3 ms·

It is hard to tell what is AI about it? It seems like GPU accelerated fluid simulation that is relatively low resolution. From the description it appears they

It is hard to tell what is AI about it? It seems like GPU accelerated fluid simulation that is relatively low resolution. From the description it appears they use AI (which they say is a neural network representation of signed distance fields) for colliders only? The performance seems good but given it is low resolution it is hard to tell if it is better than the alternatives. I wish there was a Siggraph paper on this technique.

albertzeyer 4y ago
> We use neural network ...
- snickmy 4y ago
  it was unclear to me what is the NN used for in this case?
jussipetteri 4y ago
It is supposed to be 15x faster than the alternatives. And based on my testing that figure might not be far off.
- bhouston 4y ago
  I suspect they are running at 1/3 the resolution and upscaling maybe? So the simulation is 27x faster roughly but there is an upscaling cost that brings it down to 15x?
raincole 4y ago
It's not clear to me either. But I guess it's something like a denoiser: you simulate it at a very low resolution and use AI to "upscale" it?
- Culonavirus 4y ago
  That's what it looks like. Still impressive. Even though personally I think we've had realtime fluid sims in games before and they always fail to make a splash by not being tied to gameplay properly. I think CSGO2's way of handling fluids (or at least smoke) is the way forward. Make it more than damn flipbooks, integrate it into gameplay, the rest will take care of itself in time. It just needs a spark. CSGO2 may be that spark.
  - bhouston 4y ago
    I thought the CSGO2 smoke grenades are really compelling, like you did. That said, I think CSGO2 isn't fluid simulation at all. It sort of seems like animated sprits but with the ability to have dynamic holes punched in it or to dissipate. So it is a game entity that reacts to the world, which is nice, but I think there isn't much in the way of real "fluid simulation" in it at game time.
    - jameshart 4y ago
      It's more of a cellular automaton than anything else. Which is not a bad way of getting a deterministic fluid simulation, even if it is somewhat nonphysical.
  - ehnto 4y ago
    I think you are right in that it needs to be really tightly coupled to a gameplay mechanic, so much that it won't need to be a real simulation. The gameplay availabile through more realistic physics simulation has been relegated to novel indie games that can afford the gameplay challenges. I think the main issue is structure and containment. AAA games don't avoid destruction, newtonian physics and fluid dynamics because it's computationally hard but because it probably doesn't benefit the more standard gameplay mechanics enough for all the issues it causes. Just as an example, Red Faction came out in 2001 with full terrain destruction, but they had to preempt all the shortcuts people would take else you can just tunnel to the end of the level. Every time a GTA comes out the internet rumours say it'll have fully destructable buildings, but that will never happen as you would end up blowing up all the mission locations or breaking NPC pathfinding etc.
tehsauce 4y ago
“Differentiable programming” is nice term which covers the broad application of NN/deep learning techniques to special domains

albertzeyer 4y ago

> We use neural network ...

snickmy 4y ago
it was unclear to me what is the NN used for in this case?

jussipetteri 4y ago

It is supposed to be 15x faster than the alternatives. And based on my testing that figure might not be far off.

bhouston 4y ago
I suspect they are running at 1/3 the resolution and upscaling maybe? So the simulation is 27x faster roughly but there is an upscaling cost that brings it down to 15x?

raincole 4y ago

It's not clear to me either. But I guess it's something like a denoiser: you simulate it at a very low resolution and use AI to "upscale" it?

Culonavirus 4y ago
That's what it looks like. Still impressive. Even though personally I think we've had realtime fluid sims in games before and they always fail to make a splash by not being tied to gameplay properly. I think CSGO2's way of handling fluids (or at least smoke) is the way forward. Make it more than damn flipbooks, integrate it into gameplay, the rest will take care of itself in time. It just needs a spark. CSGO2 may be that spark.
- bhouston 4y ago
  I thought the CSGO2 smoke grenades are really compelling, like you did. That said, I think CSGO2 isn't fluid simulation at all. It sort of seems like animated sprits but with the ability to have dynamic holes punched in it or to dissipate. So it is a game entity that reacts to the world, which is nice, but I think there isn't much in the way of real "fluid simulation" in it at game time.
  - jameshart 4y ago
    It's more of a cellular automaton than anything else. Which is not a bad way of getting a deterministic fluid simulation, even if it is somewhat nonphysical.
- ehnto 4y ago
  I think you are right in that it needs to be really tightly coupled to a gameplay mechanic, so much that it won't need to be a real simulation. The gameplay availabile through more realistic physics simulation has been relegated to novel indie games that can afford the gameplay challenges. I think the main issue is structure and containment. AAA games don't avoid destruction, newtonian physics and fluid dynamics because it's computationally hard but because it probably doesn't benefit the more standard gameplay mechanics enough for all the issues it causes. Just as an example, Red Faction came out in 2001 with full terrain destruction, but they had to preempt all the shortcuts people would take else you can just tunnel to the end of the level. Every time a GTA comes out the internet rumours say it'll have fully destructable buildings, but that will never happen as you would end up blowing up all the mission locations or breaking NPC pathfinding etc.

tehsauce 4y ago

“Differentiable programming” is nice term which covers the broad application of NN/deep learning techniques to special domains