This is a fascinating piece, Neil. The shift from hours on a multi-million-dollar supercomputer to under a minute on a single machine is a massive step toward democratizing global forecasting.
It’s a great example of the impact data and AI can have on society. Weather forecasting is already arguably the single most consumed type of data on the planet!
However, as you rightly point out at the end, a forecast is only as good as the infrastructure available to act on it. Having moved a few years ago from Europe to Australia, the drop-off in local forecasting reliability has been striking. It underscores that while we can democratize the computation, we haven't yet democratized the data collection (ocean data gaps in the Southern Hemisphere) or the localized emergency response systems.
I know only too well how the weather is more unpredictable having lived in Melbourne.
One of the things the article also highlights is that AI cannot do it alone. It needs the scientific data and measurements that is still a core part of forecasting. Yes AI can model - but it can only model what data it is given.
Again goes to show that you need systems thinking on how to integrate technology for the right purpose and the right time.
This is a fascinating piece, Neil. The shift from hours on a multi-million-dollar supercomputer to under a minute on a single machine is a massive step toward democratizing global forecasting.
It’s a great example of the impact data and AI can have on society. Weather forecasting is already arguably the single most consumed type of data on the planet!
However, as you rightly point out at the end, a forecast is only as good as the infrastructure available to act on it. Having moved a few years ago from Europe to Australia, the drop-off in local forecasting reliability has been striking. It underscores that while we can democratize the computation, we haven't yet democratized the data collection (ocean data gaps in the Southern Hemisphere) or the localized emergency response systems.
I know only too well how the weather is more unpredictable having lived in Melbourne.
One of the things the article also highlights is that AI cannot do it alone. It needs the scientific data and measurements that is still a core part of forecasting. Yes AI can model - but it can only model what data it is given.
Again goes to show that you need systems thinking on how to integrate technology for the right purpose and the right time.