NASA partners with IBM to Create new AI foundation model for weather and climate.
The US space agency NASA has teamed up with technology giant IBM in a bid to develop a revolutionary artificial intelligence foundation model that could transform weather and climate prediction. By pooling their considerable expertise in earth science and AI, respectively, NASA and IBM aim to craft a next-generation model capable of vastly more accurate forecasts.
Currently, existing tools like GraphCast and Fourcastnet utilize AI to generate weather projections faster than traditional meteorological models. However, as IBM notes, these are merely AI emulators that replicate forecasting based on training data - they lack the depth and versatility of true foundational models.
The upcoming joint NASA-IBM model aims to far surpass current solutions. With expanded accessibility, lightning-fast inference speeds and the ability to leverage diverse data sources, it promises significant advantages over existing technology. Both organizations also seek to apply the model's forecasting prowess to other climate applications for improved predictions.
Capabilities could include deducing high-resolution details from low-res data, anticipating meteorological phenomena and detecting conditions conducive to events ranging from turbulence to wildfires. If successful, the model would mark a huge step forward in utilizing AI to understand our complex climate systems.
This latest collaboration builds on a previous NASA-IBM foundational model deployed in May 2022. That model harnesses NASA satellite imagery for geospatial intelligence on an unprecedented scale, aiding issues like tracking reforestation efforts and analyzing urban heat islands.
With their combined astronomical and AI chops, NASA and IBM aim to develop groundbreaking forecasting tools with real-world applications. Only time will tell if their ambitious partnership can create weather prediction models far more accurate than anything before. But if so, it could boost climate resiliency worldwide.