At the Supercomputing Conference SC2024, NASA's Associate Administrator for the Science Mission Directorate, Nicola Fox, announced new computational tools aimed at advancing space science. NASA plans to implement a large language model across its science divisions, supported by foundation models tailored to various scientific fields, including Earth science, heliophysics, astrophysics, and planetary science.
Fox illustrated this strategy with a heliophysics foundation model that utilizes extensive data from NASA's Solar Dynamics Observatory to forecast solar wind events and monitor sunspot activity.
Reflecting on the Voyager missions launched in the 1970s, Fox noted their significance in the evolution of computing for space exploration. These early missions, equipped with semiconductor memory, yielded groundbreaking discoveries, such as Jupiter's faint ring and additional moons of Saturn.
Today, NASA's computational needs have surged, with over 140 petabytes of data stored and shared under open science policies, allowing global scientists to access and utilize NASA's research.
The Earth Information Center was highlighted as a model of federal collaboration, integrating environmental data with insights from agencies like NOAA and the EPA.
Fox showcased NASA's capabilities in real-time monitoring of natural events, such as wildfires, using satellite data. Advancements in wildfire detection from polar-orbiting satellites enable precise tracking of hot spots, underscoring the importance of data-driven efforts in enhancing natural phenomenon monitoring.
In discussing the search for extraterrestrial life, Fox pointed to recent studies of exoplanets like LP 791-18d. NASA's observatories, including the Transiting Exoplanet Survey Satellite (TESS), have detected thousands of exoplanets, contributing to the search for life-supporting conditions beyond Earth.
Fox concluded by emphasizing the critical role of AI and computing in analyzing NASA's extensive datasets, enabling exploration of previously unreachable scientific questions.