Zoe Kleinman

And it's basically an AI program that does weather predictions. And it performed 20 percent better in tests than the current world leader. And what's really interesting about it, not only was it better and more accurate on day to day forecasts, it was also more accurate on extreme event prediction up to 15 days in advance.

And it was better at predicting the paths of hurricanes, which, as we know, can change. change quite dramatically at very short notice. It was trained on 40 years of weather data, which is an awful lot of data and also contained an awful lot of variables. And it used that data to make its own predictions.

And it was better at predicting the paths of hurricanes, which, as we know, can change. change quite dramatically at very short notice. It was trained on 40 years of weather data, which is an awful lot of data and also contained an awful lot of variables. And it used that data to make its own predictions.

And it was better at predicting the paths of hurricanes, which, as we know, can change. change quite dramatically at very short notice. It was trained on 40 years of weather data, which is an awful lot of data and also contained an awful lot of variables. And it used that data to make its own predictions.

It does the same as traditional predictions, by the way, which is called ensemble forecasting, where you get a range of them. But it was calculating which one was the most probable and it was very good at it. Traditionally, forecast tools involve basically solving enormous amounts of equations and they use supercomputers to do it because they have to. But it still takes a long time.

It does the same as traditional predictions, by the way, which is called ensemble forecasting, where you get a range of them. But it was calculating which one was the most probable and it was very good at it. Traditionally, forecast tools involve basically solving enormous amounts of equations and they use supercomputers to do it because they have to. But it still takes a long time.

It does the same as traditional predictions, by the way, which is called ensemble forecasting, where you get a range of them. But it was calculating which one was the most probable and it was very good at it. Traditionally, forecast tools involve basically solving enormous amounts of equations and they use supercomputers to do it because they have to. But it still takes a long time.

And Gencast managed to process its forecasts in eight minutes, which is considerably faster, isn't it? But, of course, the data it was using had been gathered and processed in all of our old school ways that we've been doing for the last 40 plus years.

And Gencast managed to process its forecasts in eight minutes, which is considerably faster, isn't it? But, of course, the data it was using had been gathered and processed in all of our old school ways that we've been doing for the last 40 plus years.

And Gencast managed to process its forecasts in eight minutes, which is considerably faster, isn't it? But, of course, the data it was using had been gathered and processed in all of our old school ways that we've been doing for the last 40 plus years.

And so what we don't know is whether going forward, if we were to just rely on AI generated data, because that would become, wouldn't it, our data set, whether it would be as reliable.

And so what we don't know is whether going forward, if we were to just rely on AI generated data, because that would become, wouldn't it, our data set, whether it would be as reliable.

And so what we don't know is whether going forward, if we were to just rely on AI generated data, because that would become, wouldn't it, our data set, whether it would be as reliable.

I think it's certainly got the potential to do that. And we're going to see more and more companies focusing on that. I guess one of the ironies is that it might end up being matters related to climate change that could hold it back because we know that AI is a very power hungry tech. It's very energy hungry. It needs banks of computers in massive data centres to run it.

I think it's certainly got the potential to do that. And we're going to see more and more companies focusing on that. I guess one of the ironies is that it might end up being matters related to climate change that could hold it back because we know that AI is a very power hungry tech. It's very energy hungry. It needs banks of computers in massive data centres to run it.

I think it's certainly got the potential to do that. And we're going to see more and more companies focusing on that. I guess one of the ironies is that it might end up being matters related to climate change that could hold it back because we know that AI is a very power hungry tech. It's very energy hungry. It needs banks of computers in massive data centres to run it.

And a lot of those at the moment require large quantities of water to keep them cool. And so although there is hope that perhaps it will come up with solutions to the climate change problem that we haven't yet thought of, there is an environmental price tag attached that is... currently quite high, and some experts question whether we can actually afford it in terms of the Earth's resources.

And a lot of those at the moment require large quantities of water to keep them cool. And so although there is hope that perhaps it will come up with solutions to the climate change problem that we haven't yet thought of, there is an environmental price tag attached that is... currently quite high, and some experts question whether we can actually afford it in terms of the Earth's resources.

And a lot of those at the moment require large quantities of water to keep them cool. And so although there is hope that perhaps it will come up with solutions to the climate change problem that we haven't yet thought of, there is an environmental price tag attached that is... currently quite high, and some experts question whether we can actually afford it in terms of the Earth's resources.