Jump to content
  • Member Statistics

    17,607
    Total Members
    7,904
    Most Online
    Chimoss
    Newest Member
    Chimoss
    Joined

Dec 5-6th clipper threat


George001
 Share

Recommended Posts

21 minutes ago, RDRY said:

I find this to be nothing but voodoo and actually its kind of gross because there are people being taken advantage of for this type of stuff, particularly if they're paying a great deal of money. 

There have been a few articles on the "success" of these AI models. For example, that one tropical system last year that went into NS and there was the article that said the google AI nailed that 10 days out....

Well what do they mean by nailing something 10 or 15 days out? That what the product showed for D10 and D15 verified? But how did it do on D14, 13, 12, 11, ..., 3,2,1? Was it extremely consistent or did it waver back-and-forth and have a million different solutions? If that is the case then GARBAGE. That's no different than what models currently do. 

The idea of being able to process data much more quickly is phenomenal but if I understand correctly, these AI models are not run with equations and physics...it just looks for pattern similarities. I mean pattern recognition is great but humans can do that too and while the weather pattern is important its how the pieces move and evolve which is more important and there is zero chance in hell AI will ever figure that out because this goes beyond "patterns". 

 

  • Like 2
Link to comment
Share on other sites

6 minutes ago, weatherwiz said:

I find this to be nothing but voodoo and actually its kind of gross because there are people being taken advantage of for this type of stuff, particularly if they're paying a great deal of money. 

There have been a few articles on the "success" of these AI models. For example, that one tropical system last year that went into NS and there was the article that said the google AI nailed that 10 days out....

Well what do they mean by nailing something 10 or 15 days out? That what the product showed for D10 and D15 verified? But how did it do on D14, 13, 12, 11, ..., 3,2,1? Was it extremely consistent or did it waver back-and-forth and have a million different solutions? If that is the case then GARBAGE. That's no different than what models currently do. 

The idea of being able to process data much more quickly is phenomenal but if I understand correctly, these AI models are not run with equations and physics...it just looks for pattern similarities. I mean pattern recognition is great but humans can do that too and while the weather pattern is important its how the pieces move and evolve which is more important and there is zero chance in hell AI will ever figure that out because this goes beyond "patterns". 

 

I know from your past posts that this is your default knee-jerk reaction to any "AI" related weather posts, but you should really read the scientific paper before declaring this as just "GARBAGE".  You're a smart guy, take the time to read and learn before making proclamations.

If you took that time you'd know that this statement of yours; "The idea of being able to process data much more quickly is phenomenal but if I understand correctly, these AI models are not run with equations and physics...it just looks for pattern similarities", is wrong for this particular AI model.  It is a hybrid approach.

Link to comment
Share on other sites

3 hours ago, Damage In Tolland said:

 

14 hours ago, powderfreak said:

Big lull now after the initial push of WAA.

~2” at the ski area cam and now dry slot flurries.

IMG_1621.jpeg.3da470214053b7a19865a0035a13bf37.jpeg

 

12 hours ago, mreaves said:

Did you go to the grand opening of the new bridge the other day?

So Kevin comes up to Vermont and drives past all of our places.  The storm comes and we all get 2 inches or less and Tolland gets dumped on.   Only conclusion is Kevin stole our snow.

 

Congrats.

  • Haha 6
  • 100% 1
  • clap 1
Link to comment
Share on other sites

33 minutes ago, metagraphica said:

I know from your past posts that this is your default knee-jerk reaction to any "AI" related weather posts, but you should really read the scientific paper before declaring this as just "GARBAGE".  You're a smart guy, take the time to read and learn before making proclamations.

If you took that time you'd know that this statement of yours; "The idea of being able to process data much more quickly is phenomenal but if I understand correctly, these AI models are not run with equations and physics...it just looks for pattern similarities", is wrong for this particular AI model.  It is a hybrid approach.

I find this topic very interesting so I created a thread hoping that a discussion can occur sans whining and sobbing about current weather:

 

  • Like 1
Link to comment
Share on other sites

3 minutes ago, klw said:

 

 

So Kevin comes up to Vermont and drives past all of our places.  The storm comes and we all get 2 inches or less and Tolland gets dumped on.   Only conclusion is Kevin stole our snow.

 

Congrats.

Hey Jay Peak reported 10-14" overnight... not sure how, a friend says it skis like maybe 5" new, but maybe someone got dumped on?

Link to comment
Share on other sites

26 minutes ago, metagraphica said:

I know from your past posts that this is your default knee-jerk reaction to any "AI" related weather posts, but you should really read the scientific paper before declaring this as just "GARBAGE".  You're a smart guy, take the time to read and learn before making proclamations.

If you took that time you'd know that this statement of yours; "The idea of being able to process data much more quickly is phenomenal but if I understand correctly, these AI models are not run with equations and physics...it just looks for pattern similarities", is wrong for this particular AI model.  It is a hybrid approach.

I do want to invest some time to find any published, peer-reviewed papers which provide highly detailed analysis but IMO, much of this stuff is hyped up. All you hear about is a success story and within these success stories they fail to tell you about the non-successes. For example, if 10 storms were tracked and analyzed and the model nailed 1 of the 10...the story will be ran on that 1 and you'll get a title like you see here, "AI Weather Model masters 15-day Forecast"...but there will be nothing on the other 9. 

AI will probably help drastically improve short-term forecasting during extreme weather events, such as the flooding event that occurred in Connecticut back in August but I don't buy it will ever become more accurate overall with increasing time stamps. Weather, physics, chemical processes are all way too complex and I'm sorry but AI is not going to ever figure that out because there is way more to this than just what happened historically. That atmosphere is extremely fluid and there are influences that we still don't know about or fully understand and a computer or algorithm isn't just going to magically figure that out.

AI still struggles with generated images of people and objects...so we're to expect it will figure out the atmosphere?

Link to comment
Share on other sites

45 minutes ago, weatherwiz said:

I find this to be nothing but voodoo and actually its kind of gross because there are people being taken advantage of for this type of stuff, particularly if they're paying a great deal of money. 

There have been a few articles on the "success" of these AI models. For example, that one tropical system last year that went into NS and there was the article that said the google AI nailed that 10 days out....

Well what do they mean by nailing something 10 or 15 days out? That what the product showed for D10 and D15 verified? But how did it do on D14, 13, 12, 11, ..., 3,2,1? Was it extremely consistent or did it waver back-and-forth and have a million different solutions? If that is the case then GARBAGE. That's no different than what models currently do. 

The idea of being able to process data much more quickly is phenomenal but if I understand correctly, these AI models are not run with equations and physics...it just looks for pattern similarities. I mean pattern recognition is great but humans can do that too and while the weather pattern is important its how the pieces move and evolve which is more important and there is zero chance in hell AI will ever figure that out because this goes beyond "patterns". 

 

From Nature: https://www.nature.com/articles/s41586-024-08252-9#Fig3

 

“As shown in the scorecard of Fig. 3, the forecasts of GenCast are significantly more skilful (P < 0.05) than that of ENS on 97.2% of our 1,320 variable, lead time and vertical level combinations (and 99.6% of targets at lead times greater than 36 h)“

Seems like a significant improvement on its face, would leave the more technical analysis of the graphic to those more knowledgeable 

IMG_6085.webp

  • Like 1
Link to comment
Share on other sites

4 minutes ago, Henry&#x27;s Weather said:

From Nature: https://www.nature.com/articles/s41586-024-08252-9#Fig3

 

“As shown in the scorecard of Fig. 3, the forecasts of GenCast are significantly more skilful (P < 0.05) than that of ENS on 97.2% of our 1,320 variable, lead time and vertical level combinations (and 99.6% of targets at lead times greater than 36 h)“

Seems like a significant improvement on its face, would leave the more technical analysis of the graphic to those more knowledgeable 

IMG_6085.webp

So it has greater skill with 2m temperature, sea-level pressure, and wind? That isn't exactly mind blowing, IMO. 

Link to comment
Share on other sites

Want to clarify...I was calling the claim garbage, not the model or the work into it. If a claim is going to be made that a weather event was nailed 10 days our or 15 days out...at least indicate the model held true to that solution for each run leading up to that day. If google AI had a category 3 hitting SNE at D15 and it ended up occurring...if between it showed solutions ranging from OTS or into the mid-Atlantic, or a weak storm...well you can't say it "nailed it". That's what I am calling garbage

Link to comment
Share on other sites

7 minutes ago, weatherwiz said:

So it has greater skill with 2m temperature, sea-level pressure, and wind? That isn't exactly mind blowing, IMO. 

Seems like it’s worse at vorticity but better at geopotential heights. Worth noting also that 20% increase in temp forecast accuracy three days out should have sizable implications for the energy industry, no?

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...