Jason Cawley
Wolfram Science Group
Phoenix, AZ USA
Registered: Aug 2003
Posts: 712 
"A" goes to "A" is simple and symmetric enough that it does nothing interesting.
Something will be complicated enough to do something interesting.
Ergo, there is some continuum from dead simple to complicated, after which the behavior gets as complicated as you please.
The sole question is how soon it happens.
If one categorizes solely on the outcome, then one can call anything that produces a complex output, complex. But this is assuming a conclusion or playing with words.
If instead one categorizes on the form of the rule, then one can speak of simple rules leading to complex behavior or not doing so. The question simply cannot arise otherwise.
In the first third of the book, Wolfram strips away candidate after candidate for the supposed magic component that effortlessly produces arbitrary complexity from little. There isn't one  each special bit of 2 color CAs is removable, without the formal phenomena disappearing. How far you have to push to find it differs at the margin for register machines vs CAs, etc, that is all.
The specific hang up of this review seems to be symmetry. He notices that in the ECAs, the most complex behaviors come from the nonsymmetric rules, and that is true. It is clearly easier to arrive at a complex behavior nonsymmetrically, since symmetric anything effectively has fewer bits of distinguishable behavior to play with.
But the totalistic CAs have symmetry, and show all the complexity one could ask for, very quickly.
My personal favorite in impressive complexity combined with a symmetry I find beautiful, is a bit farther out in the rule variety tree. All the way out at 3 colors, and outer totalistic. You don't really have to go that far, and I'll give an ECA example below. But the result is to me prettier and the complexity more staggering, this extra bit out. If one isn't interested in finding an exact threshold or first, it is enough.
It is outer totalistic 3 color CA number 8640199.
JasonsRule = 8640199; Partition[IntegerDigits[JasonsRule,3,15],3]
Notice, all 3 colors occur 5 times in its rule table. The behavior for any given outer total is always symmetric between 0 and 2. Outer total 2 always leads to 2, ignoring the previous center cell value.
To appreciate it, start with at least 400 steps of evolution. (I've studied the center of the pattern to 50000 steps, and it stays spectacular).
ArrayPlot[ CellularAutomaton[{JasonsRule, {3, {3, 1, 3}}, {1}}, {{1}, 0}, 400, {All, All}], ColorRules > {0 > Yellow, 1 > Red, 2 > Black}, PixelConstrained > 2]
Symmetric? Yes. Makes it quite pretty, too. Simple? No.
Don't want to push to 3 colors or outer totalistic? OK. Look at rule 182, which appears to be simple. Now look at it with a periodic background, made of block {0,0,1,1,1}. Put a single 1 between tiles of that block.
ArrayPlot[ CellularAutomaton[ 182, {{1}, {0, 0, 1, 1, 1}}, {{1, 400, 2}, {400, 400, 2}}], PixelConstrained > 4]
Now, let's look at the effect of moving the "defect" 1 by phase one, aka the difference between a repeating background of {0,0,1,1,1} vs. one of {0,1,1,1,0}. Using the fact those two background are IntegerDigits[7,2,5] and IntegerDigits[14,2,5], you get this 
With[{data2 = CellularAutomaton[ 182, {{1}, IntegerDigits[#, 2, 5]}, {200, All}] & /@ {7, 14}}, ArrayPlot[data2[[1]]  data2[[2]] + 1, PixelConstrained > 2, ColorRules > {0 > Yellow, 1 > Red, 2 > Black}]]
Symmetric? Yes. Simple? No.
I hope this is interesting...
Report this post to a moderator  IP: Logged
