Two Stanford University graduates, John Dean and Kai Marshland, co-founded WindBorne Systems to address the lack of quality weather data for about 85% of the Earth and its atmosphere. The company designs small weather balloons that can fly around the world for up to 40 days, collecting atmospheric data that can be used to improve global weather forecast models. This data is particularly valuable for setting the initial conditions that these models start with.
Key takeaways:
- AI weather models are being developed that can run on a desktop computer, offering a revolutionary new way to forecast weather.
- Large language models, like ChatGPT, require large amounts of data to operate effectively, and as they run out of data, operators are turning to synthetic data to keep improving results.
- The European Centre for Medium-Range Weather Forecasts maintains a rich dataset known as ERA5, which has proven incredibly useful for training AI models to forecast weather.
- WindBorne Systems, a company co-founded by John Dean and Kai Marshland, aims to solve the problem of weather uncertainty by designing small weather balloons that can relay useful atmospheric data for up to 40 days.