Surely all weather forecasting other than the immediate picture is a trend.
What that trend is most to be in 3, 6, 12, 24, or even 72 hrs time depends on the analysis of many factors.
The more expertise the analyser has the more likely the trend is to be correct.
Whether AI can beat an experienced meteorologist at this is the question.
It’s a bit like a movie producer, director or actor. Those who have a track record at the box office are the ones who get the backing and the big bucks. But there are no guarantees and the backer can sometimes lose their shirt. I seriously doubt that at the moment if ever AI could direct a movie better than a top director just as I seriously doubt that AI will do a better job of weather forecasting than a top meteorologist.
Firstly I think it will take a long time before AI can develop anything like a 6th sense.
Secondly, IIUC we only use about a third of our brain’s capacity and as AI develops so will our brains. So despite science fiction books and films I don’t feel threatened by AI. But I do feel threatened by the people.who control it.
“AI” has always been a bit of a misnomer but it may not be that far-fetched. I believe it has been used because of the emergent behavior that is observed sometimes (like the famous AlphaGo game where noone could quite understand one of the AI’s moves, it didn’t follow any human playstyle). The pattern recognition is somewhat of an emergent behavior as well (you really can’t predict what patterns or predictions are going to come out). And intelligence is etymologically “making links between”, so automatic pattern recognition is not completely out of place. It’s just not the same as conscience, self-determination, empathy or thought (which is generally what’s depicted as “AI” in popular culture / Sci-Fi).
You could have an Asperger kid with no emotions or empathy whatsoever, not even capable of speaking but able to solve extremely hard puzzles of a certain kind using his hands. Whether you call this “intelligence” or not is subject to interpretation (and philosophical debate).
gallois wrote:
But I do feel threatened by the people.who control it.
Or rather, feel threatened by all the blind followers. People inclined to follow “experts” and authority will for sure be inclined to follow AIs.
Info-Pilote had a short article also mentioning Airborne Again’s ECMWF link. To briefly summarise, the AI is surprisingly good at forecasting chaotic events, but is still limited in the time dimension (i.e. when the butterfly flaps its wings). As it works on its own with a lot of raw data, it’s not easy for a human to find the reasons for mistakes or the weighting given to particular data. They estimate 5-10 years before it’s fully useable by an end user. Graphcast is more accurate than Microsoft’s Pirate Weather, Huawei’s Pangu-Weather, or other start-ups.
https://charts.ecmwf.int/?query=Graphcast for charts generated this way (need to tick the options)
Google’s full article in Science https://www.science.org/stoken/author-tokens/ST-1550/full (I’ve not read this yet)
Peter wrote:
DWD is charged for, mostly?
Some specialized products are as well as their aviation weather platform “flugwetter.de”. But particularly flugwetter.de is well worth the money.
General meteorological products such as the quoted ones are mostly free.
Peter wrote:
And I don’t think it is any better than ECMWF, which btw is available for free on windy.com which is now the standard wx briefing tool.
there is a huge difference between automatized models and tools taking the data from them than to interpreted guidance products done by human beings. Obviously, whatever is computer output only is cheaper but imho not necessarily better. Not to talk of the fact that those who interpret sites like Windy very often take the models literally, which a trained meteorologist would never do.
gallois wrote:
I don’t feel threatened by AI. But I do feel threatened by the people.who control it.
And by people who follow machine outputs blindly without any interpretation.
I agree. AI is only as dangerous as what the people controlling it are doing with it. But the fact alone that there are billions of people who will eventually follow the output those machines generate controlled by a few “superbrains” as they think of themselves, make me very very much afraid.
But particularly flugwetter.de is well worth the money.
I don’t think so – example – but I can see why pilots in the relevant countries think so 
A huge amount of information dissemination/acceptance is “cultural”. I mean, there are probably UK pilots who think the UK MO is brilliant 
Not to talk of the fact that those who interpret sites like Windy very often take the models literally, which a trained meteorologist would never do.
If a human can do a better job, this can be represented in software.
I don’t think humans are better, simply because if Method X can be formulated / documented then it can be implemented in code. And if it can’t be documented then this “trained meteorologist” may as well be taking homeopathic medicine All his/her training can be documented, and prob100 has been. Tweaking the computer models based on that has been going on since computers were invented.
In this context, “AI” is just pattern recognition + machine learning. It works in many areas e.g. MRI interpretation.
