Google AI (Neural Network) System Helps To Find Solar System with 8 Planets

Artist's portrayal of the Kepler 90 solar system, 2, 200 light years from Earth.

The media, as usual, is literally breathless about the discovery of the new 8-planet solar system ("Just like ours!") using the Kepler space telescope. The system, named "Kepler 90" is 2,200 light years from us and no - we have no idea what the planets really look like. That graphic you see above is just an artistic rendering. No astronomer actually knows what any of the eight planets looks like, it's basically educated guessing.

Before dousing expectations that we may find extraterrestrials on any of these new worlds, let me note why the discovery has scientific importance - namely in the way the new planets were discovered. This entailed "machine learning"  via an AI (artificial intelligence) neural network system when ordinarily human astronomers would be  "trying to drink through a fire hose" in the words of astronomer Derek Pitt.

Thus, the Google AI system has been able to search through the billions of gigabytes of data much faster than any humans or group of humans.  In the words of Pitt again, using an analogy of stars in the galaxy to sand grains:

"It's as if you'd be forced to search through all the grains of sand on a beach to find only the red sand grains."

The greatest attention has been focused on the third planet from the central star, called Kepler 90i. Below is another artist's depiction of this world, which the media is comparing to Earth on account of its third position in order from the central Sun:


But make no mistake there is really very little resemblance and no, you won't find any intelligent species or any species at all there. Why? It orbits so close to its star that the surface temperature is a ‘scorchingly hot’ 800F (426C). It orbits its own Sun once every 14 days.  That period, P = 14 days, gives the clue of why it is so infernally hot.

By  Kepler's 3rd law:

a ³ = P ²

The quantity a or semi-major axis is what you want. First one finds P in years, which turns out to be:  0.00147 yr.

Then take its cube root:  a = [P ²] ^1/3   =  [0.00147] ^1/3 

Which answer I leave for ambitious readers to finish.

The point is that the so-called "Earth position" planet Kepler 90i is simply too near the central Sun to support life. If is outside whatever "habitable zone" exists around it. 

Despite the fact I may have squelched any notions of alien life on these worlds (they are all relatively "scrunched up" close to the central star) the excitement remains in the method used for discovery which can now be applied to hundred, thousands, millions of other stars - and putative solar systems.

Kepler for its part has already spent four years scanning 150,000 stars and come up with 2,500 exoplanets - with 1,000 more expected. As usual,  the general analytical approach is to observe for minute dips in photo-intensity of the star's light curve as the putative planets pass in front of it in our line of sight, e.g.

Thus, any planet that passes in front of its Sun causes a measurable dip in the associated light intensity, betraying its presence via an occultation.  The longer  in duration the dip the larger the object, etc.

I should note here that the Kepler observation  mission  officially ended in 2013, but the spacecraft recorded so much data during its four year mission that scientists expect will be crunching the data for many years to come.

According to Christopher Shallue, senior software engineer at Google AI in Mountain View, California, who made the discovery, the algorithm was so simple that it only took two hours to train it to spot exoplanets. Tests of the neural network correctly distinguished  true planets from false positives 96 percent of the time. Google AI has promised to release all of the code so that amateurs can train computers to hunt for their own exoplanets.

Maybe one or more Brane Space followers will make a new solar system discovery using the data...who knows?