Research@ 2026: 10 ways we're amplifying human ingenuity
Google unveiled new experimental capabilities for Google Earth AI, including deeper geospatial analysis, a planetary prediction engine for risk assessments, and an agent‑to‑agent API that exposes Geospatial Reasoning skills for custom workflows. The rollout adds integration with Google Dataset Search (over 100 M public datasets) and TimesFM for planetary nowcasting, leveraging more than 70 geospatial tools such as the Remote Sensing Foundation Model (RSFM) and Population Dynamics Foundation Model (PDFM). These advances aim to let developers, health officials, and enterprises automate complex geospatial pipelines and proactively address public‑health and climate challenges.
⚡ Key Takeaways
- Google Earth AI now offers an experimental agent‑to‑agent API for programmatic access to satellite imagery parsing, routing, and demographic embeddings.
- The system incorporates the Remote Sensing Foundation Model (RSFM) and Population Dynamics Foundation Model (PDFM) as core geospatial foundation models.
- Integration with Google Dataset Search provides query access to more than 100 million public datasets, expanding data coverage for geospatial tasks.
- TimesFM is introduced for planetary nowcasting and forecasting iterations, enabling scenario‑based risk assessments.
- Google.org is allocating $10 million (including $2 million to The Milken Institute) to support crisis resilience and public‑health projects that will use these tools.
Engineers can now embed sophisticated geospatial reasoning directly into their pipelines via a standardized API, accelerating the development of health‑monitoring, climate‑risk, and infrastructure‑planning applications without building custom vision or routing models from scratch.
✅ Practical Steps
- Use the experimental agent‑to‑agent API to integrate Google Earth AI’s Geospatial Reasoning agent into your workflow for satellite imagery parsing and demographic embedding generation.
- Query the expanded Google Dataset Search catalog (over 100 M datasets) to source training or inference data for your geospatial models.
- Apply TimesFM within the planetary prediction engine to generate nowcasting forecasts for scenario analysis.
Want the full story? Read the original article.
Read on Google Research ↗