Machine learning algorithm beats meteorologists at predicting the weather

Jan

Jan

Who's the Boss?
Staff member
Joined
May 24, 2010
Messages
13,889
Reaction score
11,826
Location
The Rabbit Hole
AI is beating the weatherman

A recent scientific report has highlighted how the AI-powered GraphCast system was better at predicting weather than meteorologists using conventional forecasting technologies.

The tool — developed by researchers at Google's DeepMind — uses two most recent states of Earth's weather, including several variables at the time of the test and six hours earlier, to predict what the weather will be like six hours later.
 
That makes sense though. Weather prediction already runs on sophisticated algorithms. They've been using supercomputers for years to map Earths weather patterns. One would assume that AI could do this a million times better than a human forecaster.

Our local weather guru @Gordon_R days are numbered. :p
 
If ML is anything like we are seeing on Samsung or Apple devices, yeah, then they are just as bad as current meteorologists..... just throwing the bones out there and hoping for the best.
 
rvZA said:
If ML is anything like we are seeing on Samsung or Apple devices, yeah, then they are just as bad as current meteorologists..... just throwing the bones out there and hoping for the best.
Click to expand...

Showing your ignorance on the topic. :thumbsup:
 
genetic said:
That makes sense though. Weather prediction already runs on sophisticated algorithms. They've been using supercomputers for years to map Earths weather patterns. One would assume that AI could do this a million times better than a human forecaster.

Our local weather guru @Gordon_R days are numbered. :p
Click to expand...

I just finished reading an interesting book about weather forecasting and models. Each year there are incremental improvements, which are fed back into the forecasts. It is a non-sequitur to say that any one technique is better than another. They all have strengths and weaknesses, and include hundreds of parameters. Some data are crucial, others are irrelevant. It takes decades of real weather to define 'better'.

I did not respond to the OP since it is vague and hyped. The article does not specify the criteria, both location and duration. Models are poor at nowcasting, so I can see that machine learning can do better for localised short durations.

It may also be possible to provide a better 'guess' for a specific location based on past events and current conditions. However it is mathematically impossible to predict everything everywhere at once with certainty, so there is always a role for 'boring' models (and percentage rainfall, whatever that means...)
 
Because NWP can't predict the butterfly but ML can.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter