Google Research has unveiled a new suite of geospatial AI models and reasoning agents called Google Earth AI. These models leverage foundation models and cross-modal reasoning to extract actionable insights from satellite imagery and population data. The system uses a reasoning agent, powered by Gemini models, to deconstruct complex questions, execute multi-step plans using various AI models and geospatial tools, and synthesize holistic answers. New innovations include advanced imagery analysis models capable of understanding natural language queries about satellite images and population dynamics models that provide globally consistent and time-updated embeddings. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
RANK_REASON Google Research published a technical paper detailing new geospatial AI models and reasoning agents.
