #deeplearning #climatechange #weather #weatherforecast #technology #techtalk #techtalks #science #scienceandtechnology
Timestamps:
0:00 Introduction
0:12 Historical retrospective
1:46 The impact of AI on weather forecasting
2:56 How hybrid machine learning and physics models can help with weather and climate modeling
5:26 Introducing NeuralGCM, a hybrid machine learning and physics model
6:15 Exploring the architecture of NeuralGCM
7:22 Previous attempts to build hybrid models for weather and climate forecasting and why JAX, a machine learning framework for transforming numerical functions, is important
9:21 What makes NeuralGCM different from other hybrid models?
11:01 Advantages of NeuralGCM modeling
17:11 Comparison of NeuralGCM with other models in terms of runtime
17:41 Summary of the talk
19:20 Q&A
Social Links:
Our guest, Stephan Hoyer, a Senior Staff Software Engineer at Google Research, introduces NeuralGCM to the @BuzzRobot community and shares the technical intricacies behind the project.
In this lecture, Stephan also offers his broader perspective on the rapidly growing field of AI weather and climate modeling
#artificialintelligence #ai #deeplearning #climatechange #weather #weatherforecast #technology #techtalk #techtalks #science #scienceandtechnology
Timestamps:
0:00 Introduction
0:12 Historical retrospective
1:46 The impact of AI on weather forecasting
2:56 How hybrid machine learning and physics models can help with weather and climate modeling
5:26 Introducing NeuralGCM, a hybrid machine learning and physics model
6:15 Exploring the architecture of NeuralGCM
7:22 Previous attempts to build hybrid models for weather and climate forecasting and why JAX, a machine learning framework for transforming numerical functions, is important
9:21 What makes NeuralGCM different from other hybrid models?
11:01 Advantages of NeuralGCM modeling
17:11 Comparison of NeuralGCM with other models in terms of runtime
17:41 Summary of the talk
19:20 Q&A
Social Links:
Комментарии