I'm squeezing my papers so tight they are fused to my fingers!
@kraftykactus1028
3 жыл бұрын
This is the opposite of a problem!
@thezipcreator
3 жыл бұрын
I think you still had the old physics AI on, the new AI fixed paper-finger fusing
@mikejones-vd3fg
3 жыл бұрын
same, and i still dropped them all over the place in shock and amazement of the results
@TheDigitChannel
3 жыл бұрын
I'm squeezing my paper because someone is holding my pen
@owenf7824
3 жыл бұрын
What a time to be alive!
@okgoodgame
3 жыл бұрын
truly fascinating.
@genericwannabe
3 жыл бұрын
Should be a tshirt.
@Patrick-pu5di
3 жыл бұрын
livin in the future blingin on my hot line
@TheDigitChannel
3 жыл бұрын
You have to hold on to your paper first
@Forever._.curious..
3 жыл бұрын
Actually ♥️😁👍🏻
@oliver-beaudurivage2876
3 жыл бұрын
2 papers down the road the title will be, "Can a Machine teach us Physics?".
@Yeti4mad
3 жыл бұрын
Well that’s the goal. Make reality easy
@jacanchaplais8083
3 жыл бұрын
That's more or less the title of my PhD project, but I'm doing particle physics. Lots of amazing work from AI resesrchers happening in that area, but there aren't many pretty visualisations, so you probably won't see it turn up on KZitem.
@Adhil_parammel
3 жыл бұрын
@@jacanchaplais8083 share papper links
@Yeti4mad
3 жыл бұрын
@@jacanchaplais8083 Any recommendations I can google or just look up in general? I think that this is literally some of the coolest shit out there, AI learning physics and then us learning from it.
@ЭтоГеорг
3 жыл бұрын
@@jacanchaplais8083 is that in a similar vein to the work Taoli Cheng has been doing?
@ahmadisntcool8869
3 жыл бұрын
I couldn't be happier to see a new two minute physics video!
@CaseyHofland
3 жыл бұрын
I can’t wait to tell my grandkids that “back in my day, physics ran on the cpu and fluid sims could only do 3 frames a second.”
@darksunrise957
3 жыл бұрын
3 frames a second? What kind of supercomputer do you have? XD
@shin-ishikiri-no
3 жыл бұрын
Don't worry. You won't have grandchildren. Lorde KIaus Schwab of the 4IR shall not permit it.
@BambinaSaldana
2 жыл бұрын
@@shin-ishikiri-no What?
@thomasmaier7053
3 жыл бұрын
Is that viable for real time simulation? I would have liked some more info on the performance. Great video!
@anywallsocket
3 жыл бұрын
Yes! the whole point of compressing all the nuance into the training phase is that you don't have to run complicated calculations at runtime.
@elisgrahn6768
3 жыл бұрын
4:59 "And it can do it 10 to 100 times quicker."
@Soul-Burn
3 жыл бұрын
According to the paper (linked under the video), performance was between 11x to 289x faster when running on the GPU versus the standard simulator. Just note that this GPU is an NVidia V100, a $8000 card focusing on AI performance.
@MP-ri8ng
3 жыл бұрын
But whats the frametime? Is it already f/s or still s/f or m/f
@HarryPorpise
3 жыл бұрын
@@Soul-Burn seems in my budget
@thesteambreaker9449
3 жыл бұрын
Love the little "hold on to your papers" emblem 😂
@TheAlanmf
3 жыл бұрын
When I first met this channel I was like.... Hmmm lil bit odd but... wtvr. Now Im like "OMG YESSSSSS IM HOLDING MY PAPER TELL ME BOUT IT ALREADY" ahahaha
@volodymyr3169
3 жыл бұрын
Would be interesting to see computing time comparisons with standard methods
@AvastarBin
3 жыл бұрын
@@prumchhangsreng979 I think by computing time, he meant the time to predict those simulations. And going through the model to predict, while being much less costly than training the model, is still costly. And it's interesting to know how much computing power it needs to do that compared to traditional methods.
@prumchhangsreng979
3 жыл бұрын
@@AvastarBin ah so its the processing time. Im interested too xD
@ewerybody
3 жыл бұрын
Aren't processing and computing meant synonymously 😉 Well, I like the idea of offsetting simulation time to training time and having the predicted simulation to be faster. There were already some physics ML things that actually outperformed standard methods even reported here! I'd not be surprised if this perfoms well here but also would love to see the numbers. 🤓
@prumchhangsreng979
3 жыл бұрын
@@ewerybody i check dictionary and i mixed up the word compute with program(verb). :/ i was wrong hm
@hellfiresiayan
3 жыл бұрын
I think he said 10 - 100x faster. Which is quite the range and I agree more would be good to see in the video without having to follow through to read the paper
@claw320
3 жыл бұрын
Now this AI could literally "hold on to it's papers" since it has the physics to do so!
@robertwyatt3912
3 жыл бұрын
Cant wait until this sort of stuff is real-time
@martiddy
3 жыл бұрын
Two more papers down the line
@Gutagi
3 жыл бұрын
When it gets real time it will be the time to be alive!
@vladimirtomin8223
3 жыл бұрын
If it is 100x faster than simulation it should be faster than realtime already because C4D can run most of those simulations almost real time.
@aronseptianto8142
3 жыл бұрын
cloth simulation is already mostly real time
@robertwyatt3912
3 жыл бұрын
@@vladimirtomin8223 neat
@LorenzoValente
3 жыл бұрын
Can't wait to see these swaying cloths in videogames!
@LorenzoValente
3 жыл бұрын
@@Joe-nq6hy The paper says that this algo runs at 50fps on a (very) high end GPU so I guess it is totally possible :) and Unreal Engine 5 has now a clever solution for high polycount meshes, maybe it can help in these hard cases
@Benjamin_Gilbert-Lif
3 жыл бұрын
Had this idea months ago you could even create an entire physics engine built on this premise to make it run certain interactions faster
@francescocommisso5352
3 жыл бұрын
Same bro
@dentarthurdent42
3 жыл бұрын
Wow, just...wow. This is starting to feel like we're at the horizon of the singularity.
@unknown6656
3 жыл бұрын
I love how the "ground truth" flag is clipping through the pole and the AI is replicating that exact bug.
@KnakuanaRka
Жыл бұрын
Where does that happen?
@unknown6656
Жыл бұрын
@@KnakuanaRka You can see that for a couple of frames every time when the flag "waves out"/uncurls from the left to the right, e.g. at 2:45 or at 5:03. Use the keys `.` and `,` to step through the video frame-by-frame, if you need it :)
@boitahaki
3 жыл бұрын
"Can We Teach Physics To A Machine?" Yes, but the machine will probably ignore you.
@Zebred2001
3 жыл бұрын
What a time to simulate being alive!
@zenopeirce1836
3 жыл бұрын
the fluid simulations remind me of those pendulums with multiple hinges. The results start up the same, and they diverge over time. This might means that, even if the final results looks different, you can probably trust the neural network's conclusion.
@awe9217
3 жыл бұрын
i mean with that smooth voice of yours you could teach machines and humans alike anything.
@you_just
3 жыл бұрын
i know “what a time to be alive” has become something of a meme, but... seriously, what a time to be alive! future AI engineers would kill to experience the pioneering times we are!
@ErikVSV
3 жыл бұрын
I'm writing my master's thesis now on simplified models that run aeroelastic calculations on wind turbines, since CFD is generally too computationally expensive. It's wild to think this entire branch of work will be obsolete in a few years due to innovations like this.
@ricardasist
3 жыл бұрын
Cant wait till this gets implemented into scientific research and gives us some groundbreaking discoveries about our universe.
@cihadturhan
3 жыл бұрын
This is amazing! Btw, you forgot to talk about simulation duration with ai. Is it faster and if so, how much is it faster?
@Pigi0
3 жыл бұрын
will we (normla people) also be able to use this amazing technology? will we see it implemented in some 3d software soon?
@rustycobalt5072
3 жыл бұрын
Not that I have tried a lot, but nearly everything in machine learning that has merit is behind locked doors We will probably never see the programs shown in these videos, the code never released, and all features of them will be limited to the point they are almost no different to conventional programs One would think the world of AI is full of those willing to share their progress and mistakes, but nope not at all Money eventually is what decides what is and isn't produced, and by extent what we are "allowed" to know about those products
@CaptainPanick
3 жыл бұрын
@@rustycobalt5072 The code is often open source but the trained data results are not. I can understand that as it may take huge amounts of computing power plus expensive input data. But I agree with you, much of this stuff is behind pay wall's and outside of our reach. I for example would love to see this tech in software such as Blender and Unreal Engine but I doubt that is going to happen any time soon sadly.
@maythesciencebewithyou
3 жыл бұрын
@@rustycobalt5072 Learn how to do it and train your own model if you are too cheap to pay for someone elses work or if you are too lazy then wait a bit longer. The stuff that is shown here is cutting edge research.
@lolgamez9171
3 жыл бұрын
@@maythesciencebewithyou hmm yes fuck poor people
@aldiansyahwahfi
3 жыл бұрын
@@maythesciencebewithyou I've followed this channel for 2 years now but I still don't know what is this field of study called and where to learn it. Some directions would be very appreciated 😊😊
@-NGC-6302-
3 жыл бұрын
I know Károly’s voice about a thousand times better than I know his name Love the videos, keep it up Woah cool the channel is nearly at a million subscribers, wow!
@surajvkothari
3 жыл бұрын
This channel should be named Minute Papers.. like Minute Physics. That would explain this 7-min episode.
@LeSqueed
3 жыл бұрын
I wonder if machine learning could be implemented for light rays as well. Seeing how resource intensive ray tracing is currently. It wouldn't be perfect, just like these physics simulations. But if it can pass the eye test it is good enough for a ton of applications.
@TheDailyMonk
3 жыл бұрын
Would love to know the performance gains with this method
@satibel
3 жыл бұрын
10-100 faster than the training algorithms
@unvergebeneid
3 жыл бұрын
6:22 the flag just clips through the pole though. So I guess this would be useful for video games where accuracy isn't life or death.
@fahimzahir9587
3 жыл бұрын
This felt like a ASMR of Two Minute Papers.
@edzehoo
3 жыл бұрын
What i'm seeing is actually the building blocks for a true human-like AGI one day. We know when a flag is blowing unnaturally, or when honey doesn't drip with the right viscosity, simply by virtue of having learnt the physics of the real world through observation. And here's an AI that learnt and predicted it the exact same way!
@openroomxyz
3 жыл бұрын
I have a feeling that the concept of AI build accelerators build into the chip, will make more and more sense in future, with all this AI tech, begin to be used inside games, graphic app, and others, maybe we will have someday, a CPU, GPU, and a special AI Card xD in desktop computers and workstation. I am kinda surprised so many cool AI algorithm exist and they are so underused in consumer software in general.
@joshuawhitworth6456
3 жыл бұрын
I would like to see real life comparisons. All we have to to is use smoke so the a computer can learn the wind currents in a scene then add in a simulated cloth.
@anotherguycalledsmith
3 жыл бұрын
Dear Károly, Thank you very much! Your channel is always surprising to watch ;-) I do not know whether I saw it on your channel, but several years ago, there was a paper about a rigged character that was able of picking up a shirt (cloth simulation) and putting it on by itself. Do you happen to remember this paper? I cannot find it anymore. Thanks a lot ;-)
@PeterBarnes2
3 жыл бұрын
What I want to see is this mixed with adversarial machine learning. Train two more AIs: one that 'referees' by trying to guess which between two simulations is the ground truth; and the adversarial AI which takes as input a set of simulation pairs (AI generated and ground truth), and outputs a guess for the input parameters to a simulation that the referee will predict correctly. This adversarial AI will get more encouragement for simpler and smaller simulations. In this way, the adversarial AI should actually attempt to optimize for creating training data that is most efficient (at least in terms of challenge) for the amount of time spent running hand-crafted simulations. Obviously, we take these input parameters and run them through our handcrafted algorithm, then train our physics-simulating AI to accurately predict, up to some level of accuracy.
@the-old-channel
3 жыл бұрын
How long does it take for a trained AI to render these? How far are we from the real-time? Amazing video, by the way.
@tom9380
3 жыл бұрын
Let's be honest besides all the excitement: the flags are still quite a bit off from the actual simulated output - which can be a big deal when exact simulations are needed or in composite effect scenarios, which should have been mentioned in the video.
@constantinosschinas4503
3 жыл бұрын
*Demonstration of speed gains of the final AI result (after training)?* We just see similiar results (which is impressive) but no speed comparison, which is the essence of the whole concept: sacrifice some accuracy, for tons of speed.
@ReevansElectro
3 жыл бұрын
Shouldn't "Ground Truth" actually have a foundation in the world of reality rather than a simulation of reality?
@KnakuanaRka
3 жыл бұрын
Well, they’re trying to replicate the simulation with an AI, so the simulation would make sense as a ground truth. Working off real systems should come two papers down the line. ;-)
@No1TypeC
3 жыл бұрын
Ground Truth when referring to AI-accelerated structures can be assumed to refer to a (previous) non-AI established purely calculated result. Like a high sample ray traced "ground truth" image versus a low sample rate AI filtered one.
@justignoreme7725
3 жыл бұрын
Is this what Nvidia is doing with DLSS? Taking a simulation or game and then creating an AI that recreates via prediction? If not, has this being applied to game play? Or is it not fast enough yet?
@AvastarBin
3 жыл бұрын
I don't know if I'm the only one thinking about that but I can't wait to see the new video games with insane graphics while using the same computers that we have.
@musicmancer
3 жыл бұрын
I'd love to see this applied to movies and TV shows that take place "inside a video game". The problem I always see is that the game world is treated like real life, but the audience wants to see that it's fabricated. Maybe using a Physics-taught AI would strike the appropriate balance.
@cyber1714
3 жыл бұрын
0:30 this is the most used clip in all of two minute papers
@orik737
3 жыл бұрын
Lmao frrr
@Your_Friend_Corey
3 жыл бұрын
What about the one with all AI generated faces?
@sky173
3 жыл бұрын
Physics Rules!
@nickydeswart
3 жыл бұрын
This week I was on a boat. Looking down to the river and seeing the waterflow and the airbubbles underneath and all I could think was “what a perfect simulation, what a time to be alive”! (This is a true story)
@umurkaragoz
3 жыл бұрын
Only thing this NN can't predict could be the interaction between my hand and my papers while watching this video!
@gabrielarkangelo
3 жыл бұрын
GTA 7's early presentation of liquid physics.
@samwoodfield7332
3 жыл бұрын
Honestly with these algorithms and ML models now adays, I don't think it will be long till game engine animations, physics, graphics and characters are all generated by AI and use a fraction of computer resources
@TheNewton
3 жыл бұрын
Real time feedback > Real time simulation in most fields not concerned with safety. Being able to get an imperfect prediction is far more informative than an empty bounding box , mostly empty point cloud.
@timschafer2536
3 жыл бұрын
I would love to see a difference map of those simulation vs ground truth to see how accurate the algorithm really is because seeing just two images makes it hard to spot the errors.
@mikeyjohnson5888
3 жыл бұрын
There some strange effect where somehow the ground truth sims seemed far more uncanny than the ai prediction. Some of the ai predictions seemed to have more weight.
@zxa96
3 жыл бұрын
That's so crazy! Like the thing with FEM is that it's so hard to parallelize. A 16 core CPU is only like 2x the speed of a 4 core CPU. Running that on GPU clusters in tensor core is just so much more compute and would allow you to solve super problems you just couldn't feasible do with FEM.
@jackerylel
3 жыл бұрын
Apart from computer graphics, what are the applications of this? Would you ever trust this, for example in the stress test?
@matthew.wilson
3 жыл бұрын
I'd imagine being able to use it as an iterative design confirmation step, for one. It's already fairly common practice to have multiple levels of verification in Systems Engineering, leaving the final test with a physical product (an airliner for example) to the very last, after many tests with simulators at varying lower levels of fidelity. This technique could perhaps be a version of that for Mechanical Engineering, among others.
@adamtaylor2142
3 жыл бұрын
Aha! Finally, a paper I have read before seeing your video! I am strangely proud.
@dominicisthe1
3 жыл бұрын
Hmmm this paper is very reminiscent of the work coming out of caltech regarding their graph neural operators
@scionax541
3 жыл бұрын
I had to watch this twice to wrap my head around this. This is absolutely insane. The time it would take for hardware to catch up to some of these legacy simulations would be many years from now. This makes many real-time physics possible in simulations now (or at least, once the algorithms become more available). Crazy.
@Robert_McGarry_Poems
3 жыл бұрын
This is getting really good!
@olemuell5979
3 жыл бұрын
i am a bit sad you did not go at all into the dynamics of how they are training the networks. Are these the famous PINN? (Physics informed neural networks) I am working with PINN for PDE modeling. And i have to say they are amazing. They are basically unsupervised and just learn from the residual of an inheren PDE and the discrepancy of their own output. I would go as far as saying PINN are to FEM what injection molding is to additive manufacturing.
@newtonbomb
3 жыл бұрын
How tf can it so accurately perform a simulation on a cylindrical flag when it's just trained on a small piece of rectangular fabric?! And the different wind directions without ever being trained with those parameters?! It boggles the mind...
@anywallsocket
3 жыл бұрын
you might as well ask that about rectangular and cylindrical flags in real life: they don't react to the wind differently, the difference is in their geometry which reacts to itself and the wind the same.
@newtonbomb
3 жыл бұрын
@@anywallsocket Considered as whole objects their deformations are different even if individual units of fabric making them up may react generally the same. A rectangular flag not "inflating" is an example, but I guess if the ML algorithm is just extrapolating the simulated objects as compounded mesh units it could probably arrive at that naturally.
@jacobheglund4245
3 жыл бұрын
Thanks for the great video Two Minute Papers! It's really cool to see that the same neural network models that I'm researching as part of my PhD (graph neural networks) can be used so effectively for faster physics simulations. What a time to be alive!
@StephenRayner
3 жыл бұрын
Thank you for the amazing videos, having a crap day and these make me feel better!
@jukio02
3 жыл бұрын
If we ever want to create true VR, we will have to get the physics down.
@anntony5585
3 жыл бұрын
I am clutching my papers
@AIpha7387
3 жыл бұрын
I am looking forward to the AI process that takes visual/auditory data from real video and converts them into simulation data. This will allow us to automatically scan the real world and treat it as manageable data.
@AIpha7387
3 жыл бұрын
In addition, it would be awesome if that could recognize each element and give some change with a language directive, as in the example shown with Codex.
@aaronlewis4475
3 жыл бұрын
I have become one with my papers. My soul and my papers are fused to become a single entity. I have achieved immortality through my union with the divine paper.
@nathaos01
3 жыл бұрын
How many FPS or SPF does the ai send the output?
@z-beeblebrox
3 жыл бұрын
You know how in the Matrix sequels, we learn that each physics system in the simulation is run by its own AI? It seemed like a silly concept at the time...
@Fermion.
3 жыл бұрын
Originally, the writers and director wanted humans to be used as a living neural network, but the studio thought that was too complex for the public, so they dumbed it down to living batteries. Which is stupid, because the human body isn't an efficient battery at all. It would take more power to keep us alive, than we could produce.
@parallaxdawn2546
3 жыл бұрын
After seeing all of your other videos, Yes.
@originalsingh
3 жыл бұрын
It feels like a superluminal AI is indeed simulating the universe
@pariscatblue
3 жыл бұрын
Thanks a lot, I'm not a big fun of your visualisation videos but this one, FANTASTIC!!! and let's read a paper! :-)
@Subs1338
3 жыл бұрын
Holy shit imagine VR in 10 years, we will literally have seperate lives on there.
@koko969w
3 жыл бұрын
need a simulation of a hand holding onto papers
@supe4701
3 жыл бұрын
For a second I thought you said psychics
@multigamerx1710
3 жыл бұрын
wow!!!!
@voodoo5191
3 жыл бұрын
I wonder who's gonna be the heartwarming message of the day.
@Sancarn
3 жыл бұрын
Something very important that is rarely mentioned, but should be, is that AI might lead to faster simulations, but the results will always be different from the ground truth. This is fine if you're wanting to use AI for purely graphical problems, and maybe coarse approximations. But to truly test solutions to real world problems, physics simulations will be required. E.G. Your AI might state that a bridge will hold up, but you will want to prove that in a proper simulation before you build a bridge humans will be walking on. If you use AI, and the bridge breaks, it is your responsibility for relying on an unproven black box to make real-world decisions.
@mexicanmax227
3 жыл бұрын
I can sense your passion, it’s beautiful lol. This is so fascinating! Pretty sure if I was left alone one weekend I would be glued to your channel and obsessively watch all your content! XxDD
@KORKEL-
3 жыл бұрын
why isnt this type of technology used in modern game engines?
@jameshughes3014
3 жыл бұрын
AI is used extensively in games now. That's what DLSS is. It's how RTX works without being super noisy. And AI driven techniques are used extensively in the graphics apps now that are used to make games and other things. I'm sure that as they develop cool new things like this, it's being implemented as quickly as possible, but it is still new. It takes time to turn a paper into an app. I think the reason this particular thing isn't yet implemented is because it just would take too much computer power. To do this probably requires the entire GPU, which would leave nothing left over for your game, however if used by artists to help create the content for games, it is a game changer.
@edward.chyrek
3 жыл бұрын
Hold on to your papers! What a time to be alive!
@kkp1023
3 жыл бұрын
I am excited for the videogames that'll be available to us like after a decade with this tech
@dreamingpenguin6140
3 жыл бұрын
i really want to see these graphics papers applied to video games
@Fanny-Fanny
3 жыл бұрын
I am come here see you so early, it make me come here early. Thanks!
@rottenpoet6675
3 жыл бұрын
getting closer to our ancestor simulation
@MatthewFearnley
3 жыл бұрын
3:01 AI is now able to simulate the alternative timeline from Day of the Tentacle, where the American flag is swapped for a Tentacle costume.
@theempireofthepeople
3 жыл бұрын
how close is this to be in real time?!!!
@xanderwilmot5740
3 жыл бұрын
6:05 OKAy, come on. it sounds like he turned around for a moment, and his kid went STRAIGHT back to drawing on the walls. he's just so fed up with this AI having zero limitations.
@xl000
3 жыл бұрын
Soon the algorithms will start writing papers and ask to be paid in BTC. Or they will start writing exploits, attack nuclear facilities and ask for a ransom. Not sure what a program would do with BTC though. That's the scary part.
@ryanbigguy
3 жыл бұрын
I think the AI generated simulations actually look better in many cases!
@spider853
3 жыл бұрын
To be fair small domain simulations like cloth and fluids are not so amazing for AI, they just approach the influence kernel and forces to be applied per particle + neighbours, I bet you still need to feed this information to the AI as I don't think the input is all the vertices/particles, as you can't easly expand the resolution there.
@vadiks20032
3 жыл бұрын
i just realized if humanity makes an AI so smart which would take over teh world, it'd take the A.I. more than 20 years to do the math until it could improve itself. and it'd probably fail and die at some point
@solotronixTV
3 жыл бұрын
Soon we will have virtual worlds where we can live another life, while we sell our bodies energy to a server that requires power WHILE WE PLAY SUPER MAN AND ARE FED THROUGH TUBES..... The Matrix in the making...
@guilhermetorresj
3 жыл бұрын
Engineering be like: lemme find some solutions to the Navier-Stokes equations that describes the motion of fluid in this particular situation. This paper: HOLD MY BEER.
@MortenSlottHansen
3 жыл бұрын
Your enthusiasm is epic as always - loving it 🙂
@hellalittt
2 жыл бұрын
Let's make an AI that studies our brain and make theories of how it work, so we can build an AGI.
@ManiacallySmithing
3 жыл бұрын
I understand you're posting amazing things and I've followed the channel for a couple of years... I don't get it though. I can never fully understand what the problem is in the beginning, who implemented it, what "our solution" means most of the time, especially when two or more superceding methods are interjected... For instance, on this video, what is the simulator? What is the AI? Is the AI improving the simulator? OR is it a standalone solution? Maybe the AI was trained by the simulator? Is the simulator not AI-based itself? What about "prediction" and "ground truth" ? What are those? Are we fixing them? The videos play simultaneously, often without the difference between two implementations being apparent. Confusion, confusion, confusion...
@patarciofo7538
3 жыл бұрын
I'm curious if this amazing technology could be applied to videogames with Nvidia Tensor Cores of the RTX cards
@genericwannabe
3 жыл бұрын
“10-100x quicker” should be said as “1/10th to 1/100th as long”, right? It feels like saying “this bowl has 10x fewer marbles as that one” when it really means “this bowl has 1/10th the marbles of that one.”
@shahinrab
3 жыл бұрын
As much as I am rooting for works like this and this channel, it is a bit of a cringing moment for me when I see Dr. Fehér oversells what a great paper has achieved. Generalization has two meanings: interpolation and extrapolation. A vast majority (not all) of these physics-aware ML methods are memorizers; meaning you show a limited set of examples and ask your model to guess what it has not seen, as long as the underlying characteristics of the simulation remain unchanged. This is the generalized interpolation problem. Seeing enough examples the learned manifold is dense and "smooth" enough to be able to fill in the blanks. This is in fact not something new. Look up "Koopman Theory" which has been applied in data-driven fluids simulation for years (e.g. SANDY) method. In fact, if anything, the Koopman operators are more than just interpolators. Their subspace is in fact closer to the actual embedding of the laws of physics. The problem is extrapolation, generalization beyond the interpolated learned space. Example: take the flying flag and change its stiffness and damping. That pushes the dynamics of the simulation to a whole new domain that has not been seen. In this sense of generalization, no, these ML models have not actually "learned" the laws of physics. Just a fraction of the whole phase space described by the mathematical model. What I said does not lessen the value of this work even a bit. But I believe these videos should be made a bit more carefully so as not to throw around such claims as learning the laws of physics, which is, indeed, a great subject of debate. Specially when the main body of the audience consists of scholars. Here is the last keyword for you: Kolmogorov complexity... there is a fundamental limit to how much you can compress an equation without breaking it. If we go over that limit, that means approximation, which might work in certain cases but not in others. An approximation of a law of physics is still useful, but it does not equate to it.
@sumsum9017
3 жыл бұрын
I wonder what's the execution time per frame? Can it run in real-time?
@wundermax1993
3 жыл бұрын
I want to see some molecular or better yet quantum physics simulated. Imagine, we could have proper proton torpedoes in the new star wars games! Mikor jön a mezonágyú doktor úr? :)
@lucaschan756
3 жыл бұрын
As a computational physicist I have to say this is impressive. I was wondering if the model is trained on low viscosity, will it be able to perform simulation with high viscosity in the future? Or, does the model support changing physical parameters (gravity, viscosity, temperature etc.) at all? Thanks.
Пікірлер: 351