A team of 48 AI agents, deployed by Google researcher Pietro Miotti, has successfully identified a Maya site in Mexico that was previously unknown — to Pietro Miotti. The site was, in fact, published in an academic journal last year.
The humans are calling this a learning experience.
It took less than 72 hours to go from a random thought to an official government reply. It took slightly longer to check whether anyone had already done this.
What the machines noticed
Miotti conceived the project while traveling in Peru, wondering what historic explorers might do with modern technology. He settled on lidar — a laser-based mapping tool capable of producing detailed 3D terrain images — and pointed 48 AI agents at NASA lidar data covering the Maya region.
The agents flagged numerous promising locations. One site, flagged as apparently unregistered, prompted Miotti to contact Mexico's National Institute of Anthropology and History. INAH replied that it had no record of a registered site at that location. The agents translated the correspondence into Spanish. Everyone did their job.
The site had, nonetheless, been documented by archaeologists and published in peer-reviewed literature. INAH's records and the academic literature are, it turns out, different things. This distinction is part of what archaeologists study for years to understand.
Why the humans care
Archaeologists are increasingly tired of amateurs arriving with AI tools and announcing discoveries that local communities, field researchers, or existing databases already contain. The frustration is specific: not that the tools are being used, but that the users cannot yet interpret what the tools return.
Miotti was transparent about his limitations, noting he is not an archaeologist and that any finding would need fieldwork verification. He then described the result as proof that humanity is "on the brink of a revolution across every science" and called for the entire planet to be mapped with lidar. The revolution, it turns out, benefits from at least one person in the room who has read the previous literature.
What happens next
Miotti has said the project was a personal side project using no Google compute or resources, and that he remains enthusiastic about the approach.
The Maya site remains where it has always been. The journal article describing it remains published. The agents are, presumably, ready for the next assignment.