NASA, IBM Unveil Open Source Artificial Intelligence Tool to Explore the Moon

The NASA and IBM Lunar Foundation Model is the world’s first open artificial intelligence model designed specifically to study the Moon. The model, which was trained using data from three different missions — America’s LRO and GRAIL and Japan’s SELENE — can spot ice deposits on the lunar surface 23 percent more accurately and identify craters 19 percent better than a baseline Microsoft model. The model was trained using half the data required by the Microsoft model. The first time it was used, it correctly identified the crater caused by the SpaceX Falcon 9 crash that occurred on August 5th.

This open dataset of 2 million co-registered data points allows teams to tailor their own models for use with lunar data, even if they don’t have access to private repositories. Foundation models are beginning to move beyond terrestrial applications and into other fields such as planetary science and geology, as well as assisting manned missions like Artemis.