What are neural cellular automata?

COOL STUFF:

MUSIC:
Godmode: from youtube's music library. Not sure where to find that publicly.

TIMESTAMPS
(0:00) Intro
(0:40) Cellular Automata
(2:04) Neural Cellular Automata
(3:49) Filter + Convolution
(5:21) Activation Function
(7:13) Worms

[EmergentGarden]

I mentioned it briefly in the video, but didn't clarify it super well. The filter is inverted, meaning it's flipped over both it's x/y axis. This results in an inverted pattern, and this implementation is pretty common for convolutions. I may explain in a future video why I find this more intuitive, where a visual aid would be very helpful. Hope this clarifies any confusion!

[hyperteleXii]

The cells don't have to be arranged in a square grid. They often are, because that's an easy and efficient way to quantize space in computer memory; But they don't *have to* be. You could also have hexagons, or triangles, or indeed an irregular graph of random points.

[jonaw.]

This is the kind of stuff Computer Science should be about. Not creating the architecture for the newest dating algorithm. This is what the world needs right now.

[Milark]

I commented “this is the most underrated channel under 1000 subs”. And now you have 2.13k a few days later. Love to see it.

[AxisSage]

This just cracked open my mind to all kinds of new possibilities and ties right in to existing skills ive already developed. Thank you so much for the work you put into this video ❤

[Fybir_]

Your channel is a hidden gem, I just watch your videos on repeat in the background when programming

[o_enamuel]

i think your channel is going to grow pretty fast
it isnt common to see so good content so early

[flamencoprof]

Woah! this is another level above the cellular automata (and fractals) I was playing around with forty years ago on my Commodore 64, thanks to a column in Scientific American.
BASIC was too slow and I had to learn opcodes to do the actual cell processing, with operations directly on the screen memory.

[SpencerYonce]

These are the coolest videos I’ve ever seen. I’ve been coding shaders in glsl for a couple years now, and you just gave me a ton of ideas. Thank you

[micahconnor]

your program is fascinating, it almost feels dream-like. I'm having a lot of fun slightly tweaking numbers on the worm pattern to see all the new results. I feel like layers of the universe have been pulled back, and I'm getting to see how all the things function through relatively simple numbers!

[jyly]

I was litirally just watching all your videos and was just out of content lol. For my highschool project I'm also making a natural selection simulation. But your life engine is so much more interesting and made me wish I saw this earlier so I could make something more like your life engine. Anyways your videos are so interesting keep it up :).

[Diamcreeper]

It would be cool to see a part two about how the more complex cellular automata work, such as the self regenerating ones you've mentioned

[orenong]

Thank you so much! I had so much fun with it!! I made this function

float activation(float x) {
float y = 1./(1.+pow(2.7182, -x));
float z = ((x/y) + (y/x));
if (x<=0.5)
{return x/z; }

else{return z/x;}
}

once in 20 restarts or something I found another amazing thing with this function

[PunmasterSTP]

Rarely do I come across a video as inspiring as yours. I've dabbled in simulating a reaction-diffusion system as well as population dynamics, but this just makes me excited to explore even more! Thank you so much for making and sharing this video.

[immanuelt]

Fascinating stuff, please never stop.

[Barcd]

Wow this video blew me aways, very nicely produced

[tensenpark]

Love this video, impressive work!
Short comment though: This is not neural cellular automata. The neural part is of CA is TRAINING the Neural cellular automata so that the filters learn to replicate existing patterns. That's what's neural about it. In this video the filter values are a given. The way it's explained is more akin to implementing continuous cellular automata with activation functions.

[JellyMonster]

Excellent video - just the right pace for me and well explained.

[verbii-source]

I work on NCAs and this is an amazing video that gave me a much cooler intuition

[mayae]

I am considering using this for my master's thesis in music composition. Thank you for sharing your knowledge :D