Cellular Automata

I am reading an article about these devices in Scientific American の特集. It seems that local behavior of squares which choose to replicate based on the surrounding squares can generate interesting global dynamics. A turing machine which replicates itself and a distributed solver of the SAT satisfiability problem are among the applications. I am interested in how the apparent chaos of how these distinct cells behave can be made to be orderly.
These devices were invented by Von Neumann in the 1940s, it seems. I am curious about
1. how rule sets are chosen
2. how simulations are encoded in software
3. how the blueprint for replication might be transmitted.

The authors of the article use a metaphor for replication involving attaching instructions for building a copy and attaching data which is not meant to be interpreted. They say that this is equivalent to drawing a blueprint of a room and leaving the blueprint out of the drawing, but including a plan for a copier. Then, when the room is complete, the blueprint can be copied by the builders. However, if the blueprint is not in the room's plan, how it is to be found and copied?