AI Spots 10,000 New Planet Candidates in Single Survey

The search for planets beyond our solar system just took a massive leap forward, potentially tripling the number of known alien worlds in a single breakthrough.

Researchers analyzed 83 million stars using a new machine learning algorithm and uncovered 11,554 planet candidates, with more than 10,000 never spotted before. If confirmed, this would push the total number of exoplanets from about 6,000 to nearly 18,000.

The team used data from NASA's Transiting Exoplanet Survey Satellite (TESS), a car-sized telescope orbiting Earth since 2018. TESS captures wide-field photos of distant stars, watching for tiny dips in brightness that signal a planet passing in front of its sun.

Most researchers focus on the brightest stars in TESS data because they're easier to study. But this team looked at every star, including ones 16 magnitudes dimmer than the usual threshold.

That's where the machine learning algorithm became essential. A computer program learned to spot incredibly subtle patterns in the light from faint stars that would be impossible for humans to detect manually, let alone analyze across 80 million stars.

To test their model, the researchers aimed Chile's Magellan telescope at one predicted location and confirmed a "hot Jupiter" planet exactly where the algorithm said it would be. The planet, dubbed TIC 183374187 b, orbits a star nearly 4,000 light-years from Earth.

The Ripple Effect

This discovery shows how artificial intelligence can expand our view of the universe in ways previously unimaginable. About 87% of the candidates were spotted transiting their stars at least twice, giving scientists valuable data about their orbital periods.

The confirmed planets have tight orbits ranging from half a day to 27 days, meaning they're likely too close to their stars to support life as we know it. Still, the sheer number of new worlds opens countless opportunities for future study.

While TESS has already discovered 882 confirmed exoplanets since 2018, this machine learning approach unlocks access to millions of stars that would otherwise remain unexplored. Each candidate must still be verified through independent surveys, a process that could take months or years.

The researchers published their findings in a preprint study on April 20, though it hasn't been peer-reviewed yet. The work demonstrates that combining massive datasets with smart algorithms can reveal cosmic treasures hiding in plain sight.

Our cosmic neighborhood just got a whole lot bigger.