Hacker News new | ask | show | jobs
by danwills 396 days ago
Yeah awesome I totally agree that the potential seems to be there for universality in RD systems like the one I posted, and also that stability is going to be absolutely key to the idea working at all! Thankfully Gray-Scott and similar systems have already been shown to be resistant to noise (there's an example in gollygang/Ready at Patterns/GrayScott1984/Lesmes_noisy.vti)

I also 100% agree about the wiring-problem when simulating things in 2D! However I've seen how moving-solitons behave a lot, and although they are probably more sensitive, I just can't believe that they are ultimately that different to GOL gliders.. and so I would hope that all that would be required to be able to cross wires would be making enough space, and some careful synchronization (offset modulo-synced clock states) to allow streams to cross without interfering!

Here's way too much soliton-fun I had ages ago in a video:

https://www.youtube.com/watch?v=Naj_J8aznyk&t=12s

I think that how solitons can behave when they are affected by both their own diffused-history and the wave-equation results of their past states, is just so fascinating!

In terms of whether non-local effects might come-in and ruin the completeness-party, most RDs have a diffusion-coefficient and I think there's probably some fairly good guarantees around rate-of-information-propagation there especially for RDs that don't use the wave-equation. The one I posted in my first reply does use wave, but it actually might not need to use it in order to be able to achieve its overall behavior! And even when using wave-equation, the max propagation-rate is still limited by the wave-diffusion coefficient!

I think there are several ideas in your comment that are absolutely getting to the heart of the matter: Building some kind of predictable 'cell' that can take moving-solitons in its 'input ports' and results in a logical result via moving-solitons exiting its 'output ports', all perfectly synchronized to work with coupled 'cells'. It does sound pretty hard! But I think that with a concerted effort, and with the potential frameworks that could be built-into gollygang/Ready with a dedicated coder who's interested in all of this, then some progress could definitely be made!