Google introduces GenCast, its new artificial intelligence model for accurate weather forecasts

Publié le 21 February 2025 à 06h19
modifié le 21 February 2025 à 06h19

Google unveils GenCast, a revolutionary artificial intelligence model dedicated to weather forecasting. *With unprecedented capability*, it generates accurate forecasts in record time for 15 days. This development represents a decisive turning point in the fight against climate disasters, *offering a probabilistic approach* more relevant than ever. Google’s commitment to refining this technology could transform our interactions with the weather on a daily basis, influencing fields as diverse as agriculture and renewable energy.

Introduction to GenCast

Google recently revealed GenCast, an innovative artificial intelligence model for weather forecasting. Developed by DeepMind, this model provides weather estimates with speed and accuracy significantly superior to traditional methods. GenCast offers forecasts over a period of up to 15 days while employing a probabilistic approach to determine various weather scenarios.

A revolutionary technology

GenCast stands out with its ability to generate forecasts in just eight minutes, utilizing a single TPU chip. This speed represents a radical change for weather forecasting, which is typically lengthy and often imprecise. Scientists assert that GenCast has *systematically outperformed* the forecasts of the best current operational systems.

Probabilistic approach

This model adopts a probabilistic method, making over fifty predictions regarding weather changes. Utilizing data from the European Centre for Medium-Range Weather Forecasts (ECMWF), it incorporates variables such as temperature, wind, and pressure, covering decades of weather history.

GenCast’s performance in extreme conditions

GenCast’s performance continues to impress, particularly concerning forecasts for extreme weather events such as heatwaves or cold storms. It is also capable of providing accurate information on tropical cyclone paths up to five days in advance.

Transparency and public access

In a move towards transparency, Google’s research team has made the GenCast model code available. Historical and real-time forecasts will soon be accessible, thereby enhancing information exchange and continuous improvement of forecasts.

Competition in the weather forecasting sector

Google positions itself in a space already explored by other tech companies. Microsoft recently introduced Aurora, a model capable of analyzing a billion atmospheric parameters to mitigate the impacts of climate events. The competition in this sector could lead to significant advancements.

The case of Huawei and its own innovations

Huawei has also distinguished itself by developing several AI models for weather forecasting. The Pangu-Weather model, trained on 43 years of data, enables tracking the paths of typhoons and predicting temperature variations with an accuracy reflecting a 20% improvement over traditional methods.

FAQ about GenCast, Google’s artificial intelligence model for weather forecasting

What is Google’s GenCast model?
GenCast is an artificial intelligence model developed by Google and DeepMind, designed to provide accurate and rapid weather forecasts over a 15-day period using a probabilistic approach.
How does GenCast improve weather forecasts compared to traditional models?
GenCast delivers more accurate and faster forecasts by generating multiple possible weather scenarios, allowing for better anticipation of future conditions compared to traditional models that often provide only a single estimate.
How long does it take GenCast to generate forecasts?
GenCast can generate a 15-day weather forecast scenario in just 8 minutes on a TPU chip, offering unparalleled speed of execution.
What type of data does GenCast use for its forecasts?
GenCast is trained on 40 years of historical data from the European Centre for Medium-Range Weather Forecasts (ECMWF), integrating variables such as temperature, atmospheric pressure, and wind speed at various altitudes.
What types of weather phenomena can GenCast predict?
GenCast excels at forecasting events such as heatwaves, cold spells, and high winds, and it also provides improved forecasts for the paths of tropical cyclones.
Is GenCast accessible to the public?
Yes, the source code for GenCast has been made public by researchers, allowing other developers and researchers to explore and adapt it for their own uses.
How does GenCast assist in preparing for natural disasters?
By providing accurate and reliable forecasts over an extended period, GenCast enables authorities and populations to better prepare and mitigate the impacts of natural disasters associated with extreme weather events.
How does GenCast compare to other AI models for weather forecasting?
GenCast has demonstrated superior performance to existing operational weather forecasting models in terms of accuracy, which could transform the field of meteorology.
Does GenCast consider climate change?
Although GenCast is an advanced model, it still needs to take into account the uncertainties related to the current climate to continually improve its performance and reliability.

actu.iaNon classéGoogle introduces GenCast, its new artificial intelligence model for accurate weather forecasts

How Google lost over 150 billion dollars following this ‘phrase’ uttered by OpenAI’s CEO, Sam Altman

découvrez comment une simple déclaration de sam altman, pdg d’openai, a provoqué une chute de plus de 150 milliards de dollars dans la valorisation de google, bouleversant le secteur de la tech.

OpenAI unveils Atlas: an innovative internet browser powered by ChatGPT, ready to compete with Google

découvrez atlas, le nouveau navigateur web révolutionnaire signé openai, propulsé par chatgpt. innovation, rapidité et intelligence artificielle s'allient pour offrir une expérience de navigation unique, prête à concurrencer google.

OpenAI’s AgentKit: the long road ahead before giving birth to our agent

découvrez pourquoi agentkit d'openai représente une étape prometteuse mais encore incomplète vers la création d'un véritable agent autonome. analyse des défis à relever avant de voir naître un agent intelligent pleinement opérationnel.

Discover Claude Code on the web and iOS: Complete access guide

découvrez comment accéder facilement à claude code sur le web et ios grâce à notre guide complet. profitez d'une présentation claire des étapes pour utiliser claude code sur tous vos appareils.
des centaines d'experts et pionniers de l'ia lancent un appel urgent pour ralentir le développement de l'intelligence artificielle surpuissante, mettant en garde contre les risques et la nécessité d'un encadrement éthique accru.

Harry and Meghan join AI pioneers to call for a ban on superintelligent systems

harry et meghan rejoignent des experts en intelligence artificielle pour demander l'interdiction des systèmes d'ia superintelligents, soulignant les risques potentiels et appelant à une action internationale urgente.