Foundation models like Google’s Population Dynamics Foundation Model (PDFM) are trained on massive, diverse datasets like satellite imagery, mobile data, POIs, and more to learn a deep, general-purpose understanding of how people interact with places.
Like Large Language Models (LLMs), which understand and generate text, Foundational Models can be trained to recognize patterns and understand relationships in other data types, such as geospatial data. This allows them to perform tasks like creating embeddings from satellite imagery or enriching location data with semantic context. They transform spatial analytics by enabling new forms of analysis and automating complex tasks that were previously impossible at scale.
.png)
