In response to the escalating risks posed by climate change, IBM initiated a partnership with NASA six months ago. The goal? To devise an AI model that could expedite the analysis of satellite images and thereby invigorate scientific discovery. A secondary aspiration was to render nearly 250,000 terabytes of NASA mission data accessible to a wider audience.
Today, IBM's foundation model, now made public through Hugging Face, holds the distinction of being the largest geospatial model hosted on the platform, and NASA's first open-source AI foundation model collaboration. IBM has approximated that this AI model can analyze geospatial data at a pace four times quicker than its contemporaries, requiring only half the amount of labeled data.
A commercial version of this model will be incorporated into IBM's AI and data platform, watsonx, available via the IBM Environmental Intelligence Suite (EIS) by year's end.
The model, meticulously fine-tuned by IBM, can chart the extent of past floods and wildfires in the U.S., furnishing valuable insights into potential future areas of risk. But the potential applications extend far beyond this scope. With additional fine-tuning, the model could be deployed for a myriad of tasks, including tracking deforestation, predicting crop yields, or the detection and monitoring of greenhouse gasses.
IBM and NASA hope that open sourcing this model will spur researchers worldwide to refine and expand upon it, thus fostering the creation of other geospatial models and applications.
Dr. Rahul Ramachandran, IMPACT Manager and Senior Research Scientist at NASA Marshall, commented “AI foundation models for Earth observations present enormous potential to address intricate scientific problems and expedite the broader deployment of AI across diverse applications. We call on the Earth science and applications communities to evaluate this initial model for a variety of uses and share feedback.”
For companies like IBM, open sourcing also accelerates product development. Outside innovations to the model will inform improvements to IBM’s own commercial EIS offering.
The project coincides with NASA’s decade-long Open Source Science Initiative to build a more collaborative scientific community. It also aligns with IBM’s long commitment to open source, from supporting Red Hat OpenShift to streamlining AI workflows through the Ray and PyTorch communities.
As climate challenges grow more urgent, empowering the brightest minds across public and private sectors to derive insights is crucial. While not a silver bullet, democratizing access to advanced AI unlocks new potential in the fight against climate change.