I'm not a climate scientists, but I am a programmer of the os's they run their simulations on, and have deep knowledge of how the processors work. I have also seen code used in modeling, and so I can say that it is likely that all climate models are flawed to the point of extreme inaccuracy even if the inputs were absolutely correct, and the the model is at a theoretical level correct. This is because processors take mathematical shortcuts that while IEEE compliant, are not actually suited for physics modelling. Even video games run into huge problems they have to hack around, and this with a fully digital simplified physics model.
I've written a lot of models and there isn't a good way to do this. The more parameters the more guesses they have to predict (with other models probably, heh) and the fewer things they take into account the shittier the model is.
What it comes down to is "my model shows an emergency" = lots of grant money; "my model shows it can't be shown" = no money. Just look at any predictive model and its error bars from the past and you will see how stupid their conclusions were. Either that or there was a miracle and everything was outside 2 sigma.
AI is, for the most part, a black box and takes in many many "parameters" that aren't seen by the user. You can still steer it, like tell it "this word is a no-no", but you can't easily access these parameters; you feed it the history as input and it will train on that and guess and continue to tune as time goes on. researchers can add more to the historical input as processing power increases and it will get better.
I'm guessing that there isn't an AI project that shows the end of the world unless the input is dicked with (which does happen, like the nasa data, i think).
I have yet to come across a fully competent tech who didn’t confirm this principle, nor an incompetent or even just run of the mill tech who didn’t deny it :p Thanks for taking the time to teach the details.
CO2 levels are not high. They were higher than this a million years ago. They were higher than this for most of the last 20 million years. If we have to cut energy usage to save the planet, why are govenments and corporations rolling out 5G? How much more information capacity will 5G bandwith have? Thats about how much more energy it will use.
[ + ] Osmanthus
[ - ] Osmanthus 4 points 2.8 yearsAug 25, 2022 18:22:09 ago (+4/-0)
This is because processors take mathematical shortcuts that while IEEE compliant, are not actually suited for physics modelling. Even video games run into huge problems they have to hack around, and this with a fully digital simplified physics model.
[ + ] GoldenAgeWhen
[ - ] GoldenAgeWhen 2 points 2.8 yearsAug 25, 2022 20:04:32 ago (+2/-0)
What it comes down to is "my model shows an emergency" = lots of grant money; "my model shows it can't be shown" = no money. Just look at any predictive model and its error bars from the past and you will see how stupid their conclusions were. Either that or there was a miracle and everything was outside 2 sigma.
[ + ] doginventer
[ - ] doginventer [op] 0 points 2.8 yearsAug 26, 2022 07:47:26 ago (+0/-0)
That thing about guessing versus making inspired leaps is what tells me that AI can never be what the satanist hope for.
I’m guessing of course :)
[ + ] GoldenAgeWhen
[ - ] GoldenAgeWhen 1 point 2.8 yearsAug 26, 2022 09:49:43 ago (+1/-0)
I'm guessing that there isn't an AI project that shows the end of the world unless the input is dicked with (which does happen, like the nasa data, i think).
[ + ] doginventer
[ - ] doginventer [op] 0 points 2.8 yearsAug 26, 2022 07:42:03 ago (+0/-0)
Thanks for taking the time to teach the details.
[ + ] PostWallHelena
[ - ] PostWallHelena 2 points 2.8 yearsAug 25, 2022 20:27:38 ago (+2/-0)
If we have to cut energy usage to save the planet, why are govenments and corporations rolling out 5G? How much more information capacity will 5G bandwith have? Thats about how much more energy it will use.
[ + ] doginventer
[ - ] doginventer [op] 0 points 2.8 yearsAug 26, 2022 07:27:19 ago (+0/-0)