/sci/ - Science & Math » Thread #12292709

858KiB, 600x324, ns_sr_v1e-4_labelled.webm

View Same Google iqdb SauceNAO

Anonymous Sat 31 Oct 16:24:39 2020 No.12292709 View Reply Original Report

Quoted By: >>12292741 >>12292756 >>12292783 >>12292919 >>12293035 >>12293283 >>12293494 >>12293839 >>12294744 >>12294763 >>12294844

AI has solved fluid mechanics

https://www.technologyreview.com/2020/10/30/1011435/ai-fourier-neural-network-cracks-navier-stokes-and-partial-differential-equations/

Anonymous

Anonymous Sat 31 Oct 2020 16:34:26 No.12292737 Report

Quoted By:

>AI
this is really russling my jimmies b/c it looks like nickname for Allen or Albert and/or the initialization of Allen Iverson.
These nerds don't know about Pierre Woodman's series for Hustler called "Anal Intensive".
fartificial dimtelligence can gtfo with this abbreviation, that shit is taken.

Anonymous

Anonymous Sat 31 Oct 2020 16:34:56 No.12292741 Report

Quoted By: >>12292747 >>12292798 >>12293770

>>12292709
>One research topic Anandkumar is particularly excited about: climate change

Into the trash it goes.

Anonymous

Anonymous Sat 31 Oct 2020 16:38:51 No.12292747 Report

Quoted By: >>12292798 >>12294892

>>12292741
retard

Anonymous

Anonymous Sat 31 Oct 2020 16:40:48 No.12292756 Report

Quoted By:

>>12292709
Good luck trying to find a single serious company willing to model high performance vehicles or planes with that lossy approach lmao

Anonymous

Anonymous Sat 31 Oct 2020 16:52:55 No.12292783 Report

Quoted By: >>12293059

>>12292709
>AI has solved fluid mechanics
Lol, didn't solve shit. We knew how to do this for last 50 years.

Anonymous

Anonymous Sat 31 Oct 2020 16:56:31 No.12292798 Report

Quoted By: >>12293770

>>12292741
For me it was the non white name

>>12292747
Go back to plebbit

Anonymous

Anonymous Sat 31 Oct 2020 17:01:53 No.12292810 Report

Quoted By: >>12292986 >>12294866

This article is MIT blowing smoke out their ass, but I've worked with companies that have been using deep learning for physics simulations for several years now. You can get equivalent accuracy to numerical methods with ~5000x speedup. Once you have a trained model, inference time is super fast.

Anonymous

Anonymous Sat 31 Oct 2020 17:02:51 No.12292814 Report

Quoted By: >>12292910

>some literally who rediscovers fourier analysis and numerical methods
we did it reddit

BIG WOKE

View Same Google iqdb SauceNAO 1604075519138.png, 314KiB, 1102x958

BIG WOKE Sat 31 Oct 2020 17:40:44 No.12292910 Report

Quoted By:

>>12292814
thr keyword is navier stokes. Give me a explanation how navier stokes work if you mind.

Anonymous

Anonymous Sat 31 Oct 2020 17:42:42 No.12292919 Report

Quoted By:

>>12292709
Wait isn't this a millennium prize problem?

Anonymous

Anonymous Sat 31 Oct 2020 18:02:48 No.12292986 Report

Quoted By: >>12293745

>>12292810
What about proofs of convergence and consistency ?

Anonymous

Anonymous Sat 31 Oct 2020 18:11:31 No.12293035 Report

Quoted By: >>12293070

>>12292709
The amount of cope in this thread, we get it you guys have never done anything of use so you get mad when people do cool shit.

Anonymous

Anonymous Sat 31 Oct 2020 18:18:39 No.12293059 Report

Quoted By:

>>12292783
This.
People want a solution to the existance and smoothness problem, here you have some j*rnalist claiming an AI cracked the Navier-Stokes equations when in reality all they did was make a decent PDE solver and use an example of solving one case with the solver. It's like saying you figured out how to travel at 100% the speed of light and then using a photon as an example of this light-speed travel.

Anonymous

Anonymous Sat 31 Oct 2020 18:21:58 No.12293070 Report

Quoted By: >>12293084

>>12293035
It's great that they got something that works faster then the things that we have. But
1. In principle it was know that this would happen
2. It didn't SOLVE the Navier Stokes problem, there is nothing in there about blow up not occurring /occurring in finite time, for any initial conditions.

Anonymous

Anonymous Sat 31 Oct 2020 18:26:09 No.12293084 Report

Quoted By: >>12293099 >>12293107

>>12293070
>In principle it was know that this would happen
Knew what would happen?
>It didn't SOLVE the Navier Stokes problem
Only the faggot OP is claiming that. Not even the article is claiming the Navier Strokes problem was solved? It's cool that they found a faster, more accurate, and more generalizable deep learning technique to solve PDE's and everyone in this thread is being a faggot about it.

Anonymous

Anonymous Sat 31 Oct 2020 18:29:40 No.12293099 Report

Quoted By: >>12293124

>>12293084
Maybe not known, but expected that faster algorithms were possible
>Only the faggot OP is claiming that. Not even the article is claiming the Navier Strokes problem was solved?
>i-fourier-neural-network-cracks-navier-stokes-and-partial-differential-equations
Ok there.

Anonymous

Anonymous Sat 31 Oct 2020 18:31:03 No.12293107 Report

Quoted By: >>12293130

>>12293084
>Not even the article is claiming the Navier Strokes problem was solved?
>"AI has solved fluid mechanics"
So if it didn't solve the Navier Stokes problem then what part of fluid dynamics did it solve? Because that is quite a bold claim that is made by OP but not really made by the researchers themselves.

Anonymous

Anonymous Sat 31 Oct 2020 18:35:49 No.12293124 Report

Quoted By:

>>12293099
Oh my god you didn't actually read the article pass the title you faggot. Actually kill yourself, nowhere in the article does it claim that it solves the Navier Strokes Problem, it says that it's faster and more accurate then current methods at approximating PDE's. The title of the article is "AI has cracked a key mathematical puzzle for understanding our world." you didn't even open the fucking Article. Saying AI has cracked a key to understanding PDE's is not the same as saying they solved the Navier Strokes equations.

Anonymous

Anonymous Sat 31 Oct 2020 18:36:51 No.12293130 Report

Quoted By: >>12293152 >>12293209

>>12293107
>"AI has solved fluid mechanics"
Where in the article does it say this?
>then what part of fluid dynamics did it solve?
Where in the article does it claim this?

Anonymous

Anonymous Sat 31 Oct 2020 18:42:07 No.12293152 Report

Quoted By: >>12293439

>>12293130
Pajeet, I'm happy that you posted an article on arxiv, but don't expect that everybody will suck your dick over that.

Anonymous

Anonymous Sat 31 Oct 2020 18:59:43 No.12293209 Report

Quoted By: >>12293439

>>12293130
>OP claims X thing
>people call OP out on it
>"uh no sweaty YOU are the one talking about X, i didn't say ANYTHING about that"
Back to the shitting street with you

Anonymous

Anonymous Sat 31 Oct 2020 19:21:41 No.12293262 Report

Quoted By: >>12293285

>But this time, the researchers decided to define the inputs and outputs in Fourier space, which is a special type of graph for plotting wave frequencies.
I cringed so hard my butthole imploded.

Anonymous

Anonymous Sat 31 Oct 2020 19:30:50 No.12293283 Report

Quoted By:

>>12292709
Navier and Stokes are surnames of people who 'solved' those equations you fucking mongoloid and its not reasoning its fucking trained on imperfect data to make more imperfect data faster i hate you go back to fucking tiktok

Anonymous

Anonymous Sat 31 Oct 2020 19:31:53 No.12293285 Report

Quoted By:

>>12293262
Im going to explode dont do this to me

Anonymous

Anonymous Sat 31 Oct 2020 19:36:23 No.12293293 Report

Quoted By:

>Have hard math problem
>Use statistical shit that kinda sounds right
>Make some experiment to see if it kinda works
>Solve mathe problem
Do retards don't understand what math is anymore?

Anonymous

Anonymous Sat 31 Oct 2020 20:35:37 No.12293439 Report

Quoted By: >>12293454

>>12293152
>I'm a Pajeet cause I have reading comprehension.
Ok bro whatever you say.
>>12293209
>People call OP out on it.
I'm sorry, are we in the same thread? The entire thread was shitting on the article and the research?
Things like
>This article is MIT blowing smoke out their ass
>>some literally who rediscovers fourier analysis and numerical methods
What do these have to do with the OP? Keep on being a faggot instead of admitting you where wrong and didn't read the article thought. I can't believe you have the audacity to claim that I'm mad because people where calling out the OP.

Anonymous

Anonymous Sat 31 Oct 2020 20:41:27 No.12293454 Report

Quoted By:

>>12293439
Not anyone of those, but it's because OP claimed that they "solved" Navier-Stokes. Both him and the author of that article are obvious brainlets and deserve ridicule.

Anonymous

Anonymous Sat 31 Oct 2020 20:41:49 No.12293456 Report

Quoted By:

>OP solves AI
>nobody cares

Anonymous

Anonymous Sat 31 Oct 2020 20:50:28 No.12293494 Report

Quoted By: >>12293725

>>12292709
>AI has solved fluid mechanics
Some if-statements solved fluid mechanics by brute forcing through every possibility*
literally the most boring solutions, all human proofs/solutions are always very witty, smart, creative and they make you think

Anonymous

Anonymous Sat 31 Oct 2020 22:14:31 No.12293725 Report

Quoted By:

>>12293494
It's what you get when you have retards refer to machine learning as artificial intelligence trying to be fancy: it creates another generation of retards even dumber than them that think technology is magic and that scientists are wizards. The second generation at least has an excuse: they didn't know any better and were taught it wrong. The first generation, however, does not.

Anonymous

Anonymous Sat 31 Oct 2020 22:22:05 No.12293745 Report

Quoted By:

>>12292986
This is cs now brah

Anonymous

Anonymous Sat 31 Oct 2020 22:34:58 No.12293770 Report

Quoted By: >>12293833

>>12292798
>>12292741
To be fair, only non whites seems to be working on any sort of neuralnet works. All the papers on NN are published by Zhangs or Pajeets. Literally. Whites just dont have the capacity to do NN.

Anonymous

Anonymous Sat 31 Oct 2020 23:01:35 No.12293833 Report

Quoted By: >>12293849 >>12293867

>>12293770
>All the papers on NN are published by Zhangs or Pajeets
Isn't that what you would expect considering population differences? Or has the lack of success in virtually every single thing given the sick man of Asia and the streetshitters an inferiority complex?

Anonymous

Anonymous Sat 31 Oct 2020 23:03:52 No.12293839 Report

Quoted By:

>>12292709
There are well known equations for fluid dynamics, don't need AI

Anonymous

Anonymous Sat 31 Oct 2020 23:10:06 No.12293849 Report

Quoted By: >>12293869

>>12293833
>Isn't that what you would expect considering population differences?
No. Whites are living in much better developed situations. Changs/Pajeets are living in shitting streets. So one would expect the opposite, otherwise we'd see latin America/african papers as well since they're similar in numbers.

Anonymous

Anonymous Sat 31 Oct 2020 23:15:15 No.12293867 Report

Quoted By:

>>12293833
They only improve a bit. The big names in neural networks are white males cope harder

Anonymous

Anonymous Sat 31 Oct 2020 23:15:38 No.12293869 Report

Quoted By:

>>12293849
>"b-but muh money"
China has the second most millionnaires in the world, dwarfed by only one singular country: the United States of America. Or at least so they like to market themselves. So you would expect them to outperform most other countries in terms of research, certainly more than the UK which not only has less millionnaires but also has a far lower population.

Anonymous

Anonymous Sun 01 Nov 2020 05:52:07 No.12294744 Report

Quoted By:

>>12292709

Oh wow and it looks just like optimum theory. You fucking losers

Anonymous

Anonymous Sun 01 Nov 2020 05:12:27 No.12294763 Report

Quoted By:

>>12292709
Clickbait title

Anonymous

Anonymous Sun 01 Nov 2020 07:03:56 No.12294835 Report

Quoted By:

If you see a race baiting post, don’t engage. This thread has already slid into some off topic race issue just like every other board.

I get some of it. Not talking about all the racist ideas you were raised with and built your identity on is like a pressure cooker. Then you start seething once you get to post about it without consequence of how much of a brainlet idea it is

everyone who I’ve discussed with about this advancement are very excited. You can do it with traditional computational physics but the compute time is so large that one model output might take 6 hours. With this new method you can have the same accuracy with an output in near real time.

Anonymous

View Same Google iqdb SauceNAO 1600976399247.png, 818KiB, 1080x1069

Anonymous Sun 01 Nov 2020 07:08:44 No.12294844 Report

Quoted By:

>>12292709

looks like just another numerical solver, not an analytical solver

Anonymous

Anonymous Sun 01 Nov 2020 07:19:37 No.12294866 Report

Quoted By:

>>12292810

sure, most video games approximates naiver stoke in real time

Anonymous

Anonymous Sun 01 Nov 2020 07:28:13 No.12294892 Report

Quoted By:

>>12292747
This isn't plebbit. Go back there if you want to spout pseudoscience.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches