An artificial intelligence model created by Deepmind, a company owned by Google, has surpassed conventional weather forecasters for the first time, according to recent research.
Peer-reviewed research found that Deepmind’s GraphCast AI model performed noticeably better than the European Centre for Medium-range Weather Forecast’s traditional weather prediction system.
The study claims that the AI model, which can generate a precise 10-day forecast in less than a minute, marks a revolution in weather forecasting.
According to Pushmeet Kohli, vice president of research at Deepmind, weather prediction is one of the most difficult issues that humanity has faced in a very long time.
He stated that this is a very significant issue when considering the events surrounding climate change in the last few years.
39 years of historical weather data from the ECMWF are used to train GraphCast. The study revealed that GraphCast’s three to ten day weather forecasts were more accurate than traditional forecasts.
When addressing the uncertainty present in longer-term weather forecasts, Deepmind cautioned that the model did have certain limitations in comparison to non-AI forecasts.
The project’s architects revealed that, three days ahead of conventional forecasters, GraphCast had accurately predicted the landfall location of Hurricane Lee, a strong storm that slammed into North America earlier this year.
GraphCast is a major and welcome step forward for the industry, according to Matthew Chantry, an ECMWF machine learning specialist.
Speaking to the Financial Times, he said that models like GraphCast, which are trained on a larger range of historical data, will end up being far less expensive than the state-of-the-art weather forecasting techniques, which depend on potent Supercomputers.
He told that the models like GraphCast, which are trained on a larger range of historical data, will work out to be far less expensive than the current weather forecasting techniques, which depend on potent (and costly) supercomputers.
He said, they might be talking about 1,000 times cheaper in terms of energy consumption and that’s a fantastic improvement.
