Show HN: Neural Particle Automata (selforg-npa.github.io)
71 points by esychology 9 hours ago
Neural CAs model self-organizing pattern formation on grids. Now the grid is gone. Each cell is an agentic particle that can move freely in space and change its state.
While each particle follows a simple shared rule, many together can grow complex morphologies or form intricate patterns. The resulting particle system as a whole can regenerate from damage and exhibits surprising emergent behavior.
Try cutting the lizard and watch it heal itself!
waerhert 4 hours ago
On the outside it looks very similar to what Michael Levin found on electrical communication between living cells. There too, the organism's cells were able to structure and repair their larger-scale morphology: https://www.youtube.com/watch?v=XheAMrS8Q1c
sixeyes 6 hours ago
Found it much interesting that i could mess up a pattern enough that it couldn't re-form.
Would be fun if selecting a new pattern didn't refresh the image as it is. Although maybe that's a requirement?
patcon 5 hours ago
Agree! This reminded me of a post that tweaked my brain a few months ago :)
https://open.substack.com/pub/defenderofthebasic/p/why-does-...
Also reminds me of Dr Michael Levin's work, which is living rent free in my brain lately
afrodisiac 8 hours ago
Super cool work!!! Do you think it would be possible to do something like cell division here?
treyd 5 hours ago
If you look at the texture demo with the zeros, it looks a bit like lipid membranes merging/splitting as they stabilize more or less around a particular size.
esychology 8 hours ago
Thanks! Yeah I think it should be possible though it requires making the cell division/splitting a differentiable operation. But nontheless, this is indeed a very interesting and promising direction to pursue.
skimmed 4 hours ago
Can someone tell me why cellular automata are suddenly everywhere? I've seen ~10 articles regarding them in the last month.
soraki_soladead 3 hours ago
Possibly because SIGGRAPH is coming up and these were papers submitted to that conference.
Enginerrrd 4 hours ago
Because the space of people interested in such things is relatively small and so a single article has knock on effects where a reader of the article or a blogger sees it and starts exploring the space and posts more about it, increasing the exposure some more.
mattdesl 7 hours ago
This is super cool, great work. Is there a video or demo of the 3D point cloud "gaussian splat" like experiments?
Jgoauh 5 hours ago
could something similar be used for texture synthesis ? of course the particles will need to be arranged in a grid and everything, or maybe recreate the texture by interpolating between the particles to exploit low contrast areas in the data
sva_ 4 hours ago
From the original research - self-organizing textures: https://distill.pub/selforg/2021/textures/
Jgoauh 3 hours ago
thanks ! i feel stupid for only checking out the linked paper lol
hamburgererror 4 hours ago
This is the future of scientific publishing, pdf is so boring.