A new AI weather prediction system, Aardvark Weather, can deliver accurate forecasts tens of times faster and using thousands of times less computing power than current AI and physics-based forecasting systems, according to research published today (Thursday 20 March) in Nature.
Very interesting! I hope this will be able to run locally.
The paper: https://arxiv.org/abs/2404.00411
Here’s a hint: it’s gonna get real fucked up really soon and won’t show any signs of slowing down, much less reversal, for centuries.
How much energy did I just save human civilization as a whole? At a guess?
A negligible amount in this case. This isn’t an LLM, it’s a much simpler form of ML that isn’t enormously wasteful.
This is the good type of ML, because it helps scientists while being cheap and easy to run.
EDIT: This may also be a type of model with deterministic output. I’m unsure of that, but that would mean it’d be possible to correct incorrect predictions.
Fair point. For the match: how’s the weather over there?
Unseasonably warm!
How unforeseeable.
2025 Mar 20
(…) Aardvark has been developed by researchers from the University of Cambridge (UK) supported by the Alan Turing Institute (UK), Microsoft Research (USA ?) and the European Centre for Medium Range Weather Forecasting (EU) (…)
So it is safe from USA’s collapse in science.
(…) But with Aardvark, researchers have replaced the entire weather prediction pipeline with a single, simple machine learning model. The new model takes in observations from satellites, weather stations and other sensors and outputs both global and local forecasts. This fully AI driven approach means that predictions are now achievable in minutes on a desktop computer. (…)
a.i. is so far from simple “word completion”.
Nobody said ai is simple word
competitioncompletion. It’s pattern recognition, except without any logic to discern the signal from the noise (causation from correlation).i agree there are many works in AI which are far different from large language models but I saw some bad comments negating any potential for artificial intelligence so I’m happy to see we agree on this.
As for “causation”, maybe we can agree there are many levels of causation … like what’s the cause of the cause … i will not be surprised when we see artificial systems good at this game. i will not be surprised either when robots, driven by artificial intelligence, begin to read the world by themselves, then, built new models of advance robots and evolve by themselves.
