Ancient Lost Messages Uncovered With New Scientific Technology

Archeologists have discovered four new massive ancient designs etched into the ground in Peru by using a new form of artificial intelligence, the Miami Herald reported.

The ancient geoglyphs were found using the power of deep learning to uncover designs that date back thousands of years at a UNESCO World Heritage Site.

According to a new study published in the Journal of Archeological Science, four new geoglyphs were found in the northern region of the Nazca desert in Peru using the new machine learning technique.

Known as “Nazca Lines,” these geoglyphs can span hundreds of feet, often depicting animals. The newly discovered carvings are a humanoid creature holding what appears to be a club, a fish, and some kind of bird, according to the archeologists. A fourth geoglyph appears to be a pair of legs and possibly hands that stretch more than 250 feet.

The first Nazca geoglyphs were discovered around 100 years ago. Since then, more have been revealed through the use of aerial photography taken from airplanes and balloons.

But because of their size, geoglyphs are difficult to find since seeing them on the ground is impossible, and scanning through aerial photographs is immensely time-consuming.

To speed up the search, researchers from Japan’s Yamagata University developed a deep-learning artificial intelligence technique that is trained to find the designs using satellite imagery.

According to the study, this kind of artificial intelligence allows for greater power and flexibility in analyzing patterns in images, language, and speech. In archeology, this machine learning can be used to analyze text, iconography, and the writing on excavated objects. 

For the researchers from Yamagata University, searching for new Nazca Lines posed some unique challenges, as the geoglyphs vary in design and the known geoglyphs used to train the technology were so few in number.

According to the researchers, the four new geoglyphs discovered using this technique were found about 21 times faster than if the images were searched by eye.