Google PPE builds geospatial models in minutes

Google’s Planetary Prediction Engine announcement says its system turns natural-language geospatial queries into models in minutes instead of weeks, with mean R² of 76.8% across 21 CDC health indicators versus 60.0% for a manual baseline. Google’s arXiv paper reports 66.1% versus 31.5% on Nigeria food-security downscaling and Recall@10 of 83.3%, identifying 15 of 18 newly invaded Ebola health zones across five weekly forecasts. For operators, PPE makes data discovery and leakage checks look as consequential as model selection; AMIE’s clinical evidence gate is a reminder that impressive benchmark deltas still need domain validation before deployment.