Artificial Intelligence Uses NASA/JPL Data to Make Important Discovery in Far-off Solar System

New trend of analyzing old, archival data using latest AI techniques could yield many new discoveries

Published : Friday, December 15, 2017 | 6:43 AM

Kepler-90 System Compared to Our Solar System (Artist's Concept)

Image Courtesy JPL/Caltech

We are not alone.

Using data from NASA’s Kepler Space Telescope, which was developed through Pasadena’s Jet Propulsion Laboratory, scientists have discovered an eighth planet circling Kepler-90, a Sun-like star 2,545 light years from Earth. The discovery means that the first time in history, humans have now discovered a solar system with just as many planets as our own, NASA revealed Thursday.

But it’s far away, really far away. 2,545 light years away from Earth, to be exact.

The eight discovered “exo-planets,” planets outside our own solar system, orbit the the Kepler-90 star, which looks similar to the sun.

According to a NASA announcement, the newly-discovered Kepler-90i—a hot, rocky planet that orbits its star or ‘“sun,” once every 14.4 days—was found using machine learning from Google.

But, no, 90i is not another Earth. Its average surface temperature is 800 degrees fahrenheit.

Google-developed machine learning, which helped locate the far-away planet, is an approach to artificial intelligence in which computers “learn,” said JPL Thursday. In this case, computers learned to identify planets by finding in Kepler data instances where the telescope recorded changes in starlight caused by planets beyond our solar system, known as exoplanets.

According to Dr. Eric Mamajek, deputy program chief scientist for NASA’s JPL-based, Exoplanet Exploration Program, which manages the Kepler mission, the Kepler-90 system was previously known to have seven transiting planets.

“The real story,” said Mamajek in an email interview, “is that new techniques like machine learning can be applied to this archival data from the NASA Kepler exoplanet mission, and yield new discoveries.”

Paul Hertz, director of NASA’s Astrophysics Division in Washington, illustrated the significance of the discovery, saying, “Just as we expected, there are exciting discoveries lurking in our archived Kepler data, waiting for the right tool or technology to unearth them. This finding shows that our data will be a treasure trove available to innovative researchers for years to come.”

But before anyone organizes flights to the new solar system, or begins preparing for a visit from our new cosmic neighbors, Mamajek, put the discovery in perspective, explaining, “This discovery shows that new techniques like neural networks can yield the precious, often faint, signals of small Earth-size planets in large data sets taken with space telescopes. We really need as many computerized tools and automated techniques as possible to sift through the many gigabytes of telescope imagery to find these signals.

Asked if the new planet in the far-off solar system represents another Earth, Mamajek took a very long view.

“There two different aspects to that,” he began. “When you say a ‘second Earth’, I’m imagining a planet of nearly identical mass, size, and composition as Earth with similar atmosphere, surface water, and perhaps life. This is like searching for a person with almost identical properties to yourself when encountering hundreds or thousands of people. What Kepler has taught is that small rocky planets similar in size to Earth are very common in the universe. But, astronomers do not yet have a strong handle on the variety of conditions on those planets – their atmospheres, temperate temperatures, whether they have surface water, magnetic fields, plate tectonics, all the things that Earth has, and which seem to be critical to life’s development and survival on this planet.

In other words, said Mamajek, scientists can likely locate lots of planets, but not lots of Earths.

“Astronomers will probably need to discover and characterize perhaps dozens, or even hundreds, of small rocky planets around nearby stars in order to find one that is similar to Earth,” he explained. “This is a tall order, and if we are talking imaging and spectrally characterizing many planets around the nearest stars, then astronomers will need large space telescopes to be able to do this.”

Mamajek noted that JPL is currently developing technologies for coronagraphs and starshades which will enable space telescopes to reduce the bright glow of the stars in order to find the faint light from Earth-like planets.

Planets like Earth are billions of times fainter than their stars, said Mamajek, “so any telescope system flown with the purpose of discovering other Earth-like planets needs technology for blocking the bright light of their host stars.”

As to the question of whether the new machine learning telescopes can actually help find new lives in other solar systems, Mamajek was a little circumspect.

Responded Mamajek, “Machine learning techniques provide yet another tool in scientist’s belt of techniques for analyzing data from planet-hunting telescopes, and it may be helpful in classifying spectra for potentially habitable planets; in other words, looking at the chemical fingerprints of light from distant worlds which will help tell us what their atmospheres are made of and what their surfaces are like.

“We may detect the ‘biosignature’ gases of exoplanets with life, long before we hear or see “ET,” or any other life form out there,” said Mamajek, adding wisely, “Perhaps by then the machines will have learned to replace “us”, and then they’ll definitely need machine learning to find ET.”

Ames Research Center runs the Kepler mission on a day-to-day basis.