The beginning of the story is a little haphazard. Google researcher Pietro Miotti traveled to Peru in October and chatted with Gemini about Hiram Bingham, who "discovered" Machu Picchu in 1911. He asked casually: If Bingham was born today, how would he find the ancient city?
Gemini, ChatGPT, Claude, and Grok gave the same answer: use lidar. This technology can penetrate the tree canopy and sweep out the terrain beneath the forest.
Peru was not covered by NASA's public lidar data, so he changed places and turned to the Maya region of Mexico. That night, he asked a group of agents to analyze lidar data collected by NASA in 2013, and then went to bed.
Overnight, 48 agents went through the data and marked a possible Mayan settlement that had not been recorded: regular platforms, suspected building bases, and straight-line structures that did not match the natural terrain.
We should look at this matter calmly. The word "suspected" is very important. Anomalies on the lidar map will ultimately need to be confirmed by archaeologists on site and to determine whether they are modern farmland or roads. There have been previous examples of “ruins” marked by AI that turned out to be quarries.
But it illustrates something real: the bar for analyzing publicly available scientific data has been lowered. What used to take a team several weeks to conduct preliminary screening can now be done by one person in just one sleep. What is really scarce is people who ask the right questions and people who can go to the site to verify.