NASA's AI Tool Finds 7,000 New Planet Candidates

Scientists just got a powerful new ally in the search for worlds beyond Earth, and it's available to anyone with a computer.

NASA researchers at Ames Research Center in California have released ExoMiner++, an upgraded artificial intelligence tool that scans telescope data to find planets orbiting distant stars. On its first run analyzing data from NASA's TESS satellite, the software flagged 7,000 signals as likely planets.

The tool builds on its predecessor, ExoMiner, which validated 370 new exoplanets from older Kepler mission data in 2021. ExoMiner++ represents a major upgrade because it learned from both the retired Kepler telescope and the currently operating TESS satellite, despite their very different observing strategies.

"When you have hundreds of thousands of signals, like in this case, it's the ideal place to deploy these deep learning technologies," said Miguel Martinho, co-investigator for the project. The AI sifts through tiny dips in starlight that might indicate a planet passing in front of its host star, separating real planets from lookalike phenomena such as eclipsing binary stars.

What makes this development particularly exciting is its accessibility. NASA released ExoMiner++ as free, open-source software on GitHub, meaning any researcher anywhere can download it and start hunting for planets in the growing archive of public telescope data.

"Open-source software like ExoMiner accelerates scientific discovery," said Kevin Murphy, NASA's chief science data officer. "When researchers freely share the tools they've developed, it lets others replicate the results and dig deeper into the data."

More than 6,000 exoplanets have been discovered so far, with over half found in data from Kepler and TESS. But the enormous treasure trove of observations from both missions still contains many undiscovered worlds waiting to be found.

The Ripple Effect

Democratizing planet hunting could transform the pace of discovery. Instead of requiring specialized expertise or institutional resources, anyone with curiosity and a computer can now contribute to one of humanity's most profound questions: what other worlds exist in our galaxy?

The team is already working on the next version, which will identify planet signals directly from raw data rather than requiring a pre-screened list. "With not many resources, we can make a lot of returns," said project lead Hamed Valizadegan.

Every new planet discovered brings us closer to understanding how common Earth-like worlds might be, and perhaps one day, whether we're alone in the universe.