AI archeologist: tech helps scientists uncover ancient drawings

An AI model has helped scientists uncover hundreds of previously unknown giant drawings in the UNESCO World Heritage site of Nazca, potentially enhancing the discovery rate by several years.

A group of archeologists, led by Yamagata University’s Masato Sakai, utilized AI and discovered 303 new figurative geoglyphs in the Nazca valley, Peru, nearly doubling the number of known ancient drawings. The hieroglyphs were most likely created two millennia ago by people of the Nazca culture. The drawings depict anything from local animals to decapitated human heads.

The valley, located 500 meters (1,640 ft) above sea level, rarely experiences rainfall, making the area unsuitable for agriculture. The lack of human activity and relative safety from the elements preserved the drawings for countless years, with humans rediscovering the giant art in the mid-20th century.

While previous attempts focused on larger pictures, Sakai’s team found many smaller drawings, on average nine meters (30 ft) in size. The discoveries were made after scientists successfully utilized an AI model, which was fed a number of pictures of known drawings. Once the model was ready, scientists gave it high-resolution aerial photos of the valley, 3.5 times larger than Washington, DC.

“This AI model focuses on relief-type geoglyphs that are small and difficult to identify since the distribution of the large line-type figurative geoglyphs is known from previous (manual) aerial studies that do not necessarily require inspection of high-resolution images,” researchers said.

Even with the help of AI, scientists spent 2,640 hours of labor to review locations that AI deemed most likely to exhibit previously undiscovered drawings. While that might sound like a lot, scientists claim it took nearly a century to discover the first 430 figurative geoglyphs. Returning the same results without AI would have required tens of times more hours.

Scientists continue to find new venues for adapting AI models to detect patterns in vast amounts of data. For example, University of Georgia doctoral student Jason P. Terry employed AI to discover an exoplanet, a world orbiting a star 477 light years away from the Sun.