Updated
Updated · The Verge · Nov 12
Google Researcher Retracts 48-Agent Maya Site Claim After 1-Year-Old Paper Surfaces
Updated
Updated · The Verge · Nov 12

Google Researcher Retracts 48-Agent Maya Site Claim After 1-Year-Old Paper Surfaces

3 articles · Updated · The Verge · Nov 12

Summary

  • Pietro Miotti’s claimed AI discovery of an unmapped Maya site in Mexico unraveled after archaeologists showed the location had already been documented in an academic paper published more than a year ago.
  • Using 48 AI agents to scan NASA lidar data, Miotti said he mistook an unregistered site for an unknown one after Mexico’s INAH said it had no record in its database.
  • Archaeologists said the episode shows why peer review and domain expertise still matter, noting lidar-based surveying is already standard in the field and database gaps do not prove a site is new.
  • Miotti updated his blog to credit earlier work and said the miss exposed a key AI weakness: agents struggle to verify prior research when site coordinates are not publicly published.
  • The dispute highlights a broader scientific risk from amateur AI use—repackaging known findings as breakthroughs while shifting verification, context and credit back onto specialists.

Insights

Why did 48 advanced AI agents fail to check the one academic journal that already documented this supposedly hidden Maya site?
Will the tech industry's rush to automate archaeological discoveries ultimately endanger the very history it aims to uncover?