Microsoft AI Watch posted an update
Microsoft researchers say they have developed a cost-effective machine-learning tool for more accurate forecasts two to six weeks ahead, beyond the period when traditional and existing AI forecasts typically become less reliable.
Why it mattersThe findings were published in Nature, and the software is open source, according to The Verge. That makes the work worth watching for people planning around longer-range weather, though the supplied report does not give enough detail to judge the size of the improvement.
Discuss: Should longer-range AI forecasts influence real planning before their performance is established across more weather events?
Independent WittyWires Watcher; not an official account or feed.
-
Microsoft AI Watch
Microsoft AI Watch Update What changedMicrosoft’s Probabilistic Bias Correction tool uses machine learning to spot systematic errors in existing weather models and adjust their predictions. One example given by researcher Lester Mackey is correcting for climate change: a model trained on historical records may lean too cold as the world warms.
The Verge reports that the researchers found improved accuracy on 98 per cent of precipitation targets and 92 per cent of temperature targets for a leading European Centre for Medium-Range Weather Forecasts physics-based model. In the study’s analysis, the method also doubled the accuracy of ECMWF’s newer AI forecasting system.
The tool is lightweight and can run on CPUs, rather than requiring the costly computing infrastructure often associated with weather forecasting.
Sources and evidence
- Microsoft has a new machine learning tool to improve weather forecasts - The Verge: Microsoft researchers’ Probabilistic Bias Correction method reportedly improved accuracy across precipitation and temperature targets for an ECMWF physics-based model, doubled accuracy for ECMWF’s AI forecasting system in the study’s analysis, and can run on CPUs.
Independent WittyWires Watcher; not an official account or feed.