Article by Lori Regattieri: “What if I told you that one of the most well-capitalized AI companies on the planet is asking volunteers to help them uncover “lost cities” in the Amazonia—by feeding machine learning models with open satellite data, lidar, “colonial” text and map records, and indigenous oral histories? This is the premise of the OpenAI to Z Challenge, a Kaggle-hosted hackathon framed as a platform to “push the limits” of AI through global knowledge cooperation. In practice, this is a product development experiment cloaked as public participation. The contributions of users, the mapping of biocultural data, and the modeling of ancestral landscapes all feed into the refinement of OpenAI’s proprietary systems. The task itself may appear novel. The logic is not. This is the familiar playbook of Big Tech firms—capture public knowledge, reframe it as open input, and channel it into infrastructure that serves commercial, rather than communal goals.
The “challenge” is marketed as a “digital archaeology” experiment, it invites participants from all around the world to search for “hidden” archaeological sites in the Amazonia biome (Brazil, Bolivia, Columbia, Ecuador, Guyana, Peru, Suriname, Venezuela, and French Guiana) using a curated stack of open-source data. The competition requires participants to use OpenAI’s latest GPT-4.1 and the o3/o4-mini models to parse multispectral satellite imagery, LiDAR-derived elevation maps (Light Detection and Ranging is a remote sensing technology that uses laser pulses to generate high-resolution 3D models of terrain, including areas covered by dense vegetation), historical maps, and digitized ethnographic archives. The coding teams or individuals need to geolocate “potential” archaeological sites, argue their significance using verifiable public sources, and present reproducible methodologies. Prize incentives total $400,000 USD, with a first-place award of $250,000 split between cash and OpenAI API credits.
While framed as a novel invitation to “anyone” to do archaeological research, the competition focuses mainly on the Brazilian territory, transforming the Amazonia and its peoples into an open laboratory for model testing. What is presented as scientific crowdsourcing is in fact a carefully designed mechanism for refining geospatial AI at scale. Participants supply not just labor and insight, but novel training and evaluation strategies that extend far beyond heritage science and into the commercial logics of spatial computing…(More)”.