Fluid/sand physics improvement, if possible to implement

First of all I want to declare that the youtube link I’ve posted underneath isn’t to promote the youtuber in any way it’s just to show what it can do and it’s how I came to this discussion in the first place. I wasn’t part of this research and don’t claim anything to be mine on this subject.

I feel like this could have potentially an impact on fluid physics if this isn’t already implemented in some way in the engine. I also don’t know about the computing power needed to run the algorithm or how much difference it would be to the current one used.

The main goal of this discussion is to propose a certain way of using a trained neural network to improve the behaviour of smoke/water/sand in the engine. The youtube video show’s the workings as described in this paper. The link contains the link to a website which under resources shows some code/the paper produced.

This youtube video show’s how the final product behaves and I was so amazed that I thought about doing this request. I don’t really care if it will get flagged or shutdown as this is probably just a silly request to make but I wouldn’t mind seeing something done with this in the near future if it’s possible.