Summary and Some Directions for Future Research...

• CAs are
• capable of universal computation
• very easy to parallelize
• useful to model systems with finite states and only local interactions
  
  
  
• but, they are also
• very hard to program to do specific complex tasks (because programming is at such a low level, i.e. the rule table for individual cells)
  
  
  
• Much of the recent work on self-replication in CAs maybe of somewhat dubious significance, at least from the point of view of producing automata capable of open-ended evolution
  
  
• However... the local dynamics, scalability and ease of parallelisation are all very nice properties, and I might be using something like a CA for my future work...