Google’s GraphCast can predict weather with ‘unprecedented accuracy’

Google DeepMind has introduced GraphCast. This AI model takes only a minute to generate precise 10-day weather forecasts. It is faster and more accurate than the European Centre for Medium-Range Weather Forecasts (ECMWF) traditional HRES system, says Google. GraphCast is trained on decades of historical weather data and uses the ECMWF’s ERA5 dataset.
It can more accurately predict weather conditions such as cyclones and extreme temperatures. Traditional weather forecasts start with physical equations on supercomputers. GraphCast, though, quickly renders its projections on a single Google Cloud TPU v4 system rather than the hours traditional supercomputers take. DeepMind has made GraphCast’s code open-source.
Source: Google

{Categories} _Category: Platforms,*ALL*{/Categories}

Exit mobile version