NASA Science Mission Directorate’s (SMD) Strategy for Data Management and Computing for Groundbreaking Science 2019-2024 Report identified that Artificial Intelligence (AI) and Machine Learning (ML) has yet to be fully appreciated and understood by SMD and science disciplines. Specifically, Strategy 2.4 in the report states “Invest in the tools and training necessary to enable breakthrough science through application of AI/ML” and recommends that “SMD should make investments to incentivize and educate the community on how to use AI/ML to approach science in new ways. Hands-on training can be achieved through expansion of hackathons, competitions, and grant programs. Science results and lessons learned about the use of AI/ML will be shared at community meetings to increase awareness of the potential of these techniques”.
In response, SMD established an SMD AI working group consisting of AI experts from all science divisions and high end computing program to begin working on SMD level AI/ML initiatives. The goals of this initiative are:
- Provide a holistic view of AI/ML across SMD with respect to data and computing including current landscape, workforce, policies, and challenges
- Collaboratively develop a roadmap to leverage large volumes of data and computation to accelerate AI/ML across NASA science
- Demonstrate harmonized SMD AI/ML activities to the agency’s digital transformation effort
- Identify areas of natural collaborations and develop a plan for cross-division AI/ML collaboration on data, model, and computing
- Develop a long term strategic roadmap for AI to advance NASA science
The 3rd SMD (Science Mission Directorate) and ETD (Engineering and Technology Directorate) Workshop on AI (Artificial Intelligence) and Data Science: Leaping Toward our Future Goals, was held March 21-23, 2023 at NASA’s GSFC (Goddard Space Flight Center). Access the workshop presentations and posters here.
SMD AI team has established a working group led by the NASA IMPACT project in partnership with IBM through their Space Act Agreement to explore different Natural Language Technologies (NLT) in order to design, construct, and evaluate a common Large Language Model (LLM) specifically for SMD's needs. The working group consists of NLT practitioners from each of the five science mission directorates and IBM researchers who have committed to an accelerated timeline aimed at developing a SMD LLM and an accompanying comprehensive evaluation framework, by December 2023. If you have any questions related to this working group please email Ms. Elizabeth Fancher.
NASA and IBM open-sourced the Geospatial Foundation Model and datasets for three different downstream applications through the Hugging Face. This joint effort was carried out under the Space Act Agreement between NASA and IBM. More information about the model and data can be found at https://www.earthdata.nasa.gov/news/impact-ibm-hls-foundation-model.