Archaeologists in Peru, assisted by artificial intelligence (AI), have uncovered 303 previously unknown giant symbols etched in the Nazca Desert. These geoglyphs include birds, plants, spiders, human figures with head coverings, severed heads, and a knife-wielding orca.
Described in a study published in the journal PNAS, this discovery nearly doubles the number of known Nazca geoglyphs, mysterious artworks formed on the ground by the movement of stones or gravel dating back around 2,000 years. The findings shed light on the enigmatic purpose of these symbols.
Located 50 kilometers inland from Peru's southern coast, the large symbols were first discovered in the early 20th century. Situated 500 meters above sea level, the geoglyphs have survived the ages due to the sparsely populated, arid desert region, which is unaffected by floods and unsuitable for crop cultivation.
The increase in new findings has been attributed to the use of high-resolution remote imaging, with an average of 19 geoglyphs discovered annually from 2000 to 2020. However, the application of AI to narrow down potential candidates has significantly accelerated the pace of discovery, promising a 'revolution in archaeological findings.'
A team led by Masato Sakai, an archaeology professor at Yamagata University in Japan, trained an AI model using high-resolution images of 430 Nazca symbols mapped since 2020. The team included researchers from IBM's Thomas J. Watson Research Center in New York.
The primary challenge was the limited number of images available for training. Typically, such models require tens of thousands of images. The team focused on the smaller, more figurative symbols, which are usually around 9 meters long and harder to identify than the larger line-type symbols, which measure 90 meters in length and are easier to spot from the air.
The AI model suggested over 47,000 possible locations in the desert covering 629 square kilometers. The team examined and ranked these suggestions, identifying 1,309 high-potential candidates. For every 36 AI suggestions, researchers found one promising candidate.
Despite its imperfections, the AI application proved to be a 'game changer' in reducing the amount of work needed to narrow down the search. According to the study, it allowed the team to focus on targeted, valuable fieldwork in Nazca Pampa.
Between September 2022 and February 2023, the team tested the accuracy of their model in the Nazca Desert, using foot surveys and drones, ultimately 'ground-truthing' 303 figurative geoglyphs.
'The well-preserved geoglyphs were immediately recognized for what they were,' Sakai stated in an email. 'For those in poor condition, we are investigating what they are through detailed fieldwork.'
Of the 303 newly discovered figurative geoglyphs, 178 were suggested by the AI model, while 125 were additional findings. Among them, 66 were found as part of a group identified by AI, while 59 were discovered during fieldwork without AI assistance.
Many more geoglyphs may be waiting to be uncovered. The team could not survey 968 of the promising candidates during the detailed fieldwork season. Given the initial success rate based on the AI model, at least 248 additional figurative geoglyphs could potentially be discovered, according to the study's estimates.
Amina Jambajantsan, a researcher at the Max Planck Institute for Geoanthropology in Germany, called it 'amazing' to nearly double the number of known geoglyphs, especially considering the limited training data. Jambajantsan, who was not involved in the Nazca research, uses an AI model to identify burial mounds in Mongolia based on satellite images.
She noted that her work often follows a similar pattern to the Nazca team, with AI-based suggestions frequently leading to additional discoveries during fieldwork. 'Artificial intelligence is amazing, but humans are still needed,' she said.
AI has the potential to make significant contributions to archaeology, although models are not yet fully accurate. 'The problem is that archaeologists don't know how to build a machine learning model, and data scientists are usually not really interested in archaeology because they can earn much more money elsewhere,' she added.
The purpose of the Nazca symbols remains unclear. The leading hypothesis is that they formed a sacred space that may have been a pilgrimage site. Other theories suggest they played roles in calendars, astronomy, irrigation, or movements such as running or dancing.
A recent analysis of the newly discovered symbols, along with those already known, revealed interesting trends. The larger, line-type geoglyphs, easily visible from above, typically depict wild animals or plants, while the smaller relief-type geoglyphs include humans or things involving humans, such as human sacrifices or domesticated llamas.
The survey also revealed a clearer network of paths in the desert, many of which were near various groups of symbols. According to the study, the smaller geoglyphs were along ancient winding paths made by the Nazca people who viewed the symbols in small groups. The larger Nazca symbols were close to networks of straight lines, squares, and trapezoids etched into the ground.
These symbols were likely used for ceremonial activities, possibly at the end of a pilgrimage, and could be considered planned public architecture.
Sakai mentioned that efforts are ongoing to decipher the exact meaning of the geoglyphs and their mysterious purpose, which researchers hope to publish in the future. 'Our findings suggest that their meaning is formed through their combinations,' he said, referring to how the Nazca geoglyphs are grouped together.