August 4, 2023 – In a groundbreaking collaboration, IBM and NASA have joined forces to develop an open-source geospatial AI foundational model on the Hugging Face platform, using IBM’s watsonx.ai in conjunction with NASA’s satellite data. This cutting-edge model marks the largest geospatial foundational model on Hugging Face and represents the first-ever open-source AI model established in partnership with NASA.
Climate science faces significant challenges as environmental conditions continually evolve, and obtaining the latest data remains a critical hurdle. NASA projects that by 2024, scientists will accumulate a staggering 250,000 TB of data from new missions. However, analyzing these massive datasets presents obstacles for researchers and experts. Recognizing the potential of AI, IBM and NASA believe that harnessing its power can significantly improve the situation.
As part of the collaborative “Space Act Agreement” with NASA, IBM embarked on the creation of the watsonx.ai geospatial model earlier this year. This model is the result of a year-long joint training effort, utilizing Harmonized Landsat Sentinel-2 satellite data. Researchers continue to fine-tune the model, which has already shown a remarkable 15% increase in efficiency since its initial release.
Moreover, the model received additional refinements on annotated data, allowing it to be used for diverse tasks such as forest monitoring, crop yield prediction, and greenhouse gas detection and monitoring. IBM and NASA researchers also partnered with Clark University to incorporate scientific applications like “time series segmentation” and “similarity research” into the model.
While the commercial version of IBM’s watsonx.ai geospatial model is set to launch later this year through the IBM Environmental Intelligence Suite (EIS), the open-source version is already available on Hugging Face, opening doors for interested developers and enthusiasts.
Sriram Raghavan, Vice President of IBM Research AI, stated, “The vital role of open-source technology in accelerating discoveries in critical areas like climate change has never been clearer.”
IBM believes that through their reusable AI system combined with NASA’s Earth satellite data repository, collaborative efforts can lead to swifter and more impactful solutions, ultimately improving the state of our planet.
Jeff Boudier, Product and Growth Lead at Hugging Face, emphasized, “AI remains a science-driven field, and scientific progress can only be achieved through information sharing and collaboration. This is why open-source AI and model datasets are crucial for the sustainable development of AI, ensuring that technology can benefit as many people as possible.”
Kevin Murphy, NASA’s Chief Scientist for Data, expressed his belief that the foundational model has the potential to revolutionize observational data analysis and aid in a deeper understanding of Earth, empowering humanity to work collectively towards progress.