No, AI agents did not discover a previously unknown Maya site
Original reporting by The Verge
A recent claim by Google researcher Pietro Miotti of an AI-powered discovery of a "new" Maya site in Mexico has ignited a wider debate about the role of artificial intelligence in specialized scientific fields. Miotti, an AI researcher with no archaeological background, leveraged a team of 48 AI agents to scour NASA lidar data for undocumented ancient settlements. When his agents flagged what appeared to be an unrecorded major village, he contacted Mexico’s National Institute of Anthropology and History (INAH). Miotti hailed receiving an official reply within 72 hours stating INAH had no registered record of the site as proof that humanity was "on the brink of a revolution across every science."
The archaeological response
There was just one problem: the site had already been discovered and published in an academic journal more than a year prior. The incident quickly drew sharp criticism from archaeologists, who highlighted the dangers of deploying powerful AI tools into specialist fields without the necessary expertise or rigorous engagement with existing research. Experts like Sarah Parcak and Elizabeth Graham, while acknowledging the potential of lidar and AI, critiqued the project for potentially spreading misinformation and overlooking decades of archaeological work that already utilizes advanced digital tools. They underscored that local communities often already know about such sites, and the real challenge in archaeology is funding for excavation, not merely "discovery." This episode underscores a growing tension, revealing how impressive AI capabilities can still fall short without the contextual knowledge and verification protocols that define scientific inquiry, often shifting the burden of initial validation onto specialists.
The incident surrounding Pietro Miotti’s purported Maya discovery sharply illustrates the burgeoning tension between the raw computational power of artificial intelligence and the indispensable depth of human expertise in specialized fields. While AI agents demonstrated impressive capability in sifting through vast lidar data, their failure to consult existing academic literature — a fundamental step in any scholarly endeavor — led to a well-intentioned but ultimately misleading claim. This episode underscores that technological prowess, however advanced, cannot unilaterally replace the contextual knowledge, rigorous verification, and understanding of prior art that seasoned specialists bring to their work. Archeologists, far from being Luddites, have long embraced sophisticated digital tools; their concern lies with the uncredited duplication of effort and the potential for misinformation when these powerful new tools are wielded without proper domain knowledge.
Redefining Discovery
More broadly, this case foreshadows significant challenges and opportunities for scientific research across disciplines. As AI becomes increasingly accessible, the scientific community must contend with a potential deluge of "discoveries" lacking critical context, originality, or even awareness of local community knowledge. This not only risks overwhelming experts but also shifts the burden of proof and verification onto them, potentially diverting valuable resources from genuine novel inquiry. The incident highlights the urgent need for AI development that prioritizes seamless integration with established scholarly processes, including robust mechanisms for checking existing research and acknowledging the often-invisible work already done. Ultimately, the future impact of AI in science will depend not just on its ability to process data, but on its capacity to respect, contextualize, and collaboratively build upon the collective human knowledge accumulated over centuries, redefining what truly constitutes a scientific breakthrough.
Frequently asked questions
- Can AI reliably discover new archaeological sites without human expert involvement?
- AI tools, particularly when analyzing lidar data, can identify potential archaeological features. However, relying solely on AI without expert oversight can lead to false claims of discovery. Specialists are crucial for cross-referencing findings with existing published research, local community knowledge, and understanding the significance and context of sites. Without human archaeological expertise, AI-flagged locations risk being already known or misinterpreted, potentially causing misinformation.
- Why do AI-claimed archaeological discoveries sometimes turn out to be previously known?
- AI tools often analyze vast datasets like lidar imagery, flagging anomalies that resemble archaeological features. However, they may not integrate with the full body of existing academic literature or unpublished research. Archaeologists typically don't publish exact coordinates, making it hard for AI to cross-reference image data directly. Additionally, official government databases may not contain all known sites, leading AI to identify features already documented by specialists or known by local communities.
- What are archaeologists' main concerns regarding AI use in uncovering ancient sites?
- Archaeologists express concerns that AI tools, especially when used by non-experts, can lead to redundant "discoveries" already known or documented. They emphasize that while AI can help process data, it lacks the contextual understanding, peer review process, and deep knowledge required to interpret findings accurately. There are also worries about being overwhelmed by unverified claims, the potential for misinformation, and even the inadvertent publicizing of site locations, which could attract looters.