Ai2 lands Google grant for climate models

Seattle's Allen Institute for Artificial Intelligence has secured a substantial financial package from Google to develop advanced environmental forecasting tools. Google is providing the organization, widely known as Ai2, with approximately $5 million in direct funding, complemented by an additional $1.8 million in cloud credits. Announced on Wednesday, the three-year collaboration aims to transform an artificial intelligence weather model originally developed by Google Research into a comprehensive system capable of projecting climate change decades into the future.

The initiative pairs philanthropic support from Google.org with hands-on technical assistance from Google Research engineers. To accommodate the expanded project scope, Ai2 plans to incorporate four new researchers into its existing eight-person climate modeling team. Oliver Watt-Meyer, Ai2's director of research for AI climate modeling, said that climate forecasting functions as a critical public good that is frequently required by government agencies to plan for future environmental shifts.

The forthcoming open-source model, designated NeuralGCM+, will build directly upon Google's existing NeuralGCM framework, which uniquely integrates traditional physics equations with modern machine learning techniques. Developers intend to provide scientists, water management districts, and regional planners with significantly faster and more accurate forecasts concerning how rainfall, regional temperatures, and water supplies could shift over a timeline spanning five to fifty years.

While Google's previous machine learning efforts have focused primarily on shorter-term weather forecasting looking ahead up to a single year, long-term climate projections require different technical approaches and public sector applications. The expanded team will work to address previous limitations observed in long-run simulations as the three-year grant period gets underway.