NASA's Kepler AI Find Adds Eighth Planet to Kepler‑90 System

AI‑Powered Planet Hunt Hits a New Milestone

NASA is set to broadcast a Thursday press conference that will unveil a fresh planet in the Kepler‑90 system. The discovery comes from a Google‑built artificial‑intelligence model that sifted through four years of Kepler data to spot a signal humans missed.

The agency confirmed that Paul Hertz, director of NASA’s astrophysics division, a senior Google software engineer, and two other scientists will field questions at 1 p.m. Eastern (6 p.m. UK). The announcement follows a leak that the breakthrough was made with Google AI, a claim the agency repeated in a brief teaser.

The Kepler‑90 System Gets an Extra World

Kepler‑90 already matched our own solar system with eight known planets. The new addition is a scorching world that circles its star every 14.4 days. Its orbit places it among the inner, rocky planets, mirroring the layout of Earth’s neighbourhood where small bodies hug the star and gas giants sit farther out.

Astronomers first catalogued the Kepler‑90 family using the Kepler space telescope, which stared at 150,000 stars between 2009 and 2013. The telescope logged tiny dips in starlight whenever a planet transited, a method that yields planet size and orbital period. The new planet’s signal was too faint for traditional pipelines, but the AI flagged it as a candidate.

How Google’s Model Found What Humans Overlooked

The AI was trained on the full Kepler archive, learning to recognize the subtle dimming patterns that betray a planet. Instead of scanning each star manually, the model combed through the entire dataset in parallel, flagging anomalies that fell below human detection thresholds.

Andrew Vanderburg, a NASA Sagan postdoctoral fellow, noted that the algorithm “found new planets within existing systems by spotting signals that seemed to indicate something of interest but were too weak to have been spotted by humans.” His comment underscores a broader trend: the volume of astronomical data now exceeds what any team can manually vet.

The Bigger Picture: AI as a New Telescope Lens

NASA’s reliance on Google AI signals a shift in how the agency extracts science from its legacy missions. The Kepler archive alone contains billions of photometric measurements; without machine‑learning shortcuts, many hidden worlds may stay buried.

Other groups are racing to embed AI into upcoming missions like the Nancy Grace Roman Space Telescope. If Google’s model can unearth an eighth planet in a well‑studied system, the payoff for future surveys could be exponential. The competition isn’t just academic—private firms are also building proprietary pipelines that promise faster, more precise detections.

What to Watch

The Thursday livestream will reveal the planet’s official designation, its estimated radius, and any hints about atmospheric composition. Follow the NASA webcast and the accompanying liveblog for the exact data release. After the press conference, watch for a pre‑print paper detailing the AI methodology; its citation count will indicate how quickly the community adopts the approach. Finally, keep an eye on the next scheduled data‑release window for the Roman Telescope, where a similar AI pipeline is slated to run.