If you could automate the search for something in GOL, what would it be?

[ngaylinn]

I'm building a tool to search over 64x64 GOL scenarios, and I'm looking for ideas about what to search for! What would you like to see?

I'm experimenting with a novel kind of genetic algorithm, and I'm using the Game of Life as its sandbox. I released a proof of concept, with some pretty basic fitness goals. It produced some cool and mesmerizing simulations, but I presume nothing particularly useful or notable to the GOL community. For round 2, I hope to make the search process much more versatile and powerful, and I'd love some more challenging goals to match.

But what to search for? I'm very curious what you think. Are there any classes of pattern that aren't thoroughly explored, or where some AI support / competition would be welcome? Any thoughts on what qualities make a GOL simulation interesting, exciting, or delightful? Any open theoretical questions that could be answered with a "proof by example"? Something totally different? If you have ideas, or advice on where to find them, I'd love to hear from you.

Thanks!
-- Nate Gaylinn

[Apersoma]

Puffers that puff gliders for synthesis

[hotdogPi]

We don't have good ways to find c/8, c/9, etc. spaceships except for the few that we know of (c/8 diagonal, c/10 orthogonal, and c/12 diagonal). I feel like the best way to find them is to find small partials that are one or two cells off, and when there are enough of them, rub them together to see if they fit and form something complete.

The denominator is what matters, so 3c/8 orthogonal, (3,1)c/8, etc. are also ones that could be searched for this way.

As an example of a small component, here's a c/13 diagonal partial that has to be supported:

Code: Select all

x = 7, y = 7, rule = B3/S23
3b2o$b2o2bo$obo2bo$5o2$5b2o$4b3o!

(This one has been checked with the existing search tools but not with what I described above.)