52
Google’s new weather prediction system combines AI with traditional physics.
(www.technologyreview.com)
this post was submitted on 22 Jul 2024
52 points (91.9% liked)
Technology
59990 readers
2301 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Ummm was weather foresting not always done using probalistic interference models, i.e AI?
Weather forecasts have recently felt like they've been less accurate, i.e. you maybe get a good level of confidence for a day, but two days and it might be completely different. This makes sense given the climate is changing and previous models wont fit as well...
Are LLMs going to consume search data for raincoats and air-conditioning to improve the weather forecast. Clearly time to invest in AI now, the revolution is here!
Weather forecasting does create ensemble models to help constrain their forecasts. They'll adjust some of their inputs in each model, mainly as a way of embedding the uncertainty in the measured data, then run that model and see if it changed.
This resembles AI on one level, but it's at a dramatically different scale. An ensemble may contain a few hundred runs at most, but an AI needs tens of thousands of data points at minimum. In order to make predictions like what google is saying they can do, they'd need to train on billions or maybe trillions of data points.
This is still fundamentally different than ensemble modeling though. Ensembles are physically informed and the perturbations are based on real assumptions. Each model in an ensemble is based on validated physics equations. An AI model would undermine that completely. You can't possibly describe the underlying equations because there aren't any, so you can't analyze its accuracy or propose a more accurate model, you're just stuck with a bunch of coefficients that you'll never understand.
I've worked in climate modeling, and this kind of AI work is nothing more than an electricity sink for at least a decade, maybe forever.