The company's model has been compared to the European Centre for Medium-Range Weather Forecasts' (ECMWF) deterministic HRES model, which is a single realization of plausible weather. However, Silurian AI's model is currently at the top of the WeatherBench leaderboard and incorporates a lot of physics. The company is also looking to expand its technology to other industries that are highly dependent on weather, such as the energy grid, agriculture, logistics, and defense.
Key takeaways:
- Silurian AI uses machine learning to make ensemble forecasts of tropical cyclones, which is essential for useful weather forecasts of uncertain events.
- Their model, NeuralGCM, is currently at the top of the WeatherBench leaderboard and incorporates a lot of physics, proving that combining machine learning and physics models can work really well.
- They aim to branch out to industries which are highly dependent on weather, such as the energy grid, agriculture, logistics, and defense.
- Their training costs are significantly cheaper than the fixed costs of the supercomputers that government agencies require and each forecast can be generated on 1 GPU over a few minutes instead of 1 supercomputer over a few hours.