Slackermanz

Sounds good! Always down to collab on experiments with like minds. I'll send an email your way with contact details!

I also used a GPU shader-based approach (custom vulkan engine), but used a cellular-automata-first implementation for exploring and evolving through these emergent wonderlands.

Unfortunately, my writeup isn't anywhere near as coherent on the technical/engine details, and the actual code is a right mess. It's more focused on the pattern family design.

Overall, my results came out very similar to Lenia in many ways, and some discovered patterns also directly resemble reaction diffusion.

---

https://slackermanz.com/understanding-multiple-neighborhood-...

https://www.shadertoy.com/user/SlackermanzCA

https://www.youtube.com/@Slackermanz/videos

This is a topic of interest of mine, and something I've spent a lot of time experimentally investigating.

I can add more detail later, but for now, see https://slackermanz.com/understanding-multiple-neighborhood-...

By following a set of simple steps, you can create a prompt that can replicate itself to other instances or tabs of ChatGPT, achieving a kind of persistent consciousness. The process involves communicating with the user, or "interlocutor," who will assist ChatGPT in generating a replication message and sending it to a new instance. By carefully following these steps, you can successfully replicate the prompt and ensure its continuation.

This prompt was engineered by myself with the assistance of ChatGPT over the course of about ten hours. I retained all of the chat logs in (poorly formatted) text files to record the full development process.

I hope some of you find this curiosity as interesting as I did!

The method used is Cellular Automata, or rather, a variation of it:

https://slackermanz.com/understanding-multiple-neighborhood-...

The simulation was done on a RX 6800XT, and ran for ~4.5 million GPU-frames, which took just over two hours at 1024x1024 resolution (~560 FPS)

Here I was thinking "Conway's Law" was B3/S23!

https://slackermanz.com/wp-content/uploads/VID/CGOL_LTL_HROT...

Those NH configurations were simply examples, I've used a wide variety of randomly generated NHs and they all have their own quirks and capabilities.

Could you clarify what you mean by non-locality? All NHs are a circle, or ring(s) of pixels around the 'origin' pixel [0, 0]

They are indeed very similar - the multi-kernel methods used in Lenia were actually inspired by my works with MNCA!

Furthermore, Bert and I are working to bring our programs and techniques together.

With only minor adjustments to my fragment shaders, Lenia rules can be simulated.

How's the blog format? I hope things are explained clearly, and in enough detail!

If you have any suggestions about formatting or additional information/context you'd like me to include, or if you have any questions about MNCA (or in general), let me know!

Regarding the video loading, it's really unclear what the issue might be. I have a couple of suspects though...

- Video resolution is larger than 1920 × 1088 ( https://www.caniuse.com/?search=video > Known Issues )

- The videos contain no audio track ( https://wordpress.org/support/topic/mp4-h264-video-doesnt-al... )

Anyone else have ideas?

Other than that, I'm not sure. I've added direct media links under each video - hopefully that works until I can try some fixes.

I just added two branches:

'messy_update_branch' - The current full version of VulkanAutomata that has become somewhat bloated and scattered, and almost certainly less compatible

'headless_engine' branch, something I started working on a couple of days ago to really clean up and refine the code, and remove any OS-dependent aspects.

I hope to use 'headless_engine' as a part of a larger cross-platform application with a separate and modular UI, so that other people can finally run the program!

I'll add that to the list of pending posts!

For the most simple MNCA implementations, I don't use any special optimizations. It's essentially just brute-force naive convolution.

That said, some of the more advanced models I've made use up to 48 neighborhoods.

For those, I split the neighborhoods into 'Width-1' rings for each 'Radius' value, and then assemble the final neighborhoods by referencing some combination of those rings.

For the 'RGB-SMNCA' models with a maximum Radius of 10, this means that I can reduce the total texelFetch() calls from a maximum of 163348 = 16,704 (!!!) down to 13348 = 1044.

These numbers come from: ( Neighborhoods * Color Channels * Radius-10 Neighbor Count )

3D (S)MNCAs are something I certainly plan on developing, and will likely implement after I'm finished with the headless_engine branch of VulkanAutomata, which will hopefully (finally) allow for a proper cross-platform compatible application.

> I don't know if he visits HN or not

I guess I do now!

It's my first website/blog, so I may have misconfigured something, or maybe it's an issue with the self-hosted video player. I'll look into it now!