Level 2 — Elementary

A team of astronomers has used an artificial intelligence tool called RAVEN to discover more than 100 hidden planets in existing NASA data. RAVEN stands for Rapid Automated Vetting of Exoplanet candidates using Neural networks. The tool analyzed information that had already been collected by NASA's Kepler and TESS space telescopes.

Kepler and TESS search for exoplanets by watching thousands of stars at once. When a planet passes in front of its star, it causes a tiny dip in the star's brightness. Scientists call this the transit method. RAVEN was designed to detect these dips by studying light curves, which are detailed graphs of a star's brightness over time.

What makes this discovery remarkable is that human astronomers had already reviewed the same data. They missed these planets because the signals were too faint or too irregular to notice. RAVEN's neural network was able to recognize subtle patterns in the data that the human eye could not detect, proving that AI can find what people overlook.

Among the newly discovered planets, several orbit in the habitable zone of their stars, where conditions might allow liquid water to exist. Others follow unusual orbital paths that surprised researchers. This breakthrough demonstrates how artificial intelligence can accelerate the pace of scientific discovery by processing massive datasets far more efficiently than humans.

astronomers

Scientists who study stars, planets, and space.

exoplanets

Planets that orbit stars outside our solar system.

neural networks

Computer systems designed to learn and find patterns, inspired by the human brain.

brightness

How much light something gives off; how bright it is.

transit method

A way to find planets by watching for dips in a star's light.

detect

To find or notice something that is hard to see.

faint

Very weak or hard to see or hear.

subtle

Not obvious; difficult to notice.

orbital

Related to the path a planet takes around a star.

datasets

Large collections of organized information.

Level 3 — Intermediate

In a remarkable feat of computational discovery, astronomers have deployed an artificial intelligence tool called RAVEN — Rapid Automated Vetting of Exoplanet candidates using Neural networks — to uncover more than 100 previously hidden planets lurking in existing NASA data. The tool sifted through years of observations collected by NASA's Kepler and TESS space telescopes, missions originally designed to survey the sky for exoplanets using the transit method.

The transit method works by monitoring thousands of stars simultaneously and flagging tiny dips in their brightness that occur when a planet crosses in front of its host star. These periodic dips create a signature pattern in the star's light curve, a detailed record of brightness over time. However, not every signal is straightforward. Background noise, instrument glitches, and overlapping stellar signals can obscure genuine planetary transits, making them extremely challenging for human reviewers to confirm.

RAVEN's deep learning neural network was trained on thousands of confirmed and false planetary signals, enabling it to distinguish real transits from false positives with remarkable precision. When applied to archival Kepler and TESS data, the algorithm identified over 100 credible planet candidates that had eluded human vetting teams. Many of these missed detections involved planets with faint signals, long orbital periods, or configurations that produced irregular transit patterns.

The implications of this discovery extend well beyond the planet count itself. Several of the newly confirmed worlds reside in the habitable zone of their respective stars, where surface temperatures could theoretically permit liquid water. Others occupy eccentric orbits or exist in multi-planet systems with fascinating gravitational dynamics. This achievement underscores a broader trend in modern astrophysics: artificial intelligence is no longer merely an assistant to human researchers but an indispensable tool capable of extracting insights from data at a scale and depth that human analysis alone cannot match.

computational

Related to using computers to process information or solve problems.

deployed

Put into active use for a particular purpose.

sifted

Carefully examined large amounts of information to find what is useful or important.

survey

To systematically observe and record data about a large area or group.

simultaneously

At the same time; happening together.

periodic

Happening at regular intervals of time.

obscure

To make something difficult to see, understand, or detect.

false positives

Results that incorrectly indicate something is present when it is not.

archival

Relating to stored records or data from the past.

eluded

Escaped detection or understanding; avoided being found.

eccentric

Having an oval-shaped or unusual orbit rather than a circular one.

indispensable

Absolutely necessary; impossible to do without.

Level 4 — Advanced

In what may constitute one of the most consequential applications of machine learning in observational astronomy, a research team has leveraged an artificial intelligence framework called RAVEN — Rapid Automated Vetting of Exoplanet candidates using Neural networks — to unearth more than 100 previously undetected exoplanets concealed within archival data from NASA's Kepler and TESS missions. These space telescopes, which have collectively monitored the luminosity of millions of stars over the past decade, employ the transit photometry method: detecting the infinitesimal dimming that occurs when an orbiting planet traverses the face of its host star.

Despite the extraordinary volume of data amassed by these missions, the process of vetting potential planetary candidates has long been a bottleneck. Each transit signal must be scrutinized against a litany of astrophysical false positives — eclipsing binary stars, instrumental artifacts, and systematic noise — before a candidate can be elevated to confirmed planet status. Human reviewers, constrained by cognitive limitations and the sheer magnitude of the data, inevitably overlook signals that fall below their perceptual threshold, particularly those with low signal-to-noise ratios or anomalous periodicity.

RAVEN's deep learning architecture, a convolutional neural network meticulously trained on thousands of labeled transit signals, circumvents these human limitations. By ingesting raw light curve data and autonomously extracting hierarchical features — from broad flux variations down to microscopic transit signatures — the algorithm can disambiguate genuine planetary transits from spurious signals with a classification accuracy that surpasses traditional vetting pipelines. When retrospectively applied to the Kepler and TESS archives, RAVEN flagged over 100 viable exoplanet candidates that had slipped through every preceding round of human and algorithmic review.

The ramifications of this discovery are profound and multifaceted. Among the newly catalogued worlds, a notable subset resides within the circumstellar habitable zone, where equilibrium temperatures could sustain liquid water on a rocky surface. Others exhibit highly eccentric orbits or participate in resonant multi-planet configurations that offer invaluable laboratories for testing theories of planetary formation and gravitational dynamics. More broadly, RAVEN's success epitomizes a paradigm shift in astrophysics: the transition from hypothesis-driven inquiry to data-driven exploration, where machine intelligence systematically interrogates datasets of unprecedented scale to surface discoveries that would otherwise remain perpetually obscured from human cognition.

consequential

Having significant and far-reaching importance or effects.

leveraged

Used something to maximum advantage to achieve a result.

luminosity

The total amount of light emitted by a star or celestial object.

infinitesimal

Extremely small; immeasurably tiny.

amassed

Gathered or accumulated in large quantities over time.

bottleneck

A point of congestion that slows down an entire process.

scrutinized

Examined very carefully and thoroughly.

litany

A long series or list, often implying tedium or repetitiveness.

perceptual threshold

The minimum level at which a stimulus can be detected by human senses or cognition.

anomalous

Deviating from what is standard, normal, or expected.

circumvents

Finds a way around an obstacle or limitation.

disambiguate

To remove uncertainty or confusion by making something clear.

spurious

False or fake; not genuine despite appearing to be so.

circumstellar

Surrounding or existing around a star.

paradigm shift

A fundamental change in the underlying assumptions or methods of a discipline.