ConwayLife.com

How can we use Bellman/similar to find outputs for the signal? Is it going to be serendipitous at most or is there some possibility to prepare a search?

Rhombic wrote:How can we use Bellman/similar to find outputs for the signal? Is it going to be serendipitous at most or is there some possibility to prepare a search?

I think we should insist first on knowing where the LoM is coming from, so we can constrain the search appropriately. It would be nice, once we had a LoM-to-something -- or better, a half-LoM-to-something that could be a workable signal splitter -- to be able to use it.

Oddly, even with a lot of development of elementary conduits in recent years, we have very few LoM-to-X and even fewer X-to-LoM converters -- and really there's not a lot of hope of sneaking input signals in to make any of those usable in practice.

We need some lucky conduit that throws an active reaction a good long way ahead of all of its catalysts, before it turns into a LoM. A LoM is kind of an awkward shape -- almost no matter where the active reaction comes from, it's going to overlap a big fraction of the LoM's reaction envelope. That's even true if you make the LoM by colliding a couple of gliders, which would be easy enough to arrange -- maybe a search should just start there.

Seems as if people have done the basic catalyst searches, with catalyst/catfind/ptbsearch (and maybe CatForce, and maybe Bellman too? I forget) -- placing eaters and blocks and so forth and looking for anything interesting that happens. Not a lot of luck yet.

Still, there's definitely still a lot of unsearched space out there. I'd be surprised if anyone has tried a two-transparent-blocks ptbsearch search with an input LoM, for example, or even a transparent-beehive search probably. All kinds of strange things can happen, with low probability, when you throw transparent objects into the mix.

I was considering the output of the 5c/9 signal, which, to the best of my knowledge, should not involve LoM (just like in the 2c/3 signal). The signal itself is extensible and finding a way to get a Herschel or a glider out of it could be interesting.
As for the input or also using the reduced catalyst for conduits, I do agree that there are surprisingly few known catalysts to interact with, or yield, LoM due to its awkward symmetry and 2D expansion.

Rhombic wrote:I was considering the output of the 5c/9 signal, which, to the best of my knowledge, should not involve LoM (just like in the 2c/3 signal). The signal itself is extensible and finding a way to get a Herschel or a glider out of it could be interesting.
As for the input or also using the reduced catalyst for conduits, I do agree that there are surprisingly few known catalysts to interact with, or yield, LoM due to its awkward symmetry and 2D expansion.

For the newest up-to-datest modern search utility, it might be worth giving Macbi's Logic Life Search a try.

It's difficult to know what to search for, though. We can't just say "any stable construct inside MxN that produces a glider when a 5c/9 signal comes in". Or rather, we can, but it involves running thousands of separate Logic Life searches. The last frame of each search would match the first frame, but with the addition of a single glider moving away from the search area. And that glider appearing at spacetime location (X, Y, T) is a totally separate search from the same glider appearing at (X, Y, T+1).

It seems pretty obvious that any one of these searches for an (X, Y, T) output glider has almost no chance of succeeding, at least for a search area small enough that there's any hope of Logic Life Search handling it. But if we run thousands of searches (or more), that improves the probability by that much anyway.

The question is what size of search area will allow one of these searches to be run to completion in a reasonable amount of time. I'm thinking 5c/9 might be too high a period, so that searches will either have an improbably small search area or they'll take too long to complete. We should probably get some experience with Logic Life by trying some simpler searches for things that we know exist, first. A good example might be the known terminator for a 2c/3 wire, that lets a signal out to delete a beehive or affect other external still lifes.

5c/9 wires and "fizzles" were all found with Dean Hickerson's 'dr' search utility, so it will be interesting to see if Logic Life can keep up with 'dr', and/or maybe even find new elbows and converters and so on.

dvgrn wrote:For the newest up-to-datest modern search utility, it might be worth giving Macbi's Logic Life Search a try.

I haven't done extensive benchmarking, but I'm not confident that LLS will be any faster than WLS for this sort of thing. Although I'm hoping that SAT technology advances and LLS overtakes WLS without me having to get out of bed . Of course everything you've said also applies to WLS searches. (Perhaps it's hard to automate lots of searches of different sizes in WLS? Whereas in LLS you could just do some Python scripting. I have written a program that tries to orchestrate lots of LLS searches by preprocessing the candidates for next-one-to-run and then guessing which one would run fastest based on its number of variables and the statistics from previous runs. At the moment the code isn't compatible with the current version of LLS, but I'm going to try to get it working again because I'd really like to run a systematic search for a c/2 agar crawler.)

Might it be easier to search for 5c/9-to-spark? And then have the spark hit a bait still life which goes into a conduit or whatever?

5c/9 wires and "fizzles" were all found with Dean Hickerson's 'dr' search utility, so it will be interesting to see if Logic Life can keep up with 'dr', and/or maybe even find new elbows and converters and so on.

Is dr also the program that found those eaters in the Golly pattern collection that are really slow at digesting? Those are really amazing and I'm not sure that I would be able to find them with LLS. Maybe the trick is to carefully constrain which cells you want to allow to change and which have to be in the stator?

Could one of the current scripts be modified to, say, "preserve the signal still life after n generations" (feasible already with Bellman) at the same time as "at generation n, there must be more than X live cells in area (x,y,widthx,widthy)"

Because in reality, as long as some kind of spark, H, B, glider, etc. appears in a given area far enough from the signal, it will be good enough to use in some way.
Incorporating this is probably very easy (just checking that at a given generation, a bounding rectangle contains at least a specified number of live cells).

This, if I'm not mistaken, avoids the problem of predicting where the output glider might be. And again, it seems very feasible to add a short function to the script to check if any cells are alive within a bounding rectangle. Thoughts?

Rhombic wrote:Could one of the current scripts be modified to, say, "preserve the signal still life after n generations" (feasible already with Bellman) at the same time as "at generation n, there must be more than X live cells in area (x,y,widthx,widthy)"

This is a really good idea and could certainly be implemented in LLS.

I wonder if the following even easier idea would work. Just demand that every cell alive in the original still life is also alive in generation n and n+1, and that there is also at least one additional live cell in generation n. The fact that the original cells are live in generation n and n+1 more-or-less forces them to have stabilized, and then the extra cell at generation n might well be a spark.

found by Scorbie (here).
I'm not sure if it can be saved in some way to be used, with potentially different catalysts.