AI used to hunt dinosaur bones

For all the thrills involved in uncovering the remnants of a creature which lived and died millions of years ago, it certainly sounds like a lot of hard work.

This is one of the, admittedly manifold, reasons we chose not to pursue our dreams of becoming professional dinosaur hunters past the age of five.  Traipsing around in the wilting heat of the desert looking for shards of dino vertebrae sounds like something that needs the dedication to fossil hunting that we just couldn’t muster.

But the armchair dinosaur hunter may now be able to locate skeletons of long-dead beasts without having to wander through arid landscapes for months on end.

A new artificial intelligence software system has been devised, writes Nature, which will take out much of the luck which is usually needed in finding specimens.

“The role of luck in vertebrate palaeontology is legendary,” said Roger Anemone, one of the palaeontologist involved in the software development.

“People tell you, ‘I was out taking a wazz and trickled on a fossil’. Everybody recognises that it’s kind of a crapshoot.”

But he reckons that a neural computer network will be able to work out more accurately where a fossil might be lying.

Currently the most cutting edge methods for dino-hunting, using satellite images, are essentially the same as the ones used by researchers a hundred years ago.  This involves examining where others have already made finds and looking for rocks of a particular age that might hold some ancient bones, before resorting back to scouring the floor with their eyes.

However, Anemone reckons that using a neural network could help to intelligently locate fossils.  He took some satellite images of an area of a study, assigning pixels in six bands of light wavelength to different terrain types and marking whether each pixel represented a fossil site.

The network was soon able to work out which sites were likely to hold dead dinosaurs, identifying 79 percent of general areas where finds had been made.  Of the total pixels that it highlighted, 99 percent had fossils in them.

The system has even been able to identify sites before a dig has begun, with a member of Anemone’s team waiting until after a discovery was made before revealing the computer had also been on the same track.

Whether the traditional detective method of following a hunch is about to made obsolete by the computer system isn’t clear, but it’s hoped that it will at least give some good hints of where to start looking.

The system is being used right now, searching for caves which might contain the earliest human remains in the ‘Cradle of Humankind’ in South Africa.