IBM and NASA announce collaboration to research the impacts of climate change using artificial intelligence

From IBM

IBM and NASA‘s Marshall Space Flight Center, today 23, announce a collaboration to use IBM’s artificial intelligence (AI) technology to discover new pathways with a vast wealth of geospatial and earth science data from NASA. For the first time, the joint work will use AI foundational modelling technology to data from NASA’s Earth observation satellites.

>>>AI has the potential to make a huge difference in our world, but do you know how to use it?

Foundational models are types of AI that take a large unlabeled dataset and are trained to be used in different tasks and allow information about one situation to be applied to another. These models have grown rapidly in the field of natural language processing (NLP) technology over the last five years, and IBM is a pioneer in applications of foundational models beyond language.

https://www.youtube.com/watch?v=9iRULL7PA0U

Earth observations, which allow scientists to study and monitor our planet, are being collected at a frequency and volume never seen before. New approaches are needed to extract all this knowledge from this work and provide an easier way for researchers to analyze and gain insights from the vast resources of captured data to advance scientific understanding of planet Earth and answer questions rapidly. climate-related.

One project will train IBM’s foundational geospatial intelligence models on NASA’s Harmonized Landsat-Sentinel-2 (HLS) dataset, a record of Earth’s surface and land-use changes captured by Earth-orbiting satellites, analyzing the petabytes of satellite data to identify changes in the geographic presence of phenomena such as natural disasters, cyclical crop yields, and wildlife habitats. This fundamental modelling technology will help researchers provide critical analyzes of our planet’s environmental systems.

Another result of this collaboration should be a more accessible Earth science literature research corpus. IBM has developed a natural language processing model, trained on nearly 300,000 articles from scientific publications on Earth, to organize the literature and facilitate the discovery of new information. One of the largest AI projects trained with Red Hat’s OpenShift software to date, the model uses PrimeQA, a multilingual Q&A software from IBM. In addition to providing a resource for researchers, the new language model for geoscience can be inserted into NASA’s scientific data management and administration processes.

“The beauty of foundational models is that they can be used for many later applications,” said Rahul Ramachandran, a senior research scientist at NASA’s Marshall Space Flight Center in Huntsville. “Building this foundational model cannot be tackled by small teams,” he added. “You need to have teams in different organizations to bring together their different views, resources and skills.”

“Foundational models show successful natural language processing, and now it’s time to expand into new domains and modalities important to business and society,” said Raghu Ganti, one of IBM’s principal investigators on the project. “Applying foundational models to geospatial factors, a sequence of events, time series and other non-linguistic factors within geoscience data could make the knowledge and valuable information generated from these models become available to a much wider group. of researchers, companies and society at large. Ultimately, it could help more people working on some of the most pressing climate issues.”

Under this agreement, other potential projects between IBM and NASA include building a foundational model for climate predictions using MERRA2, a dataset of atmospheric observations. This collaboration is part of NASA’s Open Source Science Initiative, an initiative committed to building an open, inclusive, transparent, and collaborative science community over the next decade.