# Generative AI Quantifies Uncertainty in Weather Forecasting Google's SEEDS model measures weather forecast uncertainty at low cost. - **Source:** [Google AI Blog](https://ai.googleblog.com) - **Published:** 2024-03-29 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/ce5a27b0-cb35-46d2-855a-3281a0c4683c ## Importance of Weather Forecasting Accurate weather forecasts directly impact lives, from daily activities to emergency responses. Climate change makes these predictions even more critical. Google innovates with models like MetNet-3 and GraphCast to address this challenge. ## The Uncertainty Challenge Weather is inherently stochastic. Traditional methods use physics-based simulations to quantify uncertainty but require high computational costs. Large ensembles are needed to detect rare and extreme weather events. ## SEEDS Technology Google's SEEDS model leverages generative AI to produce large-scale ensemble forecasts at a fraction of traditional costs. This technology opens new opportunities in climate science by applying diffusion models to weather prediction. ## The Butterfly Effect Edward Lorenz's 1972 'butterfly effect' theory explained weather uncertainty. Small errors in initial conditions can lead to large forecast deviations, making deterministic predictions inadequate for decision-making. ## Probabilistic Forecasts Global weather agencies issue probabilistic forecasts using ensemble methods, but computational limits restrict their size. SEEDS solves this by generating large ensembles efficiently. ## Extreme Events Predicting hurricanes, heatwaves, or floods requires massive ensembles. For example, a 10,000-member ensemble is needed to forecast 1%-probability events. SEEDS makes this scale economically feasible. ## SEEDS Advantages SEEDS generates large ensembles using just a few operational forecasts. Its outputs match or exceed physics-based ensembles in accuracy while producing realistic weather patterns. ## Computational Efficiency SEEDS performs in minutes what supercomputers take hours to calculate. It produces 256 ensemble members in 3 minutes on Google Cloud TPUv3-32 and scales effortlessly. ## Applications SEEDS supports emergency management, energy trading, and climate research. By accurately predicting rare events, it aids critical decision-making. ## Future Outlook Applying generative AI to weather forecasting is a breakthrough for climate resilience. SEEDS enables more accurate, accessible forecasts, strengthening societal preparedness. ## Sources covering this story - http://blog.research.google/2024/03/generative-ai-to-quantify-uncertainty.html — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._