NASA is dramatically accelerating the hunt for worlds beyond our solar system by harnessing artificial intelligence (AI) to sift through massive astronomical datasets — revealing signals that human analysts might miss and speeding up the discovery process.

At the forefront of this effort is an AI-based tool called ExoMiner++ — an open-source deep learning system that builds on earlier AI models to comb through data from NASA’s space telescopes and flag promising new exoplanet candidates.

From Kepler to TESS: AI Tackles Huge Data Volumes

NASA’s Transiting Exoplanet Survey Satellite (TESS) and its predecessor, the Kepler mission, have collected enormous amounts of observational data by watching stars for tiny dips in brightness — a telltale sign that a planet might be passing (“transiting”) in front of a star. Manually analyzing all this data is a monumental task, but AI excels at recognizing complex patterns in large datasets.

ExoMiner — the original model introduced in 2021 — used machine learning to validate 370 previously unconfirmed exoplanets from Kepler’s trove of data. ExoMiner++ has since been retrained on both Kepler and TESS observations to broaden its reach.

In its initial run scanning TESS’s dataset, ExoMiner++ flagged about 7,000 potential exoplanet candidates. These are signals that could very well represent real planets, but they still require follow-up observations from ground-based or space telescopes to confirm them as bona fide discoveries.

How AI Improves Planet Detection

Machine learning tools like ExoMiner++ help researchers by:

• Filtering out false positives: Many astronomical signals look similar to exoplanet transits but are caused by other astrophysical phenomena. AI can distinguish real transit patterns from noise.
• Reducing workload for scientists: Instead of manually reviewing tens of thousands of light curves, astronomers can focus on the most promising candidates identified by AI.
• Speeding discoveries: The speed of AI classification means new worlds can be identified months or years sooner than through traditional analysis alone.

Scientists hope that future versions of the software will eventually identify transits directly from raw telescope data, further automating the search and freeing researchers to explore deeper questions — like which of these worlds might have conditions suitable for life.

Why This Matters

NASA’s exoplanet catalogue — now numbering over 6,000 confirmed planets — continues to grow thanks to improvements in detection methods and telescope technology. AI is one of the most powerful new tools in this toolbox because it can uncover subtle patterns that traditional algorithms miss.

As data from ongoing missions like TESS pours in, and as future missions such as the Nancy Grace Roman Space Telescope begin to deliver observations, AI-driven models are poised to play a central role in identifying ever more distant worlds.

A Future Fueled by Open Science

One exciting aspect of ExoMiner++ is that it’s open-source, meaning researchers worldwide can access and build on NASA’s AI models to hunt for exoplanets — democratizing discovery and accelerating progress across the scientific community.