23 Daily — According to theverge.com, Google researcher Pietro Miotti claimed earlier this week to have discovered a potentially unmapped Maya site in Mexico using a team of AI agents. However, the site had already been discovered and published in an academic journal last year. The incident sparked criticism from archaeologists online and highlights growing tensions in science as amateurs, equipped with increasingly powerful AI tools, venture into specialist fields often lacking the necessary expertise to interpret their findings accurately.
The AI-driven Search for Maya Sites and Its Findings
Miotti was inspired during his travels in Peru to consider how historical explorers might have used modern technology. He deployed 48 AI agents to analyze NASA LiDAR data, which provides detailed three-dimensional environmental images, to search for undocumented archeological sites. Due to limited public LiDAR data coverage in Peru, he shifted his focus to the Maya region. The AI agents identified several promising locations, including a major village not recorded in existing databases. Miotti contacted Mexico’s National Institute of Anthropology and History (INAH), using AI to translate his correspondence into Spanish. INAH reportedly had no record of a registered site at that location. Miotti emphasized that this was a personal side project without Google computing resources and acknowledged that fieldwork verification is necessary.
Academic Responses and Challenges in Archeological Data
While Miotti saw the discovery as proof of an upcoming revolution in various sciences due to AI, archaeologists quickly pointed out that the site had been documented in an academic journal more than a year earlier. Archaeology professor Sarah Parcak stressed the importance of peer review and thorough cross-referencing with existing research, noting a lack of government database records does not imply unfamiliarity. Other archaeologists criticized the project for spreading misinformation and overlooking decades of archaeologists’ use of advanced digital tools. Miotti admitted his lack of archaeological expertise and explained that archaeologists typically do not publish exact site coordinates, making it difficult for AI agents to confirm prior knowledge solely through database searches; instead, comparing images to figures in academic papers is required, presenting a challenge for AI. He updated his blog posts to credit original researchers once the oversight was pointed out.
Editor's take: Opportunities and Verification in AI-assisted Archaeology
Miotti’s project highlights the potential of AI in analyzing archaeological big data, especially LiDAR images, to identify possible new sites and assist experts in prioritizing fieldwork. For entrepreneurs, this suggests opportunities to develop AI-powered products or services that integrate image recognition with historical data management, targeting niches like archaeology and cultural heritage preservation. Verification requires coupling AI findings with existing academic literature and expert validation, alongside on-site expeditions to confirm discoveries scientifically. Creating collaborative platforms that facilitate expert review and feedback on AI outputs will be key to advancing the adoption and reliability of such technologies.