NASA & IBM Built An AI To Find Moon Water So We Don't Die Of Thirst Up There
IBM and NASA have teamed up to launch the NASA-IBM Lunar Foundation Model, an open-source AI built to analyze petabytes of observations collected over decades of moon missions. Hosted publicly on Hugging Face alongside its GitHub codebase, the model gives researchers a unified tool to process complex, multi-sensor datasets that previously had to be sifted through manually or fed into single-use algorithms using high-performance computing resources.
"NASA has spent decades building an extraordinary scientific record of the Moon, but collecting data is only part of the job," said Kevin Murphy, chief science data officer and acting chief data and AI officer at NASA Headquarters in Washington. "We also have to make data easier for scientists to explore and use. The NASA-IBM Lunar Foundation Model shows what's possible when we bring AI to NASA’s petabytes of scientific data. That’s a real opportunity we see with AI: turning large-scale data into new discoveries."
This also ties into the race to establish a permanent foothold on the moon, and water plays a big role in that. Beyond keeping future crews alive with potential drinking water, frozen subsurface reserves on the moon can be split into hydrogen and oxygen to manufacture fresh breathing air and rocket propellant directly on-site.
To locate these vital resources, IBM notes that the foundation model is specifically designed to scour polar craters at both the north and south poles for ice deposits. Because these permanently shadowed regions haven't seen a ray of sunlight in billions of years, hunting down surface and subsurface ice has historically been an uphill battle for researchers. By blending multi-spectral maps and high-resolution telemetry, the AI predicts ice stability inside those dark environments far more accurately than older, single-use models.
Beyond the hunt for water sources, the AI model will also be used as a mapping tool. NASA points out that since it was already pre-trained on a dataset collected by NASA's Lunar Reconnaissance Orbiter (LRO) over the past 17 years, scientists are able to adapt to the model to numerous lunar research tasks, such as spotting young volcanic features, and showcasing how deep learning AI methods continue to extract discoveries from decades of collected space data.
"For researchers studying lunar volcanism, the NASA-IBM model accelerates the identification of unusual looking volcanic features known as irregular mare patches. Because these structures appear relatively young, they challenge established timelines for lunar cooling, and mapping them could help scientists piece together a more accurate understanding of the Moon's thermal evolution," NASA says.
The model's underlying architecture is publicly accessible right now on Hugging Face and GitHub, inviting independent researchers, students, and aerospace developers to pull the code and test it against their own datasets.
