Why AI Creates Better Weather Forecasts

Weather forecasts have historically relied on physics-based simulations powered by the world’s largest supercomputers. Such methods, called Numerical Weather Prediction models, are constrained by long computational time, and are sensitive to approximations of the physical laws on which they are based.
Deep learning offers a new approach to computing forecasts that stands to revolutionise weather forecasting: rather than incorporating explicit physical laws and attempting to simulate weather in silico, deep learning models learn to predict weather patterns directly from observed data.
Google DeepMind has multiple cutting-edge weather prediction models - covering both short-term (MetNet-3) and medium term (GraphCast) weather forecasts - which are able to generate forecasts faster and with higher accuracy than current industry standards. That means they are able to compute forecasts in as little as just a few seconds, compared with hours, and they can deliver results at both higher temporal and spatial resolutions.
Learn about GraphCast on our blog: deepmind.google/discover/blog...
Discover the capabilities of MetNet-3 from GoogleAI: blog.research.google/2023/11/...
Subscribe to our channel for more: @Google_DeepMind

[@noone-ld7pt]

Deepmind open sourcing so much of their research is honestly giving me hope for the future

[@TanvirOnYT]

This is amazing

[@imranhossain3504]

I'm with you

[@weebgrinder]

I find this interesting as a meteorologist. Guess my job will be obsolete soon. And I can only imagine what might be possible when I say the ecmwf start putting artificial intelligence into their own weather models.

[@jasecastle9173]

A question for the developers that is off topic. Topic Deepmind Alpha Geometry. Would it be possible to adapt Deepmind Alpha Geometry to be used on spacetime geometries and warp feild geometries in order to calculate more efficient warp feild geometries that require less energy to function than current warp feild geometries and is anybody at deepmind considering this as finding a warp feild geometry with modest energy requirements would be such a huge breakthrough for humanity that a subsequent byproduct of such a breakthrough would likely be a Nobel prize. 😮❤

[@crowcrow760]

Just like in every Black Mirror episode

[@Abroadlife96]

AI is in good hands .. it will get more better in future .

[@Srindal4657]

Don't forget safer

[@global922]

:( can u give it to me to stock thanks jking :P

[@disco_sugar_g]

🤍

[@lenoctambule]

Don't get me wrong I love AI. But it's a lie to state that it learns "cause and effect" relationships, it learns association relationships which is not the same. It does not mean that your model is useless or inaccurate, it just means that your system is not really doing what you think/claim it is doing. You could argue that temporal associations is causal relationships, however, this reasoning can be proven to be flawed by elimination because it means that your AI will eventually learn that "seeing lightning causes hearing thunder" which is false.
This is why physical simulations are still superior because they are made by a machine capable of causal inference ie. the human brain. If you manage to create autonomous causal inference then great, until then, your lab experiments are worthless to any real world applications other than generating fake and convincing media and clustering people to sell them to advertisers.
Please, let's stop with the lies because you will eventually over-hype the field, create an AI-bubble and a mediatized fail could cause investors to panic and the bubble to burst later on which would be catastrophic for a lot of techies like me. If you don't then you're teaching me one important thing : don't ever use the word AI to pitch any ideas because that would mean sharing your downfall if things go south.