Tim Fernholtz
speaker
84 appearances
2 recordings
2 series
first heard Feb 2026
last heard 13 Feb
Tim Fernholtz’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 2 in all, peaking in Feb 2026 with 2.
Appearances
And just like with the LLMs and other forms of this technology, there's a lot to learn about what actually is happening.
And then can we take that and use that to learn more about science?
Well, it's certainly part of the trend we're seeing.
So many of the advances in deep learning technology are coming from the private sector because it takes a huge amount of money to train these models to get the data centers all going.
But once they are trained, they're actually incredibly cheap and portable.
And so there is a sense in the meteorology community that this is going to democratize access to weather forecasts.
And it's also important to say that right now, these models, while they are outperforming their government equivalents, they're still dependent on the data that is generated by the governments.
So for now, there's a very intense symbiosis between both sides.
If you're using Google, yes, it is.
Google has been feeding data from its weather model into some of its products.
If you are in Europe, you may see it in your daily forecast because the Europeans are much further ahead of the United States.
But even in December, the National Weather Service rolled out some AI models of its own that are going to start contributing to forecasts.
The most important question.
Yeah, so it's hard to say how quickly this will take effect.
But what these experiments reveal is that there is a lot of potential to extend weather forecasting out to perhaps even a month in the future.
They haven't quite figured out how to do it yet, but they can see the potential path towards it.
And so as these models continue to be refined, they will point us towards better forecasts, but they will also perhaps point us towards where we need to invest more in data collection to improve the data we're putting into them to get better forecasts, and also where we can tweak the traditional models to get better results from them.
So it's an ongoing conversation between all sort of elements of the weather system and how we improve the forecast we're getting.
You're very welcome.
Thanks for having me.
Showing 61–80 of 84 · page 4 of 5
← Previous
Next →