00:00China has been using these AI weather prediction models to help track the progress of typhoons
00:05during the typhoon season. The way these AI models work is that they can use vast historical
00:10weather pattern data and be able to predict what's going to happen next. And they have proven to be
00:16very accurate, especially at tracking the flight paths of typhoons. The recent typhoon dolphin
00:21that made landfall in China over the weekend, five days out from that landfall, they were able to
00:26predict it within 30 minutes and 30 kilometers. The way that traditional forecasting works is that
00:31they're numerical models that are run through supercomputers to predict what's going to happen
00:35with the weather. AI models are basically much, much faster and they use much less computing power than
00:41the traditional models do. They do have their strengths and weaknesses though. AI models, for
00:46example, are very, very good at predicting the track of a typhoon, but they're still not as good as the
00:51traditional models at predicting the intensity of a storm. Globally, I think people will know that
00:57there's huge companies like Google, like Nvidia, that have their own AI weather prediction models.
01:02In China, big companies like Huawei are also getting into this kind of AI weather prediction game. Some
01:09of the leading Chinese players are very, very competitive and they've been shown to be able to
01:14have accurate forecasts beyond 10 days, which even globally is quite impressive. I don't think these
01:20AI weather predicting models are going to replace the traditional ways that we predict weather. I
01:26think they're probably going to be used concurrently for quite some time, just because the strengths and
01:31weaknesses are quite different.
Comments