AI Discovers 100+ New Planets in NASA Data

Scientists at the University of Warwick just taught an AI to find planets nobody knew existed, and it's already uncovered more than 100 new worlds circling distant stars.

The team built a system called RAVEN to analyze data from NASA's planet-hunting telescope, TESS. After combing through observations of 2.2 million stars, the AI confirmed 118 planets, including 31 brand-new discoveries that had been hiding in plain sight.

Some of these worlds are truly extreme. A handful are "ultra-short-period" planets that whip around their stars in less than 24 hours, making Earth's year look leisurely by comparison. Others live in what astronomers call the "Neptunian desert," a mysterious zone where planets aren't supposed to exist but somehow do.

Finding planets used to be painstaking work. When a planet crosses in front of its star, the starlight dims slightly, but so many other things can cause that same flicker. Eclipsing binary stars, instrument glitches, and cosmic noise all create false alarms that researchers had to sort through by hand.

RAVEN changes that game entirely. The team trained it on hundreds of thousands of simulated planets and cosmic mimics, teaching it to spot the difference automatically. The AI handles everything from detecting the initial signal to statistically validating whether it's really a planet, cutting months of work into moments.

The discoveries tell us something profound about our cosmic neighborhood. About 9 to 10 percent of Sun-like stars host a close-in planet, matching earlier findings but with ten times more certainty. Those rare desert planets? They appear around just 0.08 percent of stars, giving us the first precise measurement of just how empty that region really is.

Why This Inspires

This isn't just about finding new dots on a map. Every confirmed planet helps scientists understand how solar systems form, why some planets survive in extreme conditions, and where we might eventually find signs of life.

The Warwick team released their tools and catalogs publicly so astronomers worldwide can identify the most promising targets for deeper study. Future missions like Europe's PLATO telescope will follow up on these discoveries, potentially revealing even more about these distant worlds.

Machine learning is opening doors astronomy never had keys for, turning oceans of raw data into reliable knowledge at a pace humans alone never could.

We're not just finding more planets anymore—we're teaching computers to help us understand the universe itself.